In Silico Ligand Based Pharmacophore Design of GRA and

11β-HSDI Inhibitors: An Approach to Database Mining and

Identification of New Lead Compounds

A Thesis

Submitted to

Banasthali University
For the award of degree of

DOCTOR OF PHILOSOPHY
in

(Pharmaceutical Chemistry)

Submitted By
Sucheta Das

Supervised By

Prof. Sarvesh Kumar Paliwal

Head, Department of Pharmacy

Banasthali University

Rajasthan- 304022 (India)

 Wxw|vtà|ÉÇ
This Thesis is dedicated to my Loving
Parents Mr. Dev Brat Das & Mrs.
Karuna Das; my Grandfathers Late
Pashupati Das & Late Ramesh Chandra
Das; my Uncle Late Mrinal Kanti Das
& my Aunty Late Gitanjali Das.
 Acknowledgement

“A Person can Achieve anything,

As long as
He has the Passion, the Drive, the Focus and the Support.”
It was a long road, but here I am standing on the verge of achieving my goal of
completing Doctoral Research Work and there are so many contributors to whom
thanks I extend.

Foremost, I would like to express my deep respect and thank to the almighty for his
grace and blessings and for giving me the strength, will and wisdom to carry out my
project successfully.

I would like to express my heartfelt gratitude to Prof. Sarvesh Kumar Paliwal,

Head Department of pharmacy, Banasthali University, who is not only my mentor
but always has been an inspiration. I could not be prouder of my academic roots and
hope that I can in turn pass on the research values and the dreams that he has given
to me.

I must offer my profoundest gratitude to Prof. Dharam Kishore, Dean and Secretary
Banasthali University, for providing me an opportunity to work and for his kind
cooperation and departmental facilities. I feel blessed in getting his kind guidance in
my life.

I wish to express my heartiest thanks to Prof. Aditya Shastri, Vice Chancellor,

Banasthali University, for providing me the opportunity and all necessary facilities
to accomplish this endeavor successfully.
I also owe a debt of appreciation and gratefully acknowledge the support of my
respected teachers, faculty of pharmacy, Dr. Swapnil Sharma, Dr. Manu Sharma,
Dr. Rajani Chauhan Dr. Divya Yadav, Dr. Rakesh Yadav and Mrs.Sumitra Nain,
for helping me in building up my concepts, valuable suggestions, constant
motivations and developing the ethics of my profession. I am highly thankful to all
laboratory assistants for their kind help in all phases of work.

My thesis would not have been completed without continuous contribution and
support of my lab mates and friends. I find immense pleasure in stating my cordial
thanks to my colleagues for their love, motivation and all the possible help they
could extend during my research period. I wish to extend my special thanks to Ms.
Aarti Singh Ms. Mukta Sharma & Ms. Khushbu Verma for their consistent
encouragement, constant support and friendly advice.

I am also thankful to Dr. Anupama Mittal, Dr. Anubhuti Pandey, Mrs. Anuradha
Derashri, Mrs.Richa Arya Thakur, Neetika Tripathi & Monika for their kind co-
operation.

Some special words to special persons and close friends like Ms. Vandana Singh
whose dedication, support and love reposed confidence in me to take the load off my
shoulders. I am thankful to her for all those light and cheerful moments I shared
with her.

None of this would have been possible without the love and patience of my loving
parent Mr. Dev Brat Das & Mrs. Karuna Das. They are always with me both in
my dreams and actions. Their life and teachings have taught me that excellence is
continuous journey and has no destination. I hope they will be very happy to see
that the dreams which they nurtured and the developmental inputs which they
provided have started taking the desired shape. In fact, I am short of words for their
contributions in shaping my life as they are the real architects of my life.
No one can ever thrive in life without the blessings of their elders in the family. I
bow my head with great respect before my grandfathers Late Pashupati Das & Late
Ramesh Chandra Das; my uncles Late Mrinal Kanti Das & Late Sakti Ranjan
Karar; my loving aunty Late Gitanjali Das. Though they are not today with me but
they had always been inspiration.

My deepest gratitude goes to my uncle Dr. Shyam Sundar Poddar for his unflagging
support and faith throughout my research work

I take this opportunity to express my heartiest thanks to my friends & my cousins

Mr. Vivekanad Kherwar, Mr. Somen Das, Mr. Arpan Das, Ms. Pamela Jana, Mr.
Rohit Jana, Mr. Sourav Poddar & Ms. Medha Poddar who have always stood by
me and given their moral support, no matter what.

I can never count the contribution of a special person Mr. Abhijit Naskar for his
unstinted support, whose good wishes and inspiration greatly helped me to complete
out this study with determination.

And at last I would like to thank all who had helped me directly and indirectly for
completing this work.

(Sucheta Das)
 Contents
Contents Page no.
List of Tables i-iii
List of Figures iv-vi
Chapter-1
Introduction
Diabetes Mellitus 1-41
Glucagon
Glucagon Receptor
Structure of the Human Glucagon Receptor (GCGR)
Glucagon Receptor Antagonists (GRA)
Peptide Antagonists
Non-peptide Antagonist
Monoclonal Antibodies (MAb)
Glucocorticoids
Cortisol: Major Role Player
11 Beta Hydroxysteroid Dehydogenase (11β HSD)
11 Beta Hydroxysteroid Dehydogenase type I (11β HSDI)
11β HSD1 Inhibitors
Drug Discovery
Drug Discovery Stages
Rational Drug Design
Computer Aided Drug Design(CADD)
Indirect Drug Design
QSAR
History of QSAR
QSAR Methodology
Major Parameters/Descriptors of 2D QSAR
Statistical Methods Used in QSAR Analysis
Evaluation of the QSAR Models
Ligand Based Pharmacophore (LBP) Modeling/ 3D QSAR
Pharmacophore Modeling
HypoGen
Cost Analysis in HypoGen
Validation of the model
Direct Drug Designing
Structure Based pharmacophore (SBP) modeling
Identification of Drug Target
Determination of Target Structure
Identification of Binding Site and Generation of
Pharmacophore
Molecular Docking
Search Algorithm
Scoring
Virtual Screening
Evaluation of Promising Lead Candidate
 Chapter-2
Literature Review
Glucagon Receptor Antagonists (GRA) 42-69
11β HSD1 Inhibitors
Problem Envisaged
Proposed Plan of Work
Scheme of Work
Chapter-3
Global Descriptor Based QSAR
Descriptor Based QSAR Approach 70-122
Materials And Methods
Series Selection
Data Set Preparation
Defining Substituents
Descriptor Calculation
Data Reduction
Training and Test Set
Statistical Analysis
Multivariate Linear Regression (MLR) Analysis
Partial Least Square (PLS) Analysis
Forward Feed Neural Network (FFNN Analysis)
Validation of Model
Results and Discussions
Series 1: β alanine, isoserine and thiazole
Series 2: Pyrrolidine Carboxamide Inhibitors of 11β HSD1
Series 3: Bicyclo[2.2.2] Octyltriazole Inhibitors of 11β
HSD1
Chapter-4
Ligand Based Pharmacophore (LBP) Modeling

3D QSAR Based Pharmacophore Modeling 123-167

Materials and Methods
Selection of Training Set and Test Set
Dataset Preparation and Pharmacophore Generation
Statistical Assessment of the Generated Hypotheses
Validation of Pharmacophore Model
Fischer’s Randomization Test
Internal Test Set Validation
External Test Set Validation
Result And Discussion
Series 1: β alanine, isoserine and thiazole Derivatives as
GRA
Series 2: Pyrrolidine Carboxamide as 11β HSD1 Inhibitors
Series 3: Bicyclo[2.2.2]Octyltriazole Derivatives as 11β
HSD1 Inhibitors
 Chapter-5
Virtual Screening & Structure Based Pharmacophore (SBP)
Modeling

Virtual Screening Based Identification of New Chemical 168-179

Entities
Virtual Screening
Material and Method
Pharmacophore Based Database Searching
Estimated Activity and Fit Values
Drug Likeness Screening
Result and Discussion
Series1 Pharmacophore Based Virtual Screening
Series 2 Pharmacophore Based Virtual Screening
Series 3 Pharmacophore Based Virtual Screening
Structure Based Pharmacophore Generation
Material and Method
Protein Structure Preparation
Binding Site Identification and Interaction Generation
Clustering and Pharmacophore Generation
Results and Discussions
Mapping of Clinical Trials Drugs (GRA)
Mapping of NCI Hits (GRA)
Mapping of Reference Compound (11β HSD1 Inhibitor)
Mapping of NCI Hits (11β HSD1 Inhibitor)
Conclusion
Chapter-6
Molecular Docking & Experimental Validation
Molecular Docking 180-190
Material and Methods
Target Identification
Receptor and Ligand Preparation
Receptor Ligand Docking
Results and Discussions
Molecular Docking of NCI Hits (GRA)
Molecular Docking of NCI Hits ( 11β HSD1 Inhibitors)
Biological Evaluation of Identified Lead Compounds
Experimental
Animals
STZ Diabetes Induction
Pre-Treatment Methodology
Post Treatment Methodology
Statistical Analysis
Results And Discussions
Pre Treatment
Post Treatment
Conclusion
Summary 191-196
References 197-231
Annexure-I
 List of Tables
S. No. Table no. Caption of tables
Selectivity and pharmacological effects of different 11β
1 Table 1.1
HSD1 inhibitors
2 Table 1.2 Lead discovery strategies
3 Table 1.3 Clinical drugs from CADD projects

4 Table 1.4 Type of modeling approaches in CADD

Lead structures and the substitution pattern of all three series
5 Table 3.1
used in QSAR study
Training and test set data for all three series under
6 Table 3.2
investigation
Structures and biological activities of 66 glucagon receptor
7 Table 3.3
antagonists
Correlation matrix of the independent variables used in the
8 Table 3.4
final model illustrating the degree of correlation
Statistical data of the independent variable illustrating the
9 Table 3.5
significance in terms of statistical parameters
Actual and predicted activity data obtained fromMLR, PLS
10 Table 3.6
and NN analysis of training set compounds
Actual and predicted activity data obtained from MLR, PLS
11 Table 3.7
and NN analysis of test set compounds
Correlation of biological activity of active and inactive
12 Table 3.8
molecules with Verloop L descriptors
Correlation of biological activity of active and inactive
13 Table 3.9
molecules with Dipole moment X component descriptors
Correlation of biological activity of active and inactive
14 Table 3.10
molecules with log P descriptors
Structures and biological activities of the 28 11β HSD1
15 Table 3.11 inhibitors

Correlation matrix of the independent variable used in the

16 Table 3.12
final model illustrating the degree of correlation
Statistical data of the independent variable illustrating the
17 Table 3.13
significance in terms of statistical parameters
Actual and predicted activity data obtained from MLR, PLS
18 Table 3.14
and NN analysis of training set compounds
Actual and predicted activity data obtained from MLR, PLS
19 Table 3.15
and NN analysis of test set compounds
Correlation of biological activity of active and inactive
20 Table 3.16
molecules with Dipole moment X component descriptors
Correlation of biological activity of active and inactive
21 Table 3.17
molecules with Log P descriptors
Structures and biological activities of the 32 11β HSD1
22 Table 3.18
inhibitors
23 Table 3.19 Correlation matrix of the independent variable used in the
i
 final model illustrating the degree of correlation
Statistical data of the independent variable illustrating the
24 Table 3.20
significance in terms of statistical parameters
Actual and predicted activity data obtained from MLR, PLS
25 Table 3.21
and NN analysis of training set compounds
Actual and predicted activity data obtained from MLR, PLS
26 Table 3.22
and NN analysis of test set compounds
Correlation of biological activity of active and inactive
27 Table 3.23
molecules with Inertia moment 1 length descriptors
Correlation of biological activity of active and inactive
28 Table 3.24
molecules with log P descriptors
Correlation of biological activity of active and inactive
29 Table 3.25
molecules with VAMP polarization XX descriptors
Training and test set data for all three series under
30 Table 4.1
investigation
Structure and biological activity of β-lanine, isoserine and
31 Table 4.2 thiazole core derivatives with glucagon receptor antagonistic
activity
Results of the top 10 pharmacophore hypotheses generated
32 Table 4.3
by the hypoGen algorithm
Actual and estimated values of the training set based on the
33 Table 4.4
pharmacophore hypothesis
Actual and estimated values of the test set based on the
34 Table 4.5 pharmacophore hypothesis

Structure and biological activity of Pyrrolidine carboxamide

35 Table 4.6
derivatives with 11β HSD1 inhibitory activity
Results of the top 10 pharmacophore hypotheses generated
36 Table 4.7
by the hypoGen algorithm
Actual and estimated values of the training set based on the
37 Table 4.8 pharmacophore hypothesis
Actual and estimated values of the test set based on the
38 Table 4.9
pharmacophore hypothesis
Structure and biological activity of
39 Table 4.10 Bicyclo[2.2.2]octyltriazole derivatives with 11β HSD1
inhibitory activity
Results of the top 10 pharmacophore hypotheses generated
40 Table 4.11
by the hypoGen algorithm
Actual and estimated values of the training set based on the
41 Table 4.12 pharmacophore hypothesis
Actual and estimated values of the test set based on the
42 Table 4.13
pharmacophore hypothesis
Number of compounds identified through pharmacophore
43 Table 5.1
based sequential virtual screening
NCI hits obtained from β alanine, isoserine and thiazole
44 Table 5.2
based pharmacophore
ii
 NCI hits obtained from pyrrolidine carboxamide based
45 Table 5.3
pharmacophore
NCI hits obtained from bicyclo[2.2.2] octyltriazole based
46 Table 5.4
pharmacophore
Pharmacophore mapping, fit value and lipinski’s violation of
47 Table 5.5
clinical trial drugs (BAY-27-9955 and LY2409021)
Pharmacophore mapping, fit value and lipinski’s violation of
48 Table 5.6
NCI retrieved hits (SKDGRA1, SKDGRA2 and SKDGRA3)
Pharmacophore mapping, fit value and lipinski’s violation of
49 Table 5.7
partial 11β HSD1 inhibitor (carbenoxolone)
Pharmacophore mapping, fit value and lipinski’s violation of
50 Table 5.8
NCI retrieved hits (SKDHSD1, SKDHSD2 and SKDHSD3)
Effect of SKDHSD2 and SKDHSD3 on blood glucose level
51 Table 6.1
of previously treated STZ induced albino wistar mice
Effect of SKDHSD2 and SKDHSD3 on blood glucose level
52 Table 6.2
of STZ induced albino wistar mice

iii
 List of Figures

S. No. Figure no. Caption of figures

1 Figure 1.1 Blood glucose level in normal and diabetic subjects

2 Figure 1.2 Major complications of Diabetes Mellitus
3 Figure 1.3 Physiological actions of Glucagon
4 Figure 1.4 Structure of mammalian proglucagon
5 Figure 1.5 Role of glucagon and glucagon receptor in liver
6 Figure 1.6 Structure of Human Glucagon Receptor
7 Figure 1.7 Glucagon receptor antagonists reached in clinical trials
Interconversion of cortisol to cortisone in presence of 11β HSD
8 Figure 1.8
isozymes
9 Figure 1.9 11β HSD1 and carbohydrate metabolism
10 Figure 1.10 11β Hydroxysteroid dehydrogenase type1 (11β-HSD1) inhibitors
11 Figure 1.11 The entire process of Drug Discovery
12 Figure 1.12 Phases of drug discovery and development
13 Figure 1.13 Steps involved in Drug Designing
14 Figure 2.1 Urea glucagon receptor antagonist
15 Figure 2.2 Structure of BI-32169 and its methyl ester
16 Figure 2.3 Novel human glucagon receptor antagonist
17 Figure 2.4 Conformationally constrained tri-substituted urea
18 Figure 2.5 Compound (S)-33a
19 Figure 2.6 Optimization of 1,3-oxazinan-2-one
20 Figure 2.7 Development of AZD8329 from AZD4017
21 Figure 2.8 Various steps involved in 2D QSAR model generation
22 Figure 2.9 Steps involved in ligand based pharmacophore modeling
23 Figure 2.10 Steps involved in virtual screening
24 Figure 2.11 Steps involved in structure based pharmacophore modeling
25 Figure 2.12 Steps involved in molecular docking
26 Figure 2.13 Steps involved in anti diabetic biological evaluation
Plot of actual activity versus estimated activity of training and test set
27 Figure 3.1
obtained from MLR analysis
Plot of actual activity versus estimated activity of training and test set
28 Figure 3.2
obtained from PLS analysis
Plot of actual activity versus estimated activity of training and test set
29 Figure 3.3
obtained from NN analysis
Dependency graph illustrating correlation between the Verloop L
30 Figure 3.4 (subst.3) used to train the neural network architecture versus the actual
activity data
Dependency graph illustrating correlation between the Dipole moment X
31 Figure 3.5 component (subst.2) used to train the neural network architecture versus
the actual activity data
Dependency graph illustrating correlation between the log P (whole
32 Figure 3.6 molecule) used to train the neural network architecture versus the actual
activity data

iv
33 Figure 3.7 Effect of various descriptors on different substituents of GRA
Plot of actual activity versus estimated activity of training and test set
34 Figure 3.8
obtained from MLR analysis
Plot of actual activity versus estimated activity of training and test set
35 Figure 3.9
obtained from PLS analysis
Plot of actual activity versus estimated activity of training and test set
36 Figure 3.10
obtained from NN analysis
Dependency graph illustrating correlation between the Dipole moment X
37 Figure 3.11 component (whole molecule) used to train the neural network
architecture versus the actual activity data
Dependency graph illustrating correlation between the log P (whole
38 Figure 3.12 molecule) used to train the neural network architecture versus the actual
activity data
Effect of various descriptors on the whole molecule of 11β HSD1
39 Figure 3.13
inhibitory activity
Plot of actual activity versus estimated activity of training and test set
40 Figure 3.14
obtained from MLR analysis
Plot of actual activity versus estimated activity of training and test set
41 Figure 3.15
obtained from PLS analysis
Plot of actual activity versus estimated activity of training and test set
42 Figure 3.16
obtained from NN analysis
Dependency graph illustrating correlation between the Inertia moment1
43 Figure 3.17 length (Subst.2) used to train the neural network architecture versus the
actual activity data
Dependency graph illustrating correlation between the log P (whole
44 Figure 3.18 molecule) used to train the neural network architecture versus the actual
activity data
Dependency graph illustrating correlation between the VAMP
45 Figure 3.19 Polarization XX (whole molecule) used to train the neural network
architecture versus the actual activity data
Effect of various descriptors on the whole molecule of 11β HSD1
46 Figure 3.20
inhibitory activity
47 Figure 4.1 Pharmacophore modeling workflow
Plot of correlation coefficient between actual and estimated activity
48 Figure 4.2
value of training set
Pharmacophore features (Hypothesis 1) and pharmacophore mappings of
49 Figure 4.3
GRA
Plot of correlation coefficient between actual and estimated activity
50 Figure 4.4 value test set

51 Figure 4.5 Graph of 99% Fischer’s randomization test of hypothesis1

Plot of correlation coefficient between actual and estimated activity
52 Figure 4.6
value of external test set
53 Figure 4.7 Pharmacophore mapping of clinical trial drugs
Plot of correlation coefficient between actual and estimated activity
54 Figure 4.8
value of training set

v
55 Figure 4.9 Pharmacophore features (Hypothesis 1)
56 Figure 4.10 Pharmacophore mapping with active compounds
57 Figure 4.11 Pharmacophore mapping with inactive compound no. 6
Plot of correlation coefficient between actual and estimated activity
58 Figure 4.12
value of test set
59 Figure 4.13 Graph of 99% Fischer’s randomization test of hypothesis1
Plot of correlation coefficient between actual and estimated activity
60 Figure 4.14
value of external test set
61 Figure 4.15 Pharmacophore mapping of carbenoxolone
Plot of correlation coefficient between actual and estimated activity
62 Figure 4.16
value of training set
Pharmacophore features (Hypothesis 1) and pharmacophore mappings of
63 Figure 4.17
11β HSD1 inhibitor
Plot of correlation coefficient between actual and estimated activity
64 Figure 4.18
value of test set
65 Figure 4.19 Graph of 99% Fischer’s randomization test of hypothesis1
Plot of correlation coefficient between actual and estimated activity
66 Figure 4.20
value of external test set
67 Figure 4.21 Pharmacophore mapping of carbenoxolone
Three dimensional structure of glucagon receptor and 11β HSD1
68 Figure 5.1
enzyme
69 Figure 5.2 The inter-feature distances of the glucagon receptor pharmacophore
70 Figure 5.3 The inter-feature distances of the 11β HSD1 enzyme pharmacophore
71 Figure 6.1 Crystal structure of human glucagon receptor (4L6R)
72 Figure 6.2 Crystal structure of human glucagon receptor (2BEL)
73 Figure 6.3 Binding pattern of NCI compounds SKDGRA1
74 Figure 6.4 Binding pattern of NCI compounds SKDGRA2
75 Figure 6.5 Binding pattern of NCI compounds SKDGRA3
76 Figure 6.6 Binding pattern of NCI compounds SKDHSD1
77 Figure 6.7 Binding pattern of NCI compounds SKDHSD2
78 Figure 6.8 Binding pattern of NCI compounds SKDHSD3
Effect of SKDHSD2 and SKDHSD3 on blood glucose level of
79 Figure 6.9
previously treated STZ induced albino wistar mice
Effect of SKDHSD2 and SKDHSD3 on blood glucose level of STZ
80 Figure 6.10
induced albino wistar mice

vi
CHAPTER- 1
Introduction
 CHAPTER-1
INTRODUCTION

Introduction

Diabetes Mellitus

Diabetes mellitus is a metabolic deformity of the body, commonly referred as

enhanced plasma glucose level. This metabolic condition is caused due to defect in
insulin secretion or insulin resistance. Diabetes mellitus can be broadly classified into
two major classes, Insulin Dependent Diabetes Mellitus (IDDM) commonly termed as
Type I diabetes mellitus and Non Insulin Dependent Diabetes Mellitus also known as
Type II diabetes mellitus. Type I is associated with an absolute deficiency in insulin
secretion caused by destruction of β cell mass in the pancreases.1 Type II is most
commonly occurring diabetes. It is mainly characterized by defect in insulin secretion
with a major role played by insulin resistance and an altered glucagon secretion in
response to glucose intake.2 It comprises 90% of all diagnosed cases of diabetes.3

Type II diabetes mellitus has been emerged as one of the most serious and costly
chronic metabolic disorder of the world.4 India in particular has become the leading
country in having the highest number of people suffering with type II diabetes
mellitus.5 The worldwide cases of diabetes are expected to increase by 366 million by
the end of 2030.6 Several studies have been reported which relates increase in type II
diabetes mellitus with lifestyle changes that have resulted in overweight, obesity, and
decreased physical activity levels. These changes in lifestyle affects genetic
predisposition and hence causes increase in insulin resistance and further progress to β
cell mass destruction, rise in glycemia in the non diabetic range.7

The metabolic response is found to be different for ingested carbohydrate in patients

suffering with type II diabetes when compared with people having normal plasma
glucose level also termed as glucose tolerance.8 A predictable increment in plasma
glucose level is expected after a meal in normal patients but in type II diabetes
mellitus patients, pronounced increment in plasma glucose level is recorded.9 WHO
has set a guideline for plasma glucose level to be achieved in patients with T2DM
(Figure 1.1).10

1
 CHAPTER-1
INTRODUCTION

Figure 1.1: Blood glucose level in normal and diabetic subjects

The multifaceted pathophysiology of T2DM results in imbalanced uptake, storage,

and disposal of ingested glucose in association of increased hepatic glucose
production and hyperglycemia. Diabetes tends to lead various other health
implications (Figure 1.2).11 Therefore, an alternative approach effective to reduce the
glycemic content with its control on weight, plasma pressure, lipids, cardiovascular
protection, and side effect profile, especially hypoglycemia is favored.

Figure 1.2: Major complications of Diabetes Mellitus

Glucagon

Glucagon is a polypeptide of 29 amino acid residues, a product of alpha cells located

at the fringe of the islets of Langerhans of pancreases. It is also located in the GIT and
intestinal regions of some species, and also in some special nervous tissues of the
brain.12 It is primarily released during the fasting state and acts contrary to the action
2
 CHAPTER-1
INTRODUCTION

of insulin. Glucagon is an integral component of glucose homeostasis due to its

involvement in glycogenolysis and gluconeogenesis and liberation of glucose in
liver.13 Glucagon is also effective in controlling the release of insulin through the
action of glucagon receptor expressed on pancreatic β cells.14 Other than its insulin
contradictory action, glucagon also imparts inotropic and chronotropic effects on the
heart, affects GIT emptying and gut motility of the stomach and growth hormone-
releasing activities (Figure 1.3).13

Figure 1.3: Physiological actions of Glucagon

The production of glucagon is the result of tissue specific post translation of its 160
amino acid precursor, proglucagon (Figure 1.4). Cleavage of proglucagon into
glucagon takes place in the presence of the enzyme prohormone convertase 2 which
also results in the formation of glicentinrelated pancreatic peptide (proglucagon 1-30),
and a major proglucagon fragment (proglucagon 72-158).14

3
 CHAPTER-1
INTRODUCTION

Figure 1.4: Structure of mammalian proglucagon

Glucagon level in a healthy individual varies with the glucose level in the plasma. It
reaches up to 40 pmol/l during fasting conditions when glucose concentration in the
plasma is reduced to 2-3 mmol/l and when the glucose level is elevated to 10-12
mmol/l, the glucagon level lowers down to 1-2pmol/l.15 Reduced level of glucose
generates an electric promising of Na+ and Ca2+ at pancreatic α- cells. This electrical
activity triggers Ca2+ signals to secrete glucagon. Increased glucose concentration
prevents this pathway.16 Apart from glucose concentration; there are certain amino
acids and some ANS activity which trigger the secretion of glucagon.17 The
glycogenolytic action of glucagon starts with binding of glucagon with glucagon
receptor.
Patient suffering with either type 1 diabetes or type II diabetes show increased plasma
glucagon concentration.18 This elevated plasma glucagon concentration is due to
increased basal hepatic glucose production.19 Thus, hyperglucagonemia is an
important factor of fasting hyperglycemia in diabetes. This effect is more pronounced
in the deficiency of insulin.20 Hence, diabetes is a result of insulin deficiency and
attenuated glucagon signaling.

Glucagon Receptor
Glucagon receptor is a class B GPCR. It is abundant in liver, adipose tissue, pancreas,
the gastrointestinal tract, heart, brain, and kidney.21 Abundance of glucagon receptor
in so many sites suggests its multiple roles beyond glucose homeostasis. Glucagon is
a 62-kDa polypeptide containing 4 N-linked oligosaccharide chains.22 Glucagon binds
to the glucagon receptor (GCGR) and stimulates adenyl cyclase pathway through
heterotrimeric Gs protein.23 The adenyl cyclase pathways raise circulating cAMP
levels and triggers PKA. Activated PKA in the liver is responsible of down streaming
4
 CHAPTER-1
INTRODUCTION

various biological systems like phosphoenolpyruvate carboxykinase, glucose-6-

phosphatase, fructose1,6-bisphosphatase and peroxisome proliferator-activated
receptor γ coactivator-1. These in turn suppress glycolysis and glycogen synthesis,
and as a result of which gluconeogenesis and glycogenolysis get activated. Glucagon
also activates the phospholipase C-inositol phosphate pathway in liver, inducing
intracellular Ca2+ signaling which is also responsible for causing glycogenolysis,
gluconeogenesis and urea synthesis.24 Figure 1.5 represents functions of glucagon and
glucagon receptor in liver.16

Figure 1.5: Role of glucagon and glucagon receptor in liver

Structure of the Human Glucagon Receptor (GCGR)

The GCGR is a part of class B (secretin like) family GPCRs. Class B GPCRs are
characterized with a globular N-terminal extracellular domain (ECD) having three
conserved disulphide bonds and a seven transmembrane domain. Hormonal peptides
trigger these class B GPCRs by binding to both the extracellular domain and the 7
transmembrane domains.25 Class B GPCRs is similar to class A (rhodopsin-like)
5
 CHAPTER-1
INTRODUCTION

GPCRs by having a seven transmembrane helical domain and similar signal

transduction mechanisms.26 The glucagon receptor has a number of distinct features
from known class A GPCRs in the transmembrane domain. The N-terminal end of
helix I in glucagon receptor extends three additional helical turns (approximately 16
Angstrom) above the extracellular (EC) membrane boundary from Lys 136 to Gly 125
(Fig. 1.6). This extended helical portion is the stalk and may be the glucagon
actions.27

Figure 1.6: Structure of Human Glucagon Receptor

Glucagon Receptor Antagonists (GRA)

The presence of higher glucagon concentration in diabetes clearly shows that
inhibition of glucagon receptor can be a lucrative approach for the treatment of
hyperglycemia. Three different types of glucagon receptor blockers have been tried
which includes peptide antagonists, non-peptidyl orally active antagonists and
monoclonal antibodies. But, only few candidates have reached to clinical trials
(Figure 1.7).28 Till date no drug candidate has reached the market. So, there is a strong
urge to focus on this area.

Peptide Antagonists
des-His-glucagon, structural analogue of glucagon, lacking only His1 was found to
have in vitro GRA activity, in 1972.29 In 1980, a glucagon analogue identified as [l-N

6
 CHAPTER-1
INTRODUCTION

α- trinitrophenylhistidine, 12-homoarginine]-glucagon (THG) was found to

antagonize in vitro hepatic adenylyl cyclase stimulation, and decrease plasma glucose
levels in diabetic rats. Later it was modified at His1 and Arg12, as a GRA.30 In 1987,
des-His1[Glu 9]glucagon amide was prepared, by eliminating His1, substitution of
Asp9 with Glu9 and adding a C-terminal amide. It was found to inhibit adenylyl
cyclase activity in rat liver membranes but with a deceased binding affinity to the
glucagon receptor.31 Further to this, a 19-amino acid bicyclic fungal peptide
(BI32169) isolated from Streptomyces sp., structurally different to glucagon showed
human GRA activity with an IC 50 of 440 nM. But, the pharmacokinetic and
pharmacodynamic properties of BI-32169 were not established and therefore it could
not be developed as a promising GRA.32

Non-Peptide Antagonist
In 1998 Novo Nordisk developed NNC-92-1687 as the first non-peptide human GRA.
However, it exhibited poor IC 50 of 20 μM.33 Merck came with triarylpyrrole a
promising, orally bioavailable, non-peptide GRA with an IC50 of 3.7 nM.34 Further to
this, Merck formulated substituted imidazoles exhibiting a favorable potency, IC 50
of 53 nM and high selectivity to the glucagon receptor.35 Pfizer came up with a fungal
bisanthroquinone isolated from Talaromyces wortmanni, named as Skyrin as a small
molecule GRA. Skyrin inhibited glucagon activated cAMP formation and
glycogenolysis in hepatic cells.36 An alkylidene hydrazide and its derivative
comprising an alkoxyaryl substitution as antagonists to the human glucagon receptor
were jointly developed by Pfizer, Novo Nordisk, and Anadys Pharmaceuticals.37
Bayer introduced Bay 27-9955 as the first GRA effective to reduce glucagon induced
hepatic glucose output and plasma glucose level elevation in humans. Bay 27-9955 is
a non-peptidyl, orally active, competitive antagonist of the human glucagon receptor
with an IC 50 value of 110 nM.38 Currently, Eli Lilly is engaged in clinical
development of LY 2409021 as a novel orally active GRA.39 Thus, these studies
strengthen the need to develop orally active GRAs as a promising treatment approach
for diabetes.

7
 CHAPTER-1
INTRODUCTION

Figure 1.7: Glucagon receptor antagonists reached in clinical trials

Monoclonal Antibodies (MAb)

Use of therapeutic MAb is well known in the treatment of autoimmune disorders.40
Thus, this approach has been tried to inhibit excess glucagon secretion. Certainly,
glucagon monoclonal antibody successfully reduced the glucagon excess effect in
rodents and minimized postprandial elevation in plasma glucose concentrations in the
existence of insulin in STZ treated diabetic rats.41 High affinity human MAb have
produced identical results in improving glucose homeostasis. Overall,
immunoneutralization of glucagon or its receptor is considered to be a promising
therapy to treat the excess glucagon condition in diabetes.42

Glucocorticoids
Glucocorticoids are stress hormones that regulate various biological actions involved
in metabolic, inflammatory, cardiovascular and behavioral processes. These are
secreted by the adrenal cortex under the regulation of hypothalamic-pituitary-adrenal
(HPA) axis. They impart their physiological actions in different target tissues by
binding with the glucocorticoid and the mineralcorticoid receptor. They are known as
the end-effector hormones of the hypothalamic-pituitary-adrenal axis.43

8
 CHAPTER-1
INTRODUCTION

Glucocorticoids control the energy expenditure of the body by shifting the energy
balance from anabolism to catabolism.44 They stimulate metabolism of glucose in the
liver and facilitate lipolysis in adipose tissues and protein degradation in various
organs and tissues by the process of gluconeogenesis.45 Glucocorticoids also stimulate
appetite in the central nervous system, by facilitating nutrients uptake.46 Thus, chronic
excess of circulating glucocorticoids or increase in the glucocorticoid sensitivity to
their target tissues is connected with development of visceral obesity, T2DM, and
hyperlipidemia.47 A high concentration of glucose and fatty acids affects β cell
sensitivity by triggering stress on endoplasmic reticulum, reactive oxidative species,
degenerated insulin synthesis and abnormal insulin signaling pathways.48

Cortisol: Major Role Player

Cortisol, is the main active glucocorticoid in humans. It is a secretion of adrenal gland
and can be found in its inactive form cortisone, in kidney. Its release from the adrenal
gland is controlled by circadian rhythm.49 The hypothalamic–pituitary axis (HPA)
controls cortisol secretion during increased level of stress.50 HPA triggers neuronal
signaling in the paraventricular nucleus in the hypothalamus.51 As a result of which
corticotrophin releasing hormone or factor (CRF) is secreted into the hypophyseal
portal circulation which acts on corticotrophs to secrete adrenocorticotropic hormone
(ACTH).52 ACTH further triggers the adrenal cortex to synthesize and liberate
glucocorticoids. The glucocorticoids actions on selective tissues are controlled by the
enzymatic activity of 11β HSD which reversibly converts inactive cortisone to active
cortisol and vice versa.53 Cortisol directly or indirectly triggers various insulin
synthesis and β cells signaling biological pathways.54 Cortisol triggers up-regulation
of β cell mass which stimulates the process of insulin synthesis and secretion which
causes β cells hypertrophy and hyperplasia.55 Cortisol may also affect glucose
metabolism genetically in β-cells through by modulating the gene expression causing
impaired cytoplasmic Ca2+ efficacy on the exocytotic process of insulin secretory
vehicles.56

11 Beta Hydroxysteroid Dehydogenase (11β HSD)

The binding of cortisol to glucocorticoid receptor and mineralocorticoid receptor is
regulated by 11β HSD isoenzymes.57 Both 11β HSD isozymes belong to short chain
alcohol dehydrogenase superfamily.58 11β HSD1 catalyzes the conversion of 11-keto
9
 CHAPTER-1
INTRODUCTION

metabolites into active glucocorticoids, i.e. cortisol in the presence of NADPH, while
11β HSD2 reverses the conversion in the presence of NAD (Figure 1.8).59 The
reversible inter conversion of cortisone into cortisol takes place in the endoplasmic
reticulum.60 11β HSD isoenzymes are genetically originated from different genes and
have distinct tissue distributions. 11β HSD1 is expressed in liver, adipose, kidney and
brain and 11β-HSD2 mainly in kidney and salivary glands.61

Figure 1.8: Interconversion of cortisol to cortisone in presence of 11β HSD isozymes

11 Beta Hydroxysteroid Dehydogenase Type I (11β HSDI)

11β HSD1 is a bidirectional enzyme responsible for both 11-oxo-reductase and
dehydrogenase reactions. It plays chief role in inter conversion of inactive cortisone
into active cortisol.62 It is microsomal in nature and initiates the pentose phosphate
pathway (Figure 1.9).63

Figure 1.9: 11β HSD1 and carbohydrate metabolism

10
 CHAPTER-1
INTRODUCTION

Functionally 11β HSD1 can be divided into four parts. The transmembrane domain,
which enables the enzyme to connect to the endoplasmic reticulum lumen, the
nucleotide binding domain, the catalytic site, and the C- terminal domain helps in the
oligomerisation of the enzyme. Ser170, Tyr183, Lys187 constitute the catalytic site,
and take part in the reaction.64 Increased 11β HSD1 based local cortisol production
raises hepatic glucose output and reduces glucose uptake in adipose tissue and skeletal
muscles. Apart from these, it triggers lipolysis, thus increases the free fatty acids, and
adverse glucocorticoids metabolic effects.65 Insulin inhibits 11β HSD1 expression
which shows dependency of isoenzymes activity on insulin resistance.66 11β HSD1
inhibitors block excess cortisol levels and thus results in less HGP.67

11β HSD1 Inhibitors

A diminished cortisol action in liver is speculated to be a promising therapeutic
approach to treat hyperglycemia. On the basis of this fact, 11β HSD1 has been
proposed as a novel treatment strategy for insulin resistance and type II diabetes
mellitus. 11β HSD1 inhibitors can be classified into several groups (Figure 1.10):
endogenous agents, exogenous natural products, and synthetic small heterocyclic
compounds.65
Endogenous agents include progesterone and its derivative (3α, 5α tetrahydro 11β
hydroxyprogesterone) and androgen derivative such as 3α, 5α tetra hydro 11β
hydroxytestosterone. These act as natural endogenous 11β HSD enzymes inhibitors,
but are not selective.68

Figure 1.10: 11β hydroxysteroid dehydrogenase type1(11β HSD1) inhibitors

11
 CHAPTER-1
INTRODUCTION

Exogenous natural product includes glycyrrhizic acid, a derivative of licorice, and its
hydrolytic product glycyrrhetinic acid 1 (GA). Licorice is a product of plant
Glycyrrhizia glabra. It is known for its medicinal properties for over a thousand years.
Based on 18β glycyrrhetinic acid, the hemisuccinate carbenoxolon 2 (CBX) was
prepared. Both compounds also promisingly inhibit 11β HSD2.69

Biovitrum was the first to report selective 11ß HSD1 inhibitors.70 Their
arylsulphonamidothiazole compounds efficiently inhibit 11ß HSD1 specifically both
in vitro and in vivo.71 BVT.2733 reduces in vivo plasma glucose and insulin, lowers
hepatic expression of PEPCK and glucose-6-phosphate mRNAs and thus reduces
hepatic glucose production.72
Merck’s compound 544 also found to be selective inhibitor of 11ß HSD1 activity. The
drug reduces body weight and appetite in diet-induced obesity. Analogous to previous
findings, compound 544 is effective in lowering fasting glucose, improving insulin
resistance, glucose tolerance and serum lipids in a type 2 diabetes mouse model.73
Table 1.1 summarizes selectivity and pharmacological effects of different 11β HSD1
inhibitors.

Table 1.1: Selectivity and pharmacological effects of different 11β HSD1 inhibitors

11β HSD1
Selectivity Carbohydrate Metabolism Disadvantage
inhibitors
No penetration to fat tissues
Increased Insulin sensitivity
Carbenoxolone Non selective Requires addition of
Reduced Hepatic gluconeogenesis
amiloride
Increased Insulin sensitivity
Compound 544 Increased glucose tolerance test
Selective No human studies
Merck Reduced fasting glucose
Reduced food intake
Increased Insulin sensitivity
BVT.2733 Reduced Hepatic gluconeogenesis No human studies
Selective
Biovitrum Reduced circulating glucose
Reduced circulating insulin

12
 CHAPTER-1
INTRODUCTION

Drug Discovery
Drug discovery is the process of introducing newer and safe drugs to the population in
order to improve the therapeutic value and eliminates the side effects of already
present drugs. Discovering any new drug needs a keen understanding of the disease at
the molecular level. Genomics, proteomics and computational approaches have made
it easier to understand a disease. This understanding leads to rise of “targets,” to
which promising new drugs might be able to act on. This whole process from
understanding the disease to launching a successful drug candidate in the market;
takes an average of 10-15 years. In fact out of 5,000-10,000 compounds entering the
research and development (R&D) pipeline, only one compound succeeds to get
approval as a drug candidate The whole process can cost $800 million to $1 billion.
This cost includes cost of several pitfalls while discovering the drugs.74 Figure 1.11
describes the entire process of drug discovery. The journey of drug discovery does not
end with the introduction of the drug in to the market. The post-market survey and
pharmacovigilance studies may eliminate the drug out of the market.75 Hence, there is
an urge of a careful planning even before the project is underway.

Figure 1.11: The entire process of Drug Discovery

Recent technologies can predict the possible outcomes with the help of computers and
which make this process less risky. This will help in minimizing the efforts in clueless
directions.

13
 CHAPTER-1
INTRODUCTION

Drug Discovery Stages

The whole drug discovery process can be classified into four major discovery stages
and four clinical stages (Figure 1.12)

Figure 1.12: Phases of Drug Discovery and Development

D0 Stage: Selection of Therapeutic Area

The process of drug discovery starts with the selection of therapeutic area of interest.
Criteria to select therapeutic area include:76
• Medical needs required in comparison to expected therapy.
• Possible scientific phenomena.
• Number of affected population.
• History of the therapeutic area.

DI Stage: Selection of Biological Target

The very next step of drug discovery process after selecting the therapeutic area is
selecting the appropriate therapeutic or biological target. The human body functions
are normally controlled by the enzymes, hormones and endogenous molecules
binding to their respective receptors. A disorder generally arises due to abnormal
functioning of these enzymes, hormones or these endogenous molecules. Thus, these
abnormal behaviors make them the biological targets for drug discovery. They
include:76
• Receptors
• Enzymes
• Signaling pathways
• Genes

14
 CHAPTER-1
INTRODUCTION

DII Stage: Validation of Biological Targets

After identification of a target, validation of that target is carried out to confirm
correct target identification. This validation can be done with the use of animal
models, gene targeting and expression techniques. Validation of biological target
helps us to identify any other secondary target that may show affinity with the drug
and may show unwanted physiological response. An ideal drug should bind
specifically to the target, which seldom happens. This binding of drug to other target
can be minimized up to certain extent.77 For example most of the biological targets
are G-protein coupled receptors (GCPRs) to which a drug binds. But, expression of
each GPCR is different to each other. Thus, by measuring the mRNA expression on
the receptor with the help of quantitative polymerase chain reaction (qPCR) analysis,
this problem can be solved.78

DIII Stage: Lead Discovery

The process of drug discovery actually starts with the discovery of the lead
compound. A lead compound has all the basic physical/chemical features necessary to
impart the desired pharmacological action. A lead compound can be further developed
to produce a new chemical entity with improved biological activity and
pharmacokinetic properties. A lead compound can be selected by several approaches.
The very first requirement of any lead discovery approach is assay of compounds for
a particular biological activity. It determines the activity of a compound.79
High throughput screening is the most common technique to identify the lead
compound from the pool of various chemical entities or naturally derived
compounds.80 This high-throughput screening approaches, is capable to screen
100,000 compounds in a day. The combinatorial chemistry is a modern approach of
lead discovery which can supply huge database of compounds in a short period of
time. Table 1.2 describes various lead discovery strategies. Compound screening or
lead finding can also be carried out by electrospray ionization mass spectrometry,
NMR spectrometry, virtual screening, informatics, pharmacaphore mapping and high
throughput docking.81

15
 CHAPTER-1
INTRODUCTION

Table 1.2: Lead discovery strategies

Description
Screening Strategies
Large numbers of compounds analyzed in a assay generally designed to run in
High throughput screening
plates of 384 wells and above

Compounds previously identified as hitting specific classes of targets (e.g.

Focused screen
kinases) and compounds with similar structures

Soak small compounds into crystals to obtain compounds with low µM

Fragment screen
activity which can then be used as building blocks for larger molecules

Structural aided
Use of crystal structures to design molecules
drug design

Docking models: interrogation of a virtual compound library with the X-ray

Virtual screen structure of the protein or, if have a known ligand, as a base to develop further
compounds on.

A tissue-based approach for determination of the effects of a drug at the tissue

Physiological screen
rather than the cellular or sub cellular level, for example, muscle contractility

Screen small compounds (fragments) by soaking into protein targets of known

crystal or NMR structure to look for hits with low µM activity which can then
NMR screen
be used as building blocks for
larger molecules

D IV Stage: Lead Optimization

Lead optimization is the final phase of the drug discovery after identification of the
lead structure. This phase includes optimization of properties of the lead compounds
discovered in the lead discovery phase. This step needs careful study of the structure
activity relationship of the lead compounds and synthesis of their derivatives, to
broaden the scope of the desired biological activity. There are various approaches
which are employed to optimize a lead compound in order to have the desired
pharmacological activity with minimal side effects.

Identification of Active Part (Pharmacophore)

A drug molecule consists of a pharmacodynamic part and a pharmacokinetic part. The
pharmacodymamic region is responsible for the pharmacological action where as the
pharmacokinetic region is for ADME of the drug. These regions are the
characteristics of the functional groups present in the drug. These functional groups

16
 CHAPTER-1
INTRODUCTION

are responsible for imparting specific property and nature to the drug. Representation
of these specific functional groups along with their interatomic distances is known as
pharmacophore. The functional groups or atoms which interfere with the binding of
the lead compounds with the receptors are known as auxophores. Out of all those
functional groups present in the lead compound some may be interfering with the
binding of the pharmacophore, and such functional groups need to be optimized.

Functional Group Optimization

The pharmacological activity of a drug depends on its structure which is due to the
contribution of its functional groups. The functional group of any compound governs
lipophilicity and electronic features of the compound. Thus, by optimizing proper
functional group, the drug distribution pattern and side effects can be optimized.

Structure – Activity Relationship Studies

Structure activity relationship (SAR) study is based on the effect of chemical
constitution of a drug on its physiological action. The activity and potency of a drug is
susceptible to small changes in its chemical structure. The motive of SAR studies is
to synthesize as many analogs as possible of the lead compound and analyzing their
effect of structure on activity. Conclusions are drawn based on the SAR studies of the
lead compound and its various analogues.

Homologation
A group of compounds generally differing by methylene group is termed as a
homologous series. A small change in the substituent can affect the polarity, alter the
pKa, & change the electronic properties of a molecule. Optimization of homologous
series is one of method used method to induce the optimization in the lead compound.

Cyclization of side chain

Over simplified structure of drug molecule is sometimes responsible for increased
side effects & reduced activity or selectivity. Conformational flexibility and effect of
substituents are two important governing factors for enhancing binding affinity that
bring additional binding interaction. Transformation of the alkyl side chain into cyclic
analogs may bring change in potency or in the activity spectra.

17
 CHAPTER-1
INTRODUCTION

Bioisosterism
Certain substituents or groups having similar physical or chemical properties and
hence similar biological pattern are known as isosters. Isosteric groups have same
number of electrons in their outermost electron shell. Bioisosteric replacement may
affect various physiological actions of the lead compound, for example molecular
size, steric shape, electron distribution, lipid solubility, water solubility, the pKa, the
chemical reactivity to cell compartments, capacity to undergo H- bonding.82

Preclinical Studies
Preclinical study is meant to check the safety profile of the lead compound. A newly
developed molecule has to clear various pharmacological and toxicological tests
based on animal studies before its testing on the human body. Preclinical phase
includes elucidation of the pharmacokinetics and pharmacodynamic profile of the
molecule by depicting its mode of action. However, priority is given to the animal
study data, which roughly estimate the possible adverse reactions and side effects
during treatment. These preclinical studies are performed in two ways, in vitro studies
and in vivo studies. The in vitro studies are carried out on distinct cell-lines and tissue
preparations. The in vivo studies include use of live animals and then changes in the
their behavior is analyzed.83

Clinical Trials
Preclinical studies are followed by the clinical studies, which involve evaluation of
the lead compound in the human volunteers. This stage helps to analyze the toxic
effects in the human body which cannot be analyzed in animals. But, further to this
stage an “Investigational New Drug (IND)” application should be filed in order to get
FDA approval based on the preclinical data. Clinical trials is manly consists of three
stages, stage I, stage II and stage III. There is one more stage 4 study which is carried
out post marketing of the drug.

Stage I Clinical Trial

Stage I clinical studies include study of effect of drug on healthy humans
volunteers and on a small population. This stage measures the safety, tolerability,
pharmacokinetics and pharmacodynamics of the lead compound.

18
 CHAPTER-1
INTRODUCTION

Stage II Clinical Trial

Stage II studies are target oriented. It involves study of the lead molecule on a
disease affected population. In this stage, the drug’s efficacy and safety,
metabolism and pharmacokinetics are measured in a diseased human body.
Stage III Clinical Trial
Stage III is a confirmatory stage of stage II. It involves study of all the
parameters of stage II. Stage III studies are widespread multiple site studies. This
stage involves study of lead molecule on a large group of diseased population.
This stage takes 3-6 years to complete. After completion of stage III, “NEW
DRUG APPLICATION (NDA)” application should be filed.
Stage IV Clinical Trial
Stage IV is also known as post marketing surveillance. It gets started, after the
drug has entered into the market. The main objective of this stage is to look for
any possible adverse or serious reactions, missed during the earlier stages and
might arise later. In such cases, the company stops the production of the drug and
draws back the drug from the market.84

Rational Drug Design

Main objective of ‘drug design’ is to develop a drug with a desired degree of
chemotherapeutic index and specific action. Basically, it is a study of the drug and the
biological receptor or enzyme interaction. Drug design explains whole journey of a
drug from the time it is synthesized, enters into the body, imparts desired
pharmacological effect, metabolize and at the end excretes out. It also explains the
relationship between the pharmacological effects and the chemical nature of the drug.
The ‘drug design’ in a wide spectrum deals with various time consuming and labor
extensive processes which starts with the stages of drug discovery and ends with a
precise design of a drug (Figure 1.13). The problem of excessive time and cost is
somehow reduced by the rise of computer software used for drug designing and
understanding the biological targets.

19
 CHAPTER-1
INTRODUCTION

Figure 1.13: Steps involved in Drug Designing

Computer Aided Drug Designing (CADD)

Computer assisted drug design (CADD) is an emerging technology that makes use of
knowledge of the structural and physical/chemical aspects of the receptor or ligands
interaction to identify pharmacophores or aid in the design of molecules.85 It increases
the efficiency of the drug discovery process. Computers have become a vital
contributor in the drug design process and have a large number of applications, which
include structure analysis, structure comparisons, lead compound design,
identification of active conformations and pharmacophores, combinatorial library
design, protein and binding site structure, ligand binding, QSAR studies etc. CADD
was founded in the early 1970s with its application in modifying the biological
activity of insulin by altering its structure and in designing the preparation of human
hemoglobin ligands.86 Table 1.3 enlists some marketed drugs which has been
designed by CADD.
The basic theory of CADD relies in the fact that molecular and electronic structure
underlines all physical and chemical properties of the molecule, including their
biological activity.

20
 CHAPTER-1
INTRODUCTION

Table 1.3: Computer aided drug designed marketed drugs

Year Generic Name Brand Name Manufacturer Against/Imhibits

1989 Zanamivir Relenza GlaxoSmithKline Neuraminidase
1997 Nelfinavir Aviracept Hoffmann-La Roche HIV protease
1998 Raltitrexed Tomudex AstraZenecca Thymidylate
1999 Amprenavir Agenerase GlaxoSmithKline HIV protease
2007 Raltegravir Isentress Merck HIV integrase

The physical/chemical properties correlate the reactivity and interaction of the

molecule with other molecules. There are two modeling approaches in CADD, which
is used for designing of new candidate molecules; these are direct and indirect design
(Table 1.4).87

Table 1.4: Type of modeling approaches in CADD

Direct design Indirect design

Accurate
Required for both the ligand and protein Required for ligand only
Structures

Minimum, simple and intuitive if not

Information Precise
very accurate

Requires multiple scoring functions, less

Predictivity Highly predictive (within the SSS).
predictive.

Variable selection, consistency with

Concerns Pose versus score correlation
the active site

As good as the training set, so proper

De Novo, more reliable when screening
Reliability emphasis on its design, limitations to
against diverse molecules
the diversity of the database to be used.

Indirect Drug Design

The indirect drug designing is also known as ligand based approach. The indirect
approach consists of subjecting a large number of the molecules (with known
potency) of varied structures to the computerized molecular modeling and then on the
basis of mathematical calculation 2D and 3D QSAR models are generated. These
models are then used to optimize the structures for the required biological activity.
This approach is applicable in the absence of receptor structure. The main objective of
(LBDD) drug design is to apply the structure and conformation of already synthesized

21
 CHAPTER-1
INTRODUCTION

molecules having some desired biological activity onto the conformation of a lead
compound, with the objective of improved activity or drug like properties.88 A series
of compound having diverse structure and activity is prerequisite feature of the
methodology. For this approach the lowest possible conformers of the rigid
compounds are generated and superimposed on them. Conformations of the flexible
compounds are searched by applying distance constraints of the rigid compounds.
Finally, all the types of compound structures are aligned to develop a pharmacophore
model. This model may be used to identify and synthesize novel compounds with
specific functional groups at specific positions. The most common tool in a LBDD is
2D and 3D QSAR of which the widely used method is CoMFA, HypoGen and
Hiphop.

QSAR
QSAR studies have become a boon for medicinal chemist in modern chemistry and
biochemistry. It speeds up the process of drug designing by saving the money and
time. The basic theory behind QSAR states that there is a connection between the
physical/chemical properties of a compound with their biological activity i.e.
biological activity is a function of structure of the compounds and an equation can be
deduced by relating the biological activity to the physical/chemical parameters.89
These physical/chemical parameters, which relates to hydrophobicity, topology,
electronic properties, and steric effects, are calculated by computer softwares. The
correlation between the chosen parameters and the structure-activity helps in selecting
the promising chemical entities to be synthesized and evaluated in the laboratory.
Activities used in QSAR include in vitro biochemical assays. The most applicable
approach of QSAR is that it helps in designing new chemical entities based on in vivo
situation.
The linear equation for QSAR is depicted below-

BA = Const + (C1 * P1) + (C2 * P2) + (C3 * P3) +...+ (Cn * Pn) Eq. (1)

Where, P1, Pn are the parameters of all the compounds of the series and C1, Cn are
coefficients of the parameters and the biological activity.

22
 CHAPTER-1
INTRODUCTION

History of QSAR
Crum-Brown and Fraser were the first to hypothesize the relationship between the
physiological actions of a substance as a function of its chemical composition and
constitution.90 They postulated the first generalized equation suggesting that
physiological activity Φ of a series of quaternized strychnine derivatives depends on
constitution (c), and suggested the following mathematical equation 2.
Φ= f(c) Eq. (2)
In 1893, Richet found inverse correlations between the toxicities of simple
compounds (alcohol, ether) and water solubility.91 In the late 19th century, Meyer and
Overton showed a relationship between the narcotic actions of organic compounds
with their oil/water partition coefficients.92 In 1939, Ferguson published a
thermodynamic equation to show the correlation between depressant actions of
volatile compounds and their relative saturation.93
In the era of physical organic chemistry, Hammett, introduced to the "sigma-rho"
culture, based on the substituent effects on organic reactions.94 Taft introduced the
first steric parameter, Es and developed a technique to isolate polar, steric, and
resonance effects.95 In 1962, Hansch and Muir correlated SAR of plant hormones to
their Hammett constants and hydrophobicity dependency.96 Hansch and Fujita came
up with the Hansch’s equation by combining hydrophobic constants with Hammett's
electronic constants and Taft’s steric constants.97
Log 1/C = aσ + bπ + ck Eq. (3)

Following are the most common approaches of QSAR

1. Hansch Analysis (The Extra Thermo Dynamic Approach).
2. The Free-Wilson (Additivity Approach).
3. Combined Approach.
Hansch’s Analysis (The Extra Thermodynamic Approach)
In 1962, Hansch suggested that the drug acts on the physiological system in two ways.
First, the journey of drug from the oral cavity to the target site and the second is its
interaction with the target site.98
Log (1/Co) = log p -ΔG/ RT +A Eq. (4)
Activity Transport Interaction

23
 CHAPTER-1
INTRODUCTION

He suggested the linear (Equation.5) and parabolic or non-linear dependence

(Equation.6) of biological activity (BA) on different parameters.
log BA = a log P + bσ + cEs + d Eq. (5)
log BA = a log P + b (log p)2 + cσ + dEs + e Eq. (6)
The coefficients (a, b, c, d, e) are calculated by multiple regression analysis. Though
the drug transport process and drug receptor interactions are complex in nature, yet
they are essentially physical/chemical and can be factored into electronic,
hydrophobic and steric parameters. The variations in biological activity depend upon
changes in these physical/chemical parameters like hydrophobic (π), electronic (E)
and steric (S) and can mathematically be expressed by equation98 (7)
BA = a (hydrophobic) + b (electronic) + c (steric) + constant Eq. (7)

The Free-Wilson (Additivity Approach)99

The Free-Wilson approach is based on additive property of mathematics in which
particular substituent at a specific position is presumed to affect the biological activity
of a compound of a series in additive and constant manner. It is assumed that a
particular functional group at a particular substitution site of the compound always
produces identical quantitative effects on biological activity of the whole compound,
as expressed by the equation (8):

log BA = µ+ Σ (a1.a2) Eq.(8)

Where, a1 = no. of position at which substitution occurs

a2 = no. of the substituents at that position
µ = overall average
The equation is solved by multiple linear regression using the presence or absence of
the different substituent as independent dummy parameter, while the measured
activity serves as a dependent variable. The Free Wilson analysis has one major
advantage over Hansch analysis and i.e., a QSAR model development is based on the
pharmacological activity values and the structures of the compounds are required.
Free Wilson analysis has several pitfalls:
• At least two distinct chemically modified substitutions must be present;
• New arrangement of substituents which are already present in the analysis can be
predicted.

24
 CHAPTER-1
INTRODUCTION

• Single point determinations (i.e. the single occurence of a certain structural feature
in the whole data set) obscure the statistical results;
• Substituents are described by using many degrees of freedom.

Combined Approach100
The two approaches, Hansch and Free-Wilson, are practically interrelated. Free-
Wilson approach suffers by its inability to attribute physical/chemical significance to
the substituent contribution and fails to yield good correlations in the cases where
parabolic model of Hansch is applicable. Likewise, sometimes it is difficult to
correlate the biological activity only in terms of known physical/chemical parameters,
which has necessitated the use of an indicator variables (I) which constitutes a part of
correlation equation so the most commonly used equations now contain a Hansch
parameters are used and of indicator variables based on the assumption of Free-
Wilson as described in equation (9) contains Hansch parameters and Free-Wilson
assumption.
BA = ∑ai, + ∑cj j + constant Eq. 9
The term ai denotes the contribution for each ith substituent, whereas j is any
physicochemical property of a substituent Xj.

QSAR Methodology
Approach 1: Based on Dimensions of the Descriptors/Models
QSAR methods can be classified on the basis of the structural representation or the
way of deriving the descriptor values101:
1. 0D QSAR characterizes molecular formula e.g. molecular weight, number and
type of atoms etc.
2. 1D QSAR only reflects a composition of a compound. This technique includes
correlation of activity with global molecular properties like pKa, log P etc. It
can be used as a tool for independent virtual screening.102
3. 2D QSAR correlates biological activity to the chemical features like
connectivity indices, 2D-pharmacophores etc. These models characterize only
the topology of the molecule. 2D QSAR helps in finding the necessary
features required for the biological activity of the compounds.103
4. 3D QSAR provides full information of structural composition, topology, and
spatial shape of molecule for one conformer only. These models are
25
 CHAPTER-1
INTRODUCTION

widespread. Through the 3D QSAR we can easily understand the orientation

of structure in space, effect of other forces, distance between the molecules.104
5. 4D QSAR is developed by Hopfinger and others. It is based on the
conformational information as the fourth dimension additionally including
ensemble of ligand configurations in 3D QSAR. These models differ to 3D
QSAR in considering the structural information for a group of conformers,
instead of one fixed conformation.105
6. 5D QSAR characterizes different induced-fit models in 4D QSAR.
7. 6D QSAR characterizes solvation models in 5D QSAR.

Approach 2: Chemometric methods

QSAR methods can be categorized on the basis of the correlation technique employed
to determine the connection between structural features and biological activity into
following two methods:
1. Linear methods including linear regression (LR), multiple linear regression
(MLR), partial least-squares (PLS), and principal component
analysis/regression (PCA/ PCR).106
2. Non-linear methods as in artificial neural networks (ANN), k-nearest
neighbors (kNN), and Bayesian neural nets.107

Major Parameters/Descriptors of 2D QSAR

Descriptors determine the intermolecular forces involved in drug-receptor
interactions. There are three major descriptors, which are electronic, hydrophobic, and
steric. Apart from these there are some other molecular descriptors as topological
descriptors. These descriptors are based on the 2D representation of a chemical
structure. 108
1. Electronic i.e. total dipole moments, HOMO and LUMO energies.
2. Hydrophobic parameters- log P, lipole
3. Steric i.e. molecular surface area, Verloop, Molar refractivity.
4. Topological parameters i.e. Wiener, Balaban, Randic, Connectivity indices,
shape flexibility, kappa, electrotopological state (E-state) indices.

Electronic Parameters
Electronic features of compounds are closely linked to their chemical and biological
activities. The electronic need of a reaction is decided by its mechanism. The
26
 CHAPTER-1
INTRODUCTION

substituent groups involved in the reaction and its effect on the reaction rates helps in
determining the mechanism of reaction. The first and still most widely used electronic
substituent parameter σ was developed by Hammett in 1935 on the basis of ionization
constants for benzoic acid derivatives.109 Hammett’s equation encompasses solution
effects like hydrogen bonding, dipole interactions etc.110 But, Hammet’s equation has
certain pitfalls which are as follows:111
1. The Hammett relationship only applies to systems where the substituents are
attached to the reaction centre via aromatic rings and are situated meta or para.
2. Ortho- substituted aromatic rings cannot be calculated due to steric effects.
3. Aliphatic acids do not follow the equation, because of steric hindrance in the
transition state and their flexibility may affect correlation in negative manner.
4. The substituent constant (σ) for meta- and para-substituents in benzene
derivatives as defined, on the basis of the ionization constant of substituted
benzoic acid in water at 25 oC. ie. X=H.
5. σ is a collective measure of the total electronic effects of a substitutent
[resonance and inductive effects].
6. σ does not depend on the nature of the reaction site.

Total Dipole Moment

Total dipole moment is the vector sum of bond dipole moments for each
compound.112 Dipole moment is a partial charge dependent parameter calculated on
the basis of center of charge over the substitution as the origin. The total dipole
moment represents the electrostatic interaction between the ligand and the target. It
can be calculated using the following formula.

μ = eΣrіqі Eq. (10)

Where rі is the distance of the ith atom from the origin and qі is the atomic charge of
the ith atom. Total dipole moment is calculated in Debye units. It is often correlated
with the polarity of a compound.113

HOMO and LUMO

HOMO (highest occupied molecular orbital) and LUMO (lowest unoccupied
molecular orbital) are the quantum chemical descriptors.114 These energy orbital
desrciptors control various chemical reactions and determine electronic band gaps in
solids; formation of many charge transfer complexes. Frontier molecular orbital
27
 CHAPTER-1
INTRODUCTION

theory (FMO) of a chemical reaction states that HOMO and LUMO of reactants
interact to form a transition state.115
The HOMO energy is analogous to the ionization promising and correlates with the
likeliness of the electrophilic attack on the molecules. The LUMO energy is
analogous to the electron affinity and correlates with the likeliness of the nucleophilic
attack on the molecules. Both HOMO and the LUMO energies govern radical based
reactions. Strong electrophilic or nucleophilic natures of the molecules are governed
by the relative energies of the HOMO/LUMO orbitals. Strong nucleophiles have a
low HOMO energy and vice versa; strong electrophiles have a high LUMO energy
and vice versa.116
The gap between two orbitals, i.e. the energy difference between the HOMO and
LUMO, is an important indicator of stability. Larger the gap higher is the stability and
hence less reactive is the molecule.

Steric Parameter
Quantitative determination of steric effects is quite an intriguing process, particularly
at the molecular level. However, steric effects are major role player in the drug-
receptor interactions and also in cellular transport systems. Taft’s ES constant is the
first steric parameter of QSAR.117

ES = log(kX/kH)A Eq.(11)
Where, kX and kH are the rates of acid hydrolysis of esters, X=CH2COOR and
CH3COOR
Taft quantified the steric (spatial) effects using the hydrolysis of esters: Here, the size
of R limits the reaction kinetics by obstructing the nucleophilic attack of water.
ES = log kRCO2CH2-log kCH2CO2CH2 Eq. (12)
=log (kRCO2CH2/ kCH2CO2CH2)
In this case, the steric effects are quantified by the Taft parameter Es: k is the rate
constant for ester hydrolysis.

Molecular Surface Area

Molecular surface area is the outer surface area of the volumes of the solute molecule
present in a solution. It is based on the Van der Walls molecular surface which
consists of small gaps and cervices, inaccessible to the solvent molecules present in

28
 CHAPTER-1
INTRODUCTION

solution. The molecular surface area is determined by removing these gaps and
cervices.118

Verloop
Verloop parameters are comprised of multi-dimensional steric descriptors defined in a
box for characterizing the shape and volume of the substituent. The shape and volume
of a substituent plays a critical role in deciding the steric effects produced by the
substituent in the ligand receptor interaction. The Verloop parameters B1-B5
represent the width of the substituent in the direction perpendicular to the length of
the substituent.119

Molar Refractivity (MR)

Pauling and Pressman proposed a parameter and equation which relates the dispersion
forces of the haptens-antibodies complex. It is evaluated by calculating the refractive
index, (n), the molecular weight, (MW) and the density of a crystal, (d).

MR (n2 -1) (MW) = (n2 -1)/d Eq (13)

Larger the molecular weight of the compound, larger is the steric effect, while larger
is the density, the smaller will be the steric. A smaller molar refrativity for a similar
molecular weight indicates stronger interactions in the crystal and hence stronger
dispersion interactions with the receptor.120

Hydrophobic Parameter
The absorption and distribution of any compound inside the biological systems is
determined by its hydrophilic or hydrophobic properties, which are represented by
partition coefficient (P). It is defined by:
P= [drug]octanol Eq. (14)
[drug]water

The hydrophoicity of a substituent is calculated by the lipophilicity coefficient π or

from the partition coefficients. Partition coefficient is the natural log of the
octanol/water partition coefficient (Log P), which is the relative affinity of a
compound for an aqueous or lipid medium and is closely related to the environment of
the biological system.121

29
 CHAPTER-1
INTRODUCTION

Log P
Log P is hydrophobic parameter. It explains the effect of the substituents on the
hydrophobicity of the main compound.122 To understand the effect of relative
hydrophobicity of substituents the hydrophobicity substituent constant, p was
introduced. The parameter p has been defined analogous to the Hammett parameter

p = log PX - log PH Eq. (15)

where PX and PH are the partition coefficients of the substituted and H substituted
compounds respectively. A higher value of p means higher affinity for the octanol
stage and vice versa.123

Topological Indices
Topological indices are techniques to convert chemical constitution of the compounds
into numerical values which further can be correlated with physical/chemical
properties and for QSAR studies. It represents the compounds in molecular graphs,
where the atoms are denoted by dots (vertices) linked to each other by lines (edges),
which are chemical bonds.124 With the help of this molecular graphs, certain path
lengths can be determined. A specific path length (m) is the distance between two
atoms connected with m bonds.
Topological indices are easy to calculate on the molecular two dimensional structures
only, thus ditching the need of conformational analysis of the compounds.
Topological indices suffer the drawback in their complexity in interpretation.

Wiener Number
Wiener number was introduced in 1947. It is so named after the inventor Harry
Wiener (1924–1998). His life and his achievements first in chemistry and later in
medicine have been well–presented by Rouvray.125
Wiener number is the sum of all the distances between the carbon atoms. The smaller
this number, the larger is the compactness of the molecule. It can be calculated by
multiplying the number of carbon atoms on one side of any bond by those on the other
side; W is the sum of these two values for all bonds. W can also be obtained by
simply adding all the elements of the graph distance matrix above the main diagonal.

30
 CHAPTER-1
INTRODUCTION

Balaban
The Balaban index was proposed by A. T. Balaban which also called the average
distance-sum connectivity or J index. It appears to be a very useful molecular
descriptor with attractive properties.126
The Balaban index is a graph index defined for a graph G with vertex-set V (G) and
edge-set E(G), dG(u, v) denotes the distance between vertices u and v in G, and
DG(u) = Σ dG(u, v) is the distance sum of vertex u in G, i. e., the row
vεV(G) Eq. (16)
sum of distance matrix of G corresponding to u. The Balaban index of G is defined as
J(G) =m/μ + 1 Σ 1/√DG (u)DG(v) Eq. (17)
uvεE(G)

Where, m is the number of edges and μ is the cyclomatic number of G, respectively.

Randic
Randic introduced molecular connectivity Randic index (c) to characterize the
molecular branching. Its quantitation is based on the degree Di of the vertex i in the
hydrogen-suppressed molecular graph. The Di is number of bonds from the atom
(vertex) i to non-hydrogen atoms.

Connectivity Indices
The connectivity index order k is derived from the adjacency matrix and is normally
written as, kχt, The order k is between 0 and 4 and is the number of connected non-
hydrogen atoms which appear in a given sub-structure.127
k
nt
k
χt= Σ 1/2
[Пδi ]- Eq. (18)
j-1 iεSj

In above equation, δi is the number of simple (i.e., sigma) bonds of the atom i to on
hydrogen atoms, Sj represents the jth sub-structure of order k and type t, and knt is the
total number of sub-graphs of order k and type t that can be identified in the molecular
structure. The types used are path (p), cluster (c), and path-cluster (pc).

31
 CHAPTER-1
INTRODUCTION

Molecular Shape Indices (Kappa)

The molecular shape indices are the techniques to quantify the molecular structure of
the compound provided that attributes of molecular shape are encoded into three
indices (Kappa values).128 Kappa values are calculated from counts of one-bond, two-
bond and three bond fragments relative to fragment counts of the reference structures
having a maximum and minimum values. It begins with the hydrogen-suppressed
skeleton of the compound. Bond fragments are calculated, IP, 2P and 3P which are
further used in calculating the Kappa indices lx, 2 1C, and 31C.

Electrotopological State (E-State) Indices

The electrotopological state index (E-state) is the topological index at atomic level. It
came into existence in 1990 and is computed as a graph invariant for individual atom
in the molecular graph.129 The E-state index represents the electron density and the
accessibility of electrons which participate in non-covalent intermolecular
interactions. It also characterizes the structural configuration of the surrounding
neighbors of the atom.130 The E-state indices are defined in terms of the ∂i and ∂vi
values of an atom i similarly to the connectivity indices.131 The E-state index for an
atom i (Si, also called atom-level E-state index) is composed of an intrinsic state term
(Ii), plus a sum of perturbations (∆Iij) from all other atoms in the molecule:

Si = Ii + ∑j ∆Iij Eq. (19)

where the summation is over the remaining atoms in the molecule. The intrinsic state
(Ii) and the perturbation (∆Iij) terms are calculated as follows:

Ii = ((2 / Ni) 2 * ∂iv+ 1)/ ∂I Eq. (20)

where Ni is the principal quantum number of valence electrons (the row, at which the
chemical element is placed in the periodic table)

∆Iij = (Ii – Ij)/rij2 Eq. (21)

Where, rij is the topological distance between atoms i and j, equal to the number of
atoms in the shortest path between them.

The quantification of E-state indices for atoms (such as >C<, >N–, =O, –Cl) in a
molecule, and for every hydride group (such as –CH3, >NH, –OH) is possible. The

32
 CHAPTER-1
INTRODUCTION

atom-type E-state indices have recently been introduced as an extension of the E-state
indices. These indices are the integration of the E-state values of each atom type. E-
state indices calculation of hydrogen atoms for deriving hydrogen atom-type is also
possible. Atom-type E-state indices determine the nature of electron accessibility for
the atoms of same types. Thus, the E-state values describe the feasibility of a
molecule to take part in a non-covalent intermolecular interaction.

Statistical Methods Used In QSAR Analysis132

Statistical methods are essential components of QSAR analysis. They help in building
and estimating a model's predictive abilities and in validation of an already existing
model. Thus they play a major role in establishing relationship and co-relationship
among variables and activities.

Linear Regression Methods

It is a mathematical procedure, which co-relates dependent (X) variable with the
independent (Y) variables. There can be different forms of regression analysis:
• Simple Linear Regression Analysis
An independent variable is correlated with a dependent variable to produce a
linear one-term equation. It helps in the discovery of some of the important
descriptors.
• Multiple Linear Regression Analysis (MLR)
More than one independent variable is correlated with dependent variable and a
single multi term equation is formed. The number of descriptors should be kept
less than one-fifth of the training set. Smaller the number of descriptors, more
robust and biologically relevant will be the QSAR model. The model creates a
linear relationship. In regression analysis, conditional mean of dependant variable
Y depends on (descriptors) X.
• Stepwise Linear Regression Analysis
This is useful when number of independent variables is very high and it thus
correlates independent variables in a stepwise manner with the dependent variable
to produce a multi term linear equation.
• Partial Least Square (PLS)
The PLS is a robust multivariate generalized regression method used to
summarize multitudes of promising collinear variables. PLS is an iterative
33
 CHAPTER-1
INTRODUCTION

procedure effective when the numbers of descriptors are more than the number of
compounds modeled. Normally, it is calculated by the leave one out cross
validation procedure.

Non Linear Methods

• Artificial neural networks (ANN): Artificial neural network explores nonlinear
relationships between the dependent and independent variables without using
an explicit mathematical function.133

Evaluation of the QSAR Models 134

1. Correlation Coefficient (r): The correlation coefficient ‘r’ assesses the quality of
the fit of the model. It depends on the overall variance of the dependent variable, i.e.
the biological activity. The correlation coefficient ‘r’ of any two subsets is small in
comparison to the correlation coefficient of a combined set, due to increment in the
overall variance
r =√1- ΣΔ2/Syy Eq. (22)

Where, Syy is the overall variance i.e Syy =Σ (yobs -ymean)

High value of regression coefficient (r>0.9) indicates the statistical significance of the
regression equation is high. While the low value of r indicates that the substituents
constant is not important in describing the variation under consideration.
2. Square of the Correlation Coefficient (r2): It can be envisioned as the fraction of
total variance in the data, which is illustrated by the regression model.
r2 =l-ΣΔ2/Syy Eq. (23)

Where, Syy=Σ (yobs -ymean)

ΣΔ2= Σ (yobs –y cal) 2 Eq. (24)

Where ‘yobs’ is observed biological activities, ‘ymean’ is mean of biological activities

value ‘ycalc’ is calculated biological activity used in the equation.
3. Standard Deviation (SD): This is a measure of dispersion or scatter of the
observation from the mean and indicates the ability of the QSAR analysis derived
function in predicting the observed biological activity. The value of SD depends on
the number of object n and the number of variable k. Therefore, SD depends on the

34
 CHAPTER-1
INTRODUCTION

quality of fit and the number of degrees of freedom. The smaller the value of SD the
better is the QSAR.
DF=n-k-1. Eq. (25)
2
SD= (yobs –y cal) / n-k-1 Eq. (26)
4. F-value: It calculates the statistical significance of the regression model, the
influence of the number of variables included in the model is larger than the standard
deviation.
F-value= r2 (n-k-1)/k (1-r2) Eq. (27)
5. Predictive residual sum of squares (PRESS): The sum of overall compounds, of
the square difference between the actual and the predicted value dependent variables.
P=Σ(yobs-y calc) 2 Eq. (28)
6. Outliers: An outlier is defined as a data point with a residual greater than two
times the standard deviation of residual.
7. Cross Validated r2 (q2): This is squared correlation coefficient generated using the
equation during a validation procedure.
r2Cv = SD-PRESS/SD Eq. (29)
Where SD is standard deviation
A cross-validated r2 is smaller than the overall r2. It is used to evaluate the
predictibility of an equation.

8. Jack-knife: The jackknifed residual, also called standardized residual, is calculated

as the ordinary residual in prediction divided by the residual standard deviation.

Ligand Based Pharmacophore (LBP) Modeling/ 3D QSAR

Pharmacophore Modeling
A pharmacophore is a 3D spatial display of chemical features of a compound
necessary for interaction with the target. The pharmacophores are used as 3D queries
in searching compound databases containing “druglike” chemical entities to discover
promising lead compounds or in mapping a new compound onto a pharmacophore.
This approach has huge potential in the field of drug designing in minimizing the cost
of drug discovery process and identifying new chemical entities. A number of
softwares can be used for pharmacophore generation like Catalyst. Accelrys (initially
BioCAD) in 1992, established pharmacophore generation in Catalyst as a technique to
recognize the automated pharmacophore pattern in a series of compounds on the basis

35
 CHAPTER-1
INTRODUCTION

of chemical features correlated with three-dimensional structure and biological

activity data.135
The software Catalyst came up with two types of hypothesis:
1. Chemical feature based hypothesis (HipHop)
2. Structure activity based hypothesis (HypoGen)
The hypothesis generation methods (HipHop and HypoGen) of the Catalyst software
is widely accepted methods of drug discovery research and toxicology.136

HypoGen
The HypoGen is similar to HipHop in using the same features but is a quantitative
hypothesis. The pharmacophore generation by HypoGen is divided into three
stages.137
• Constructive Stage constitutes generation of pharmacophore models in which the
most active compounds fit.
• Subtractive Stage removes the pharmacophore models common to the inactive
compounds.
• Optimization Stage randomly modifies the generated pharmacophore models by
moving features, rotating vectors, or adding or removing features. The best
modification having the lowest error cost values are kept and finalized as best
hypothesis. The cost values are the number of binary bits to generate
pharmacophore hypotheses. Lower the cost values, stable the hypothesis.

Cost Analysis in HypoGen137

Cost analysis is a technique to select best hypotheses out of all the generated
pharmacophore hypotheses. Three cost factors, weight cost, error cost, and
configuration cost constitute the total cost for each hypothesis. There are two
theoretical costs, the Null and Fixed cost, which determine the significance of the
generated hypotheses. The total cost value should be in between these two theoretical
cost values. If closer to the Fixed cost value and away from the Null cost value, the
hypothesis is statistically significant. All the cost values are expressed in bits, and
follow the principle of Occam’s razor.

36
 CHAPTER-1
INTRODUCTION

Error Cost138
The error cost is directly linked to the RMS difference. It increases with the increase
in root mean-square (RMS) difference between the estimated and the actual affinity
for the training set. It calculates the quality of prediction of the model.
Weight Cost138
The weight cost relates to the actual and ideal weight of the features. It is directly
proportional to the difference between the actual and ideal weights of the features.
The ideal value of the weight is 2. Weight values more than 2 causes fitting of
unrealistic conformations of the compounds to the features.
The error and the weight components are estimated using the probability distribution
shown in the equation 30

Eq. (30)

The error and weight components of the hypothesis cost also be estimated by taking -
lnP of as shown in eq. 31

Eq. (31)

Where, sigma and delta are constant parameters derived from expected values of
weights and errors, and x is the deviation from the expected values for each. The
sigma values are log (uncertainty) and 0.302 for the error and weight components,
respectively. The delta values are log (1.1) and 0.2 for the error and weight
components. The deviation for the errors is the actual minus the predicted scaled
activity values for the training set members, and the deviation for the weights is the
difference between the average of the feature weights and the ideal weight.
Configuration Cost138
The configuration cost is a fixed cost representing the complexity of the hypothesis
space to be optimized. It is expressed as log2 ‘P’, where P is the number of hypothesis
generated during the constructive stage and existed in the subtractive stage. The cost
value should not cross the value of 17. A configuration value of more than 17 leads to
a chance correlation as Catalyst software has the ability to form 217 hypotheses only.

37
 CHAPTER-1
INTRODUCTION

Hypothesis having a true correlation is represented by a high cost difference between

the total cost and the null cost. A cost difference of more than 60, there represents a
true correlation. A cost difference of 40-60 bits represents a chance of 75-90% good
correlation. Lower the cost difference values lower is possibility of having a true
correlation.
Total Cost138
It is sum of the aforementioned three cost components multiplied by a coefficient
(default coefficient is 1.0 for each) as shown in equation 32.

Cost = eE+wW+cC Eq. (32)

Where, e, w, and c are the coefficients associated with the error (E), weight (W), and
configuration (C) components, respectively.
Fixed Cost138
The fixed cost represents the ideal model fitting all the data perfectly i.e. all
compounds lay on a single line of slope. It is a sum of lowest possible error cost and
weight cost and the constant configuration cost

Fixed Cost = eE(x =0) +wW(x=0) +cC Eq. (33)

Null Cost (Upper bound cost)138

The null cost is representation of a zero hypothesis having no features. It has zero
contribution of the weight and configuration costs.

Null Cost = eE(x=x) Eq. (34)

In other words

• Fixed Cost = X + [entropy cost]

• Null Cost = X + [null error]
• The difference between the Fixed and Null Costs is the balance between the
entropy and null error.
• The entropy is a free energy relationship and is directly proportional to
negative logarithm of the activity. The difference between these two costs
states the statistical significance of the hypothesis.

38
 CHAPTER-1
INTRODUCTION

Validation of the Model

Cat-Scramble Validation
The Cat-Scramble validation procedure which is a cross-validation technique based
on Fischer’s randomization test is frequently employed to check the validity of the
model.139 This validation method confirms a strong correlation between the structures
of the compound with their biological activities. This is in consequence of
randomization of the activity data of the training set compounds, which generates
pharmacophore hypotheses using the same features and parameters to develop the
original pharmacophore model.140 A test set is employed in order to check the
predictive power of the hypothesis.
Internal Test Set Prediction
The ability of the models to predict the biological activity of compounds outside the
model development is a common method of validation.
External Test Set Prediction
The generated pharmacophore can also be validated by employing a structurally
diverse external test set.

Direct Drug Design

Direct drug design or structure-based design has emerged as a powerful tool in
medicinal chemistry. The concept of structure-based drug discovery is an admixture
of several fields: X ray crystallography and NMR, molecular modeling, synthetic
organic chemistry, QSAR, and biological evaluation. Initial binding site information
is gained from the three-dimensional solution of a complex of the target with a lead
compounds. Molecular modeling is usually next applied with the intent of designing a
more specific ligands with higher affinity. Synthesis and subsequent in vitro
biological evaluation of the new agents produces more candidates for crystallographic
or NMR analysis, with the hope of correlating the biological action with precise
structural information. It makes good sense at the early stages of design to use lead
molecule structural scaffolds that retain low toxicity profiles, given that the latter
most often derails successful drug discovery.141

39
 CHAPTER-1
INTRODUCTION

Structure Based Pharmacophore (SBP) Modeling

Identification of Drug Target
A drug target is a receptor or an enzyme whose functioning or deformation regulates a
disease condition to happen. Such drug target can be identified by understanding the
structural biology of those targets. Function prediction, pathway information, disease
associations, and structural data need to be considered when choosing a drug target.

Determination of Target Structure

To understand the structural biology of the target, accurate structural information is
required. The structure of the target entails the functioning of that target. The
structural data can be obtained by following three techniques.

1. X-ray crystallography (XRC)

2. Nuclear Magnetic Spectroscopy (NMR)
3. Homology Modeling or Comparative Modeling
X-ray crystallography (XRC) is a diffraction technique of crystals. It is a main source
of determining the three dimensional structures of the proteins.142 A three dimensional
structure of protein provides co-ordinates of the atoms which are further used by the
computer to build the protein structure. Nuclear magnetic resonance (NMR) has its
application in using magnetic fields and electromagnetic radiation to determine the
structure of organic compounds. It is widely used for characterizing the receptor
structure.143 Homology modeling determines the approximate structure of the protein
on the basis of available sequence, with at least 30% sequence identity.

Identification of Binding Site and Generation of Pharmacophore

A ligand binds with the target protein structure at a specific site known as binding site
or active site. Identification of the active site is needed after obtaining the structure of
the target. A binding site is a hollow cavity-like structure having hydrogen bond
donor/acceptor and hydrophobic features. On the basis of the binding site
environment a pharmacophore model is created and on the basis of that model new
compounds are designed.

Molecular Docking
The fate of a successful drug candidate depends on its interaction with suitable
receptor at a suitable orientation. Molecular docking is a technique which determines

40
 CHAPTER-1
INTRODUCTION

the possibility of binding of ligand to the receptor and also evaluates the proper
orientation of ligand receptor complexes during binding.144 Molecular docking
comprises of two basic steps: search algorithm and scoring.

Search Algorithm
A search algorithm searches all the possible binding poses for the drug-receptor
complex. This phenomenon includes exploration of all the degree of freedoms for the
ligand, which creates all the possible conformers for the complex.145 Lowest energy
conformers are considered as best and presented as binding poses.146 There are
numerous algorithms used for the process; Monte Carlo simulations and Genetic
algorithms are the most commonly used.

Scoring
Evaluation of the binding poses is done on the basis of the computed ligand–receptor
interaction energy known as docking score of the pose.147 The calculation of this score
is based on energy functions such as energy of interaction (pKd), electrostatics, Van
der Waal’s interactions, hydrogen bonds, solvation effects, loss of entropy, active site
waters. Score represents the number given to the goodness or the energy of the
conformer.

Virtual screening
Identification of novel and promising leads is one of the approaches of drug discovery
and designing process. Pharmacophore models are used to search the compound
databases and library on the basis of perfect fit. The compounds perfectly fitted onto
the pharmacophore model are developed or optimized further as a lead compound.
Two search methods Fast or Best flexible methods are used to retrieve known
compounds having unidentified biological activities.

Evaluation of Promising Lead Candidate

Identification of possible hits by virtual screening gives rise to promising lead
compounds. The compounds should be evaluated for binding affinity before
progressing to the preclinical trials

41
 CHAPTER-2
Literature Review
 CHAPTER-2

LITERATURE REVIEW

Overview
Diabetes comprises of a group of disorders having the phenotype of hyperglycemia.
Insulin is the key player and a number of insulin products are available in the market
but majority of patients are still fighting to bring the normal glucose level in the
plasma. This demands search for some biological targets that would shift the focus of
diabetes treatment from insulin dependency. The role of glucagon in the
pathophysiology of diabetes has drawn attention towards glucagon as a rational
therapeutic target. Attempts have been done to develop clinically usefull GRA.
Ability of 11β HSD1enzyme to convert inactive cortisone to active cortisol, which has
a major role in mobilization of energy in the form of glucose in the liver, makes it a
promising therapeutic approach to treat type 2 diabetes induced hyperglycemia.
Inhibitors of 11β HSD1 block the synthesis of local cortisol which results in less
hepatic glucose production.

Glucagon Receptor Antagonists (GRA)

Gurd, RS and others (1982), reported synthesis of N-biotinyl-N-acetimidoglucagon

with potency equal to the original glucagon.148

Coy, DH and others (1984), used solid-phase methodology, preparative medium and
high performance reverse phase liquid chromatography to synthesize glucagon and its
Arg12 analogue in approximately 5% yields.149

Hruby, VJ and others (1986), synthesized seven glucagon analogues: [D-

Phe4,Tyr5,Arg12] glucagon; [D-Phe4,Tyr5,Lys17,18] glucagon; [Phe1,Glu3,Lys17,18]
glucagon; [Glu3,Val5,Lys17,18] glucagon; [Asp3,D-Phe4,Ser5,Lys17,18] glucagon; I4-
[Asp3,D-Phe4,Ser5,Lys17,18] glucagon and [Pro3] glucagon having partial GRA
activity.150

Hruby, VJ and others (1986), suggested that hepatic cells comprises of two distinct
glucagon receptors, GR-1 receptor which triggers inositol phospholipid breakdown
and GR-2 receptor which triggers adenylate cyclase activity.151

Hruby, VJ and others (1987), synthesized two new glucagon analogues having
antagonistis activities using solid-phase methodology. These two new analogues,

42
 CHAPTER-2

LITERATURE REVIEW

[Asp3,D-Phe4,Ser5,Lys17,18,Glu21] glucagon and [D-Phe4,Tyr5,3,5-I2-

Tyr10,Arg12,Lys17,18,Glu21] glucagon had good activity.152

Unson, CG and others (1989), designed and synthesized des-His1[Glu9] glucagon

amide with antagonist activity.153

Post, SR and others (1993), investigated the mechanisms of action of des-His'-

[Glu9] glucagon amide as a peptide antagonist on the glucagon receptor/adenylyl
cyclase system.154

Hruby, VJ and others (1995), synthesized a new glucagon analog [des His1, des
Phe6, Glu9] glucagon amide with IC50 value of 48 nM.155

Hruby, VJ and others (1996), synthesized a series of glucagon analogues by

introducing β-methylphenylalanine isomers (2S, 3S, 2S, 3R, 2R, 3R, and 2R, 3S) at
10th position. This modification was done to find out the structural and topographical
needs of the glucagon receptor. They also suggested that antagonism could be
enhanced by introducing (des-His1, Glu9) and a bulky group at position 5.156

Azizeh, BY and others (1997), designed, and synthesized five new glucagon
analogues, having excellent antagonistic properties.157

Hruby, VJ and others (1997), designed and synthesized fifteen glucagon analogues
by modifying N-terminal region of glucagon, in particular histidine-1, phenylalanine-
6, and aspartic acid-9. Modifications were done to study the effect of phenylalanine at
6th position on the glucagon action.158

Rault, S and others (1998), synthesized new pyrrolo[l,2-a]quinoxaline derivatives as

GRA using nitroanilines / orthophenylenediamines as precursor material.159

Hruby, VJ and others (1998), synthesized eight glucagon analogs. The compounds
were reported to have full or partial adenyl cyclase agonistic activity.160

Madsen, P and others (1998), synthesized the first non peptide competitive human
GRA, 2-(benzimidazol-2-ylthio)-1-(3,4-dihydroxyphenyl)-1-ethanone, NNC 92-1687
with binding affinity of 20 μM (IC50) and Ki = 9.1 μM.33

43
 CHAPTER-2

LITERATURE REVIEW

De Laszlo, SE and others (1999), reported the structure activity relationship of 2-

pyridyl-3,5-diaryl pyrroles, as glucagon receptor ligand and p38 kinase inhibitors.
This structure activity relationship study led to 2-(4-pyridyl)-5-(4-chlorophenyl)-3-(5-
bromo-2-propyloxyphenyl) pyrrole (L-168,049), a promising (Kb = 25 riM), glucagon
antagonist.161

Cascieri, MA and others (1999), reported a series of promising, orally bioavailable,

non-peptidyl, triarylimidazole and triarylpyrrole GRAs, 2-(4-Pyridyl)-5-(4-
chlorophenyl)-3-(5-bromo-2-propyloxyphenyl) pyrrole (L-168,049).162

Hruby, VJ and others (2000), designed and synthesized six cyclic glucagon
analogues, c[Asp9, Lys12][Lys17,18, Glu21] glucagon-NH2, c[Asp9, Lys12] glucagon-
NH2, c[Lys12, Asp15] glucagon-NH2, c[Asp15, Lys18] glucagon-NH2, [Lys17-c[Lys18,
Glu21] glucagon-NH2, and c[Lys12, Asp21] glucagon-NH2.163

Sivarajah, P and others (2001), cloned glucagon-like receptors from two species of
frogs to understand the biology of proglucagon-derived glucagon-like sequence.164

Hruby, VJ and others (2001), designed and synthesized 23 truncated glucagon

analogues containing the middle and the C-terminal region (16-23) sequence of
gluacgon.165

Hruby, VJ and others (2001), designed and synthesized a series of conformationally

constrained glucagon analogues by disulfide or lactam bridges to understand the
conformational requirement for the N-terminal region which is essential for signal
transduction.166

Ladouceur, GH and others (2002), discovered a novel GRA 5-Hydroxyalkyl-4-

phenylpyridines with IC50 of 0.11 µM.167

Ling, A and others (2002), optimized a series of alkylidene hydrazide derivatives

with alkoxyaryl moiety as competitive human GRAs.168

Smith, AR and others (2002), established a narrow SAR for the 4-aryl group in 4-
aryl-pyridine glucagon antagonists.169

44
 CHAPTER-2

LITERATURE REVIEW

Ladouceur, GH and others (2002), optimized arrangements of substitutions in 4-

aryl-pyridine to produce GRAs.170

Ladouceur, GH and others (2002), evaluated the SAR of a new class of 4-aryl-
pyridine glucagon receptor antagonists bearing the hydroxyl moiety.171

Madsen, P and others (2002), modified the core (proximal) dimethoxyphenyl group,
the benzyl ether linkage, as well as the (distal) benzylic aryl group of the lead 2, 3-
cyano-4-hydroxybenzoic acid (3,5-dimethoxy-4-isopropylbenzyloxybenzylidene)-
hydrazide to generate a promising human glucagon receptor (hGluR) antagonists.172

Unson, CG and others (2002), reported that for high affinity binding of any ligand
with the glucagon receptor multiple contacts with the residues of N-terminal tail and
first extracellular loop domain of the receptor are required.173

Sapse, AM and others (2003), performed ab initio calculations on the two series of
non-peptide glucagon receptor antagonists (pyrrolo[1,2-a] quinoxalines and
mercaptobenzimidazoles). The results suggested that substituted phenyl ring in the
pyrrolo[1,2-a]quinoxaline series does not affect binding affinity efficiently.174

Hruby, VJ and others (2003), determined the solution structure of the promising
glucagon antagonist, [desHis1, desPhe6, Glu9] glucagon amide by homonuclear 2D
NMR spectroscopy at pH 6.0 and 37 °C in perdeuterated dodecylphosphocholine
micelles.175

Kurukulasuriya, R and others (2004), evaluated biaryl amides derived from a

reported series of ureas (Figure 2.1) as promising human GRAs.176

Figure 2.1: Urea glucagon receptor antagonist

45
 CHAPTER-2

LITERATURE REVIEW

Jiang, G and others (2004), designed an in vivo glucagon receptor binding assay
125
using I labeled glucagon as a radiotracer. The results revealed that glucagon
receptor is chiefly located in liver, and to some extent in kidney; and peptidyl and
non-peptidyl GRAs bind to the hepatic glucagon receptor in vivo.177

Hsu, WH and others (2004), investigated the function of arginine vasopressin (AVP)
on glucagon secretion in both normal and diabetic rats.178

Potterat, O and others (2004), isolated a new bicyclic 19-peptide, BI-32169 (Figure
2.2), from the culture broth of Streptomyces sp. (DSM 14996), which showed human
glucagon receptor antagonistic activity (IC 50 440 and 320 nM, respectively) in a
functional cell-based assay.179

Figure 2.2: Structure of BI-32169 and its methyl ester

Shen, DM and others (2005), discovered a novel class of spiro-ureas as promising

human GRAs in both binding and functional assays.180

Duffy, JL and others (2005), discovered a novel human GRA (Figure 2.3), with high
activity (IC 50 = 34 nM) and moderate bioavailability (36% in mice).181

46
 CHAPTER-2

LITERATURE REVIEW

Figure 2.3: Novel human glucagon receptor antagonist

Bolli, GB and others (2006), reviewed the role of glucagon in physiology,

physiopathology and clinical medicine. They concluded that most frequent risk
connected to the glucagon excess or deficiency is diabetes mellitus. Drugs having
ability to reduce glucagon secretion may contribute to improve the overall glycaemic
control in diabetes.182

Duffy, JL and others (2006), demonstrated a technique to evaluate the efficiency of

a GRA in the surgically removed perfused liver. The technique involved use of 13C
NMR in analyzing the biosynthesis of [13C] glycogen from [13C] pyruvate in the
process of gluconeogenesis.183

Liang, R and others (2007), synthesized a series of conformationally constrained tri-

substituted ureas as promising GRA. The efforts resulted in the identification of
compound (Figure 2.4), having a binding IC50 of 4.0 nM.184

Figure 2.4: Conformationally constrained tri-substituted urea

Qureshi, SA and others (2007), cloned a cDNA corresponding to the glucagon

receptor from dog liver RNA.185

47
 CHAPTER-2

LITERATURE REVIEW

Mortensen, OH and others (2007), characterized the promoter regions of the human
glucagon receptor gene.186

Lau, J and others (2007), optimized a weak human GRA with an IC50 of 7 μM to a
promising glucagon antagonist. . The GRA efficiently reduced plasma glucose levels
in a diabetic animal model and exhibited oral bioavailability in animals.187

Kim, RM and others (2008), described the discovery and optimization of promising
and selective aminobenzimidazole based GRAs.188

Kodra, JT and others (2008), optimized a series of small molecule human GRAs by
counter screening against the human GIT inhibitory polypeptide receptor. Through
SAR studies, they were successful in finding compounds with nanomolar affinities
against hGluR, selective hGIPR and human glucagon-like peptide-1 receptor (hGLP-
1R).189

Madsen, P and others (2009), designed and synthesized compounds with cyclic
cores (5-aminothiazoles) as promising human GRAs with advanced pharmacokinetic
(PK) properties as a therapeutic approach towards type 2 diabetes.190

Zhuo JL and others (2009), reviewed the role of glucagon and its receptor signaling
in the pathogenesis of Type 2 diabetic renal injury, and explored the possibility of
targeting glucagon receptor signaling as a anti diabetic strategy in Type 2 diabetes.191

Marahiel, MA and others (2010), performed tandem mass spectrometric and nuclear
magnetic resonance spectroscopic studies on the GRA BI-32169. The studies revealed
that BI-32169 is a lasso-structured peptide constituting the new class III of lasso
peptides.192

Rafferty, EP and others (2010), examined the efficacy and stability of two novel
glucagon antagonist peptides in vitro and their duration of action in vivo in normal
mice.193

Shen, DM and others (2011), discovered a novel class of 1,3,5-pyrazoles as

promising human GRAs.194

48
 CHAPTER-2

LITERATURE REVIEW

Sinz, C and others (2011), reported a novel class of N-aryl-2-acylindole human

glucagon receptor (hGCGR) antagonists.195

Filipski, KJ and others (2012), discovered a novel series of pyrazole ethers and
aminopyrazoles as GRAs.196

Kieffer, TJ and others (2012), reviewed the biology of GIP, GLP-1 and glucagon
and examined the different therapeutic strategies to activate and antagonize the
receptors of these peptides.197

Xiong, Y and others (2012), discovered a promising, selective GRA 9m, N-[(4-
{(1S)-1-[3-(3,5-dichlorophenyl)-5-(6-methoxynaphthalen-2-yl)-1H-pyrazol-1-
yl]ethyl}phenyl)-carbonyl]-β-alanine.198

Chung, JYL and others (2012), described asymmetric synthesis of 1,1,2,2-

tetrasubstituted ethane core substituted with a propyl and three aryl groups including a
fluoro-indole as a GRA drug candidate for the treatment of type 2 diabetes.199

Guzman-Perez, A and others (2013), reported (S)-3-[4-(1-{3,5-dimethyl-4-[4-

(trifluoromethyl)-1H-pyrazol-1-yl]phenoxy}butyl)benzamido]propanoic acid, a novel
and promising small molecule GRA for the treatment of diabetes mellitus.200

Irwin, N and others (2013), evaluated two novel glucagon analogues, desHis1Glu9-
glucagon-[mPEG] and desHis1Glu9(Lys30PAL)-glucagon for GRA properties.201

O’Harte, FPM and others (2013), tested in vitro and in vivo biological activities of
four novel structural analogues of glucagon, desHis1Pro4-glucagon, desHis1Pro4Glu9-
glucagon, desHis1Pro4Glu9Lys12FA-glucagon and desHis1Pro4-Glu9Lys30FA-
glucagon.202

DeMong D and others (2014), developed a novel series of spiroimidazolone-based

antagonists of the human glucagon receptor (hGCGR).203

Siu, FY and others (2014), reported the crystal structure of the seven transmembrane
helical domain of human glucagon receptor at 3.4 A˚ resolutions, complemented by
extensive site-specific mutagenesis, and a hybrid model of glucagon bound to

49
 CHAPTER-2

LITERATURE REVIEW

glucagon receptor to understand the molecular recognition of the receptor for its
native ligand.27

11β HSD1 Inhibitors

Alberts, P and others (2002), examined the plasma glucose lowering ability by
selective inhibition of 11β HSD 1 in a hyperglycaemic and hyperinsulinaemic mouse
model.204

Alberts, P and others (2003), evaluated the 11β HSD1 inhibitory effects on plasma
glucose levels, glucose tolerance and insulin sensitivity in type 2 diabetes mouse
models.205

Andrews, RC and others (2003), evaluated the approach of lowering circulating

cortisol concentrations and its effect in enhancing insulin sensitivity and hepatic lipid
catabolism in metabolic disorders by inhibiting 11β HSD1 using the nonselective 11β-
HSD inhibitor, carbenoxolone.206

Tomlinson, JW and others (2005), described the role of 11β HSD1 as an enzyme
that liberates active glucocorticoid, cortisol, within liver and adipose tissue.207

Gu, Xin and others (2005), reported that heteroaryl substituted

bicyclo[2.2.2]octyltriazoles are promising and selective 11β HSD1 inhibitors with
excellent pharmacokinetic profiles.208

Olson, S and others (2005), identified adamantyl triazoles as selective inhibitors of

11β HSD1. The compounds were found to be active both in vitro and in in vivo
pharmacodynamic models.209

Xiang, J and others (2005), described the unanticipated discovery of β-keto sulfones
as promising 11β HSD1 inhibitors.210

Odermatt, A and others (2006), developed a pharmacophore based on LBDD for

11β HSD1 inhibitors. The results suggested that the model associated with suitable
cell-based activity assays can be used for the identification of selective and promising
inhibitors.211

50
 CHAPTER-2

LITERATURE REVIEW

Potter, BVL and others (2006), synthesized a series of novel benzothiazole

derivatives as 11β HSD1 inhibitors. The SAR studies revealed that insertion of a
chlorine substituent at the 4 position of the benzothiazole ring increases the inhibitory
property.212

Richards, S and others (2006), discovered that dichloroaniline amide are novel and
potent 11β HSD1 inhibitors.213

Yeh, VSC and others (2006), synthesized and biologically evaluated a series of
metabolically stable butyrolactam as 11β HSD1 inhibitors.214

Yeh, VSC and others (2006), synthesized and biologically evaluated a series of
metabolically stable adamantane amide as 11β HSD1 inhibitors.215

Sutin, L and others (2007), reported 2,5,5-trisubstituted oxazolones as promising

inhibitors of 11β HSD1.216

Fujisawa, Y and others (2007), investigated the cause of metabolic disorders in the
infants of diabetic mothers and proposed that 11β HSD1 might be a key player behind
this.217

Potter, BVL and others (2007), discovered and synthesized three structurally
different series of novel 11β HSD1 inhibitors that inhibit human 11β HSD1 in the
micromolar range.218

Potter, BVL and others (2007), reported the discovery and synthesis of some 18β-
glycyrrhetinic acid (18β-GA) derivatives and their inhibitory activities against rat
hepatic11β HSD1 and rat renal 11β-HSD2.219

Patel, JR and others (2007), discovered a novel class of adamantane ethers as 11β
HSD1 inhibitors.220

Webster, SP and others (2007), discovered and chemically modified a series of

adamantyl amide as 11β HSD1 inhibitors.221

Fotsch, C and others (2007), synthesized a series of 2-anilinothiazolones as 11β

HSD1 inhibitors.222

51
 CHAPTER-2

LITERATURE REVIEW

Flyren, K and others (2007), identified a series of piperidine amide inhibitors of

human 11β HSD1.223

Bhatt, BG and others (2008), reported a glucocorticoid induced diabetic animal

model using 11β HSD1 antisense oligonucleotide (ASO) as an inhibitory tool to
evaluate the inhibition of 11β HSD1 as anti diabetic target.224

Uckaya, G and others (2008), determined 11β HSD1 mRNA expression levels in
adipose tissues in diabetic patients. The down regulation of 11β HSD1 activity
suggested its contribution to the pathogenesis of type 2 diabetes mellitus.225

Iwasaki, Y and others (2008), examined the effects of multiple humoral factors
related to the metabolic syndrome on the transcriptional activity of 11β HSD1 gene in
hepatocytes in vitro.226

McCormick, KL and others (2008), examined the regulation of microsomal 11β

HSD1 activity.227

Sundbom, M and others (2008), reported BVT116429 as a selective 11β HSD1

inhibitor which works on the circulating concentrations of adiponectin with
concomitant effects on glucose homeostasis.228

Fotsch, C and others (2008), designed and synthesized a series of compounds

containing the 2-amino-1,3-thiazol-4(5H)-one core as promising 11β HSD1
inhibitors. Compound (S)-33a (Figure 2.5) was further progressed to
pharmacodynamic studies in cynomolgus monkeys, where it inhibited adipose 11β
HSD1 activity.229

O
N

N S
H
OH
F

Figure 2.5: Compound (S)-33a

Xiang, J and others (2008), synthesized, in vivo evaluation and performed structure-
activity relationship, of piperazine sulfonamides as 11β HSD1 inhibitors.230
52
 CHAPTER-2

LITERATURE REVIEW

Kang, NS and others (2008), performed 3D-QSAR CoMFA and CoMSIA studies on
70 11β HSD1 inhibitors, based on molecular docking conformations obtained by
using FlexX-Pharm.231

Navarrete-Vazquez, G and others (2008), synthesized and evaluated N-(6-

Substituted-1,3-benzothiazol-2-yl) benzenesulfonamide derivatives for in vivo anti
diabetic activity in a non-insulin-dependent diabetes mellitus rat model and in vitro as
11β HSD1 inhibitors. The most active compounds were docked into the crystal
structure of 11β HSD1. Docking results indicated hydrogen bond interactions with
catalytic amino acid residues.232

Powers, JP and others (2008), through high-throughput screening of a small-

molecule compound library identified a series of arylsulfonylpiperazines as promising
and selective (11β HSD1) inhibitors.233

Zhu, Y and others (2008), identified 4-Methyl-5-phenyl-(1,2,4)-triazoles as selective

inhibitors of 11β HSD1.234

Aster, SD and others (2008), identified 3-Aryl-5-phenyl-(1,2,4)-triazoles as selective

inhibitors of 11β HSD1.235

Shen, J and others (2008), discovered novel inhibitors of 11β HSD1 by using
docking and pharmacophore modeling. Several compounds, with large structural
diversity and good potency against 11β HSD1, were found and their potency was
determined by the enzyme assay. New scaffolds of 11β HSD1 inhibitors were also
reported.236

Zhu, Y and others (2008), identified 3-(Phenylcyclobutyl)-1,2,4-triazoles as

selective inhibitors of 11β HSD1.237

Wang, H and others (2008), identified several series of pyridine amides as selective
and promising 11β HSD1 inhibitors.238

Gumy, C and others (2009), evaluated six traditional anti diabetic medicinal plants
extracts for inhibition of 11β HSD1 activity and glucocorticoid receptor (GR)
activation in transfected HEK-293 cells. Leave extracts of Eriobotrya japonica
inhibited 11β HSD1 over 11β-HSD2. Extracts of roasted but not native coffee beans
53
 CHAPTER-2

LITERATURE REVIEW

inhibited 11β HSD1 over 11β-HSD2, emphasizing the importance of sample

preparation.239

Miyamoto, Y and others (2009), investigated the correlation between variations in

the 11β HSD1 gene with metabolic syndrome, by a case–control study using a
population-based urban Japanese cohort. The findings suggested that polymorphisms
and haplotypes in the 11β HSD1 gene are not significantly associated with metabolic
syndrome in the Japanese population.240

Potter, BVL and others (2009), reported the discovery of a series of novel
adamantyl carboxamides as selective inhibitors of human 11β HSD1 in HEK-293
cells transfected with the 11β HSD1 gene.241

Classen-Houben, D and others (2009), using a pharmacophore model based on the

crystal structure of the GA-derivative carbenoxolone in complex with human 11β
HSD1, compared the biological activity of 18β-GA and its diastereomer 18α-GA
against the two enzymes in lysates of transfected HEK-293 cells and showed that 18-
GA selectively inhibits 11β HSD1 but not 11β-HSD2.242

Srivastava, RAK and others (2009), reported that fenofibrate down regulated
expression of hepatic and adipose 11β HSD1 gene, which contributed in attenuating
diabetic state.243

Stewart, PM and others (2009), reported the contribution, of 11β HSD in generation
of cortisol from liver.244

Morgan, SA and others (2009), determined the mechanisms supporting

glucocorticoid-induced insulin resistance in skeletal muscle and described the
mechanism of improving insulin sensitivity by 11β HSD1 inhibitors.245

Odermatt, A and others (2009), evaluated the impact of the NADPH/NADP+ ratio
and the extracellular glucose on 11β HSD1 activity by applying microsomal
preparations and intact HEK-293 cells expressing 11β HSD1 in the presence or
absence of hexose-6-phosphate dehydrogenase (H6PDH). Measurements in intact
cells suggested that the ER-luminal NADPH concentration is highly sensitive to
fluctuating extracellular glucose levels.246

54
 CHAPTER-2

LITERATURE REVIEW

Lloyd, DJ and others (2009), examined the in vivo effects of a specific 11β HSD1
inhibitor on atherosclerosis while assessing glucose homeostasis. The results
suggested that 11β HSD1 inhibition leads to improved glucose metabolism.247

Siu, M and others (2009), converted an initial lead, N-(Pyridin-2-yl)

arylsulfonamides in to an efficient ligand with improved physiochemical properties
using a deletion strategy. This strategy allowed for further optimization of potency
leading to the discovery of the clinical candidate PF-915275.248

Sun, D and others (2009), described the synthesis and SAR of a series of
arylsulfonylpiperazine inhibitors of 11β HSD1.249

McMinn, DL and others (2009), optimized novel 4,4-disubstituted

cyclohexylbenzamide as 11β HSD1 inhibitors.250

Rew, Y and others (2009), described discovery and optimization of a piperidyl

benzamide series of 11β HSD1 inhibitors.251

Neelamkavil, SF and others (2009), discovered a series of azepine sulfonamides as

promising inhibitors of 11β HSD1.252

Shen, J and others (2009), reported the synthesis, structure–activity relationships,

and efficacy evaluation of 4-(phenylsulfonamidomethyl)benzamides as 11β HSD1
inhibitors.253

Christmann-Francka, S and others (2009), identified benzylamides of pentanedioic

acid as inhibitors of 11β HSD1.254

Christmann-Francka, S and others (2009), identified spiro-carboxamides as

inhibitors of 11β HSD 1 by HTS.255

Karlsson, C and others (2010), investigated the relationship between 11β HSD1
mRNA levels in adipose tissue and in skeletal muscle and anthropometric and
metabolic measurements in humans with and without impaired glucose
homeostasis.256

55
 CHAPTER-2

LITERATURE REVIEW

Odermatt, A and others (2010), reviewed progress on the biochemical

characterization of 11β HSD1, with a focus on cofactor and substrate specificity and
on possible alternative functions of this enzyme.257

Stewart, PM and others (2010), reviewed the role of 11β HSD1 in the pathogenesis
of obesity, type II diabetes and the metabolic syndrome.258

Huber, R and others (2010), evaluated the efficacy and safety of the 11β HSD1
inhibitor INCB13739, in combination therapy with metformin in patients with type 2
diabetes exhibiting inadequate glycemic control (A1c 7-11%).259

Edgerton, DS and others (2010), investigated the effect of acute 11β HSD1
inhibition on HGP and fat metabolism during insulin deficiency in conscious dogs.260

Wang, M and others (2010), reported structure activity relationship studies on

thiazolones as promising 11β HSD1 inhibitors. The C-5 center was prone to
epimerization in vitro and in vivo, forming a less promising diastereomer. A methyl
group was added to the C-5 position to eliminate epimerization, leading to the
discovery of (S)-2-((1S,2S,4R)-bicyclo[2.2.1]heptan-2-ylamino)-5-isopropyl-5-
methylthiazol4(5H)-one (AMG 221).261

Ge, R and others (2010), reported novel compounds which inhibits 11β HSD1
reductase activity by increasing its oxidase activity without inhibition of 11β-HSD2.
They proposed dual modulation on 11β HSD1 as a better therapeutic strategy for type
II diabetes.262

Yan, X and others (2010), reported a novel series of diarylsulfones as human 11β
HSD1 inhibitors.263

Shah, U and others (2010), reported the discovery of novel azabicyclic sulfonamide
based 11β HSD1 inhibitors.264

Cheng, H and others (2010), reported the design and development of a series of
highly selective pyrrolidine carboxamide 11β HSD1 inhibitors.265

Tice, CM and others (2010), reported the synthesis of 2-adamantyl carbamate

derivatives of piperidines and pyrrolidines as human 11β HSD1 inhibitors.266

56
 CHAPTER-2

LITERATURE REVIEW

Ahn, JH and others (2010), synthesized and evaluated a new series of cyclic
sulfonamide derivatives for their ability to inhibit 11β HSD1.267

Tice, CM and others (2010), identified a class of spirocyclic ureas which

promisingly inhibit human 11β HSD1 in vitro with the help of structure guided drug
design.268

Ye, XY and others (2011), designed and synthesized a series of polycyclic acids as
novel, promising, and selective inhibitors of human 11β HSD1.269

Liu, Y and others (2011), observed that 11β HSD1 expression in livers increases
under non-fasting state. They suggested that insulin is a regulator of 11β HSD1
expression and inhibition of this enzyme can be an effective therapeutic approach.270

Simunkova, K and others (2011), reported the cortisone acetate test in 12 young
adult type 1 diabetic patients. The results showed adaptive changes of 11β HSD
activity associated with altered cortisol tissue supply.271

Tice, CM and others (2011), obtained hit 1,3-oxazinan-2-one (9a) from structure
based design with an in vitro IC 50 of 42 nM against 11β HSD1. Optimization of 9a
for improved in vitro enzymatic and cellular potency yielded 25f (Figure 2.6) with IC
50 values of 0.8 nM for the enzyme and 2.5 nM in adipocytes.272

Figure 2.6: Optimization of 1,3-oxazinan-2-one

Lipson, VV and others (2011), explained the importance of glucocorticoids in

metabolic syndrome. They highlighted approaches to reduce the action of the
hormones in the liver and adipose tissue as a basis for novel therapies of type II
diabetes mellitus and surveyed structural types of 11β HSD1 inhibitors.65

57
 CHAPTER-2

LITERATURE REVIEW

Kim, KY and others (2011), investigated the anti diabetic and anti adipogenic effects
of 2-(3-benzoyl)-4-hydroxy-1,1-dioxo-2H-1,2-benzothiazine-2-yl-1-phenylethanone
(KR-66344) as 11β HSD1 inhibitor.273

Sun, D and others (2011), described the synthesis and SAR studies of 4,4-
disubstituted cyclohexylbenzamide as 11β HSD1inhibitor.274

Shen, J and others (2011), reported discovery of seven hit compounds as novel 11β
HSD1 inhibitors from in silico screening of the the commercially available Maybridge
database. They further optimized the lead compounds to have two promising
compounds for further development.275

Wu, SC and others (2011), discovered a series of hydroxyisoquinolines as selective

11β HSD1 inhibitors.276

Maletic, M and others (2011), reported the discovery of a promising ethyl sulfone as
11β HSD1 inhibitors.277

Venier, O and others (2011), identified a novel class of ureas as 11β HSD1
inhibitors by high throughput screening.278

Sun, W and others (2011), identified 3-(Phenylcyclobutyl)-1,2,4-triazoles as 11β

HSD1 inhibitors.279

Ahn, JH and others (2011), synthesized and evaluated a new series of thiazolidine
derivatives with an adamantyl group for their ability to inhibit 11β HSD1.280

Valeur, E and others (2012), identified indole-pyrrolidines as 11β HSD1 inhibitors

by HTS. They further optimized the initial hit through SBDD which led to 7-
azaindole-derivatives, with the best analogues displaying single digit nanomolar IC 50
potency.281

Xiong, B and others (2012), designed and synthesized a novel class of 1-arylsulfonyl
piperidine-3-carboxamides derivatives as human 11β HSD1 selective inhibitor.282

Maser, E and others (2012), showed that M. charantia extract capsules have
selective 11β HSD1 inhibitory activity.283

58
 CHAPTER-2

LITERATURE REVIEW

Dhanesha, N and others (2012), examined the pharmacological inhibitory effect of

11β HSD1 in a genetic mice model.284

Scott, JS and others (2012), reported the optimization and development of a

carboxylic acid class of 11β HSD1 inhibitors (Figure 2.7).285

Figure 2.7: Development of AZD8329 from AZD4017

Goldberg, FW and others (2012), described the co-optimization of human and

murine potency within the (5-hydroxy-2-adamantyl)-pyrimidine-5-carboxamide
series. Two approaches were described- a data-driven (Free-Wilson) analysis and a
structure-based design approach.286

Shen, J and others (2012), discovered a new series of 2-alkyl-1-

arylsulfonylprolinamides on the basis of scaffold hopping, as 11β-HSD-1
inhibitors.287

Wan, ZK and others (2012), described the SAR studies of a series of piperazine
sulfonamide-based 11β HSD1 inhibitors. (R)-3,3,3-Trifluoro-2-(5-(((R)-4-(4-fluoro2-
(trifluoromethyl)phenyl)-2-methylpiperazin-1-yl)sulfonyl)thiophen-2-yl)-2-
hydroxypropanamide 18a (HSD-621) was identified as a promising and selective 11β
HSD1 inhibitor.288

Kim, KY and others (2012), investigated the anti diabetic and anti adipogenic effects
of 4-(2-(1,1-dioxido-6-(2,4,6-trichlorophenyl)-1,2,6-thiadiazinan-2-yl)acetamido)
adamantane-1-carboxamide (KR-67183), a novel selective 11β HSD1 inhibitor.289

59
 CHAPTER-2

LITERATURE REVIEW

Xiao, X and others (2012), investigated placental 11β-HSDs expression, and

analyzed their relationship with cortisol levels in maternal and umbilical vein.290

Rew, Y and others (2012), devloped a novel series of benzenesulfonanilide

derivatives as 11β HSD1 inhibitors through modification of the sulfonamide core of
the arylsulfonylpiperazine lead structures.291

Scott, JS and others (2012), optimized a series of neutral 2-thioalkyl-pyridine as 11β

HSD1 inhibitors with excellent bioavailability.292

Christmann-Franck, S and others (2012), identified indole-pyrrolidines as

inhibitors of 11β HSD1 by HTS.293

Shen, J and others (2012), reported the design, synthesis and biological evaluation of
novel amide and urea as 11β HSD1 inhibitors.294

Ritter, K and others (2013), isolated racemic cis-1,1-dioxo-5,6-dihydro-

[4,1,2]oxathiazine derivative 4a as an impurity in a sample of a hit from a HTS
campaign on 11β HSD1. After separation by chiral chromatography the 4a-S, 8a-R
enantiomer of compound 4a was identified as the true, promising enzyme inhibitor.295

Anagnostis, P and others (2013), reviewed the role of 11β HSD1 inhibition in the
treatment of metabolic disorders and provided rationale for the development of 11β
HSD1 inhibitors in diabetes.296

Kim, KY and others (2013), investigated the anti-diabetic and anti-inflammatory

effects of N-(5-carbamoyladamantan-2-yl)-3-((2-fluorophenyl)sulfonyl)thiazolidine-
2- carboxamide (KR-67105 a novel 11β HSD1 inhibitor.297

Harno, E and others (2013), hypothesized that increasing the substrate for 11β
HSD1 (11-dehydrocorticosterone, 11-DHC) would adversely affect metabolic
parameters. They analyzed that chronic administration of 11- dehydrocorticosterone
to male C57BL/6J mice results in increased circulating glucocorticoids, and down-
regulation of the hypothalamic-pituitary-adrenal axis.298

Nair, SK and others (2013), identified N-(Pyridin-2-yl) arylsulfonamides (PF-

915275) as promising 11β HSD1 inhibitors.299

60
 CHAPTER-2

LITERATURE REVIEW

Navarrete-Vazquez, G and others (2013), synthesized a tetrazole isosteric analogue

of clofibric acid as 11β HSD1 inhibitors.300

Venier, O and others (2013), reported a novel series of adamantane urea as a

promising 11β HSD1 inhibitor.301

Ge, RS and others (2013), synthesized and biologically evaluated a series of

structurally novel mono-carbonyl curcumin analogues as 11β HSD1 inhibitors.302

Odermatt, A and others (2013), discovered a novel class of 11β HSD1 inhibitors
bearing an arylsulfonylpiperazine scaffold using molecular modeling approach and
classical bioisosteric studies.303

Robl, JA and others (2014), designed and synthesized a series of 2-adamantylmethyl

tetrazoles bearing a quaternary carbon at the 2-position of the adamantane ring as
novel, promising, and selective human 11β HSD1 inhibitors.304

Fardella, CE and others (2014), implemented a combined ligand and structure-based

virtual screening approach using crystal structures of human 11β HSD1 in complex
with inhibitors as source of structural information, to identify novel 11β HSD1
inhibitors. Selected compounds were identified and evaluated in cell-based assays for
cytotoxicity and 11β HSD1 mediated cortisol production inhibitory capacity.305

Huang, Y and others (2014), designed and synthesized series of mono-carbonyl

curcumin analogues by deleting the reactive beta-diketone moiety responsible for the
pharmacokinetic limitation of curcumin. They demonstrated that out of nine, four
curcumin analogues were selective inhibitors of human and rodent 11β HSD1.306

Akiyama, N and others (2014), elucidated adipose dysfunction via 11β HSD1
activation in the development of obesity-related diabetes.307

Navarrete-Vazquez, G and others (2014), designed and synthesized a series of

compounds using a bioisosteric approach and evaluated their in vitro for 11β
HSD1inhibitory activity.308

Ahn, SK and others (2014), optimized 11β HSD1 inhibitors (3-amino-N-adamantyl-

3-methylbutanamide derivatives) using structure based drug design tools.309

61
 CHAPTER-2

LITERATURE REVIEW

Problem Envisaged
Diabetes is a chronic glucose metabolic disorder affecting millions of people
worldwide. The number of affected population is continuously increasing. Various
therapies have been exercised for the treatment of type 2 diabetes induced
hyperglycemia, but tight glycemic control is still a challenge. Hence there is a
pressing need for additional therapeutic agents which could give a fight to this life
threatening metabolic disorder. In earlier days researchers completely focused on
insulin supplement or enhancing β cell mass; but type 2 diabetes mellitus is not just
insulin impairment. It is actually caused due to imbalanced insulin-glucagon action.
Insulin is responsible for conversion of plasma glucose into its stored form, glycogen
whereas glucagon acts contrary to insulin by converting glycogen back into free
glucose. If insulin impairment is responsible for non conversion of free glucose into
glycogen, glucagon worsens the situation by converting back already stored glycogen
into free glucose, which leads to hyperglycemia. Relative insulin deficiency with
glucagon excess apparently bestows to afflicted glucose production and
hyperglycemia. Thus, targeting reducing glucagon signaling can be a promising
therapeutic approach for treatment of type 2 diabetes mellitus. Up to now, three GRAs
have been reported to be effective in human subjects. The first attempt of glucagon
receptor blockade to reduce the glucose output was achieved with (+)-BAY 27-9955.
Later on, MK0893 were disclosed as a promising GRA in preclinical and human
clinical trial results. Similar to (+)-BAY 27-9955 and MK-0893, LY2409021 was also
reported to block exogenous glucagon-stimulated glucose production in humans. (+)-
BAY 27-9955 and MK-0893 succeeded in short term human studies by safely
blocking exogenous glucagon actions, however clinical development for these
compounds were not pursued and LY2409021 is currently in phase II clinical trial.

Similarly, there are several other biochemical processes in the human body which
increases the plasma glucose concentration by metabolizing other nutrients into
glucose. Gluconeogenesis is a similar process in which glucocorticoid circulates
glucose by accelerating the transcription of rate limiting gluconeogenesis and
glycolysis enzymes.310 Cortisol is the centrally acting glucocorticoid and is the main
player of gluconeogenesis. The activity of cortisol is regulated by 11β HSD isozymes.
11β HSD1 converts inactive cortisone to active cortisol, whereas 11β HSD2 converts
62
 CHAPTER-2

LITERATURE REVIEW

cortisol back into cortisone. Hence, inhibiting 11β HSD1 can be a potential target to
reduce hyperglycemia. 11β HSD1 inhibitors block the circulating concentration of
glucocorticoid and thus interfere with the process of gluconeogenesis and hence,
result in less hepatic glucose production (HGP). Liver selective 11β HSD1 inhibitors
actions include decreased hepatic glucose production, enhanced insulin sensitivity,
retardation of progressive β cell dysfunction, and lower risk of hypoglycemia when
compared to sulfonylurea or insulin treatment. Several attempts have been made to
synthesize a promising 11β HSD1 inhibitor; but till now only carbenoxolone has
exhibited some efficacy in humans. But, carbenoxolone also found to inhibit 11β
HSD2 and also have some side effects like hepatic ulcers.

Thus, there is a compelling need to identify a promising GRA and a promising 11β
HSD1 inhibitor that can be considered as alternative clinical development candidates
for the treatment of type 2 diabetes.

Work Plan
The utility of virtual screening in lead identification is unquestionable and the success
stories can be traced in previous literature. In view of this we have made an effort to
identify novel 11 βHSD1 inhibitors and GRA through sequential use of ligand based
pharmacophore modeling, virtual screening, molecular docking and biological
evaluation protocols.
The specific objectives of the present research work:
1. To develop QSAR based models that establish the correlation among
physical-chemical properties of compounds with the biological activity of
GRA and 11β HSD1 inhibitor by using TSAR 3D version 3.3 software.
2. To generate pharmacophoric features essential for ligand receptor/enzyme
interaction.
3. To elucidate important amino acid involved in ligand receptor/enzyme
interaction.
4. To identify new promising GRA and 11β HSD1 inhibitor from chemical
compound databases.
5. Biological evaluation of compounds retrieved from databases.

63
 CHAPTER-2

LITERATURE REVIEW

Scheme of Work
Scheme 1: Descriptor based QSAR (Hansch analysis)

Figure 2.8: Various steps involved in QSAR model generation

64
 CHAPTER-2

LITERATURE REVIEW

Scheme 2: Ligand based pharmacophore (LBP) modeling

Antidiabetic target (GRA

Data set preparation
and 11 β HSD type 1)
(Sketching,CharmM forcefield and
energy minimization)

Feature selection Conformational Analysis

1. Hydrogen Bond (Acceptor / Donor/

Acceptor lipid)

2. Charge (Positive / Negative)

Feature generation
3. Hydrophobic, Hydrophobic
Aliphatic/ Aromatic

4. Ring-Aromatic

Hypogen
Algorithmm
Test set Training set Pharmacophore generation

No
Statistically relevant ?

Model validation Yes

Fischer validation

External test set validation No

Yes
Best model obtained Model validated ?

Figure 2.9: Steps involved in ligand based pharmacophore modeling

65
 CHAPTER-2

LITERATURE REVIEW

Scheme 3: Virtual screening

Figure 2.10: Steps involved in virtual screening

66
 CHAPTER-2

LITERATURE REVIEW

Scheme 4: Structure based pharmacophore (SBP) modeling

Figure 2.11: Steps involved in structure based pharmacophore modeling

67
 CHAPTER-2

LITERATURE REVIEW

Scheme 5: Molecular docking (Libdock)

Figure 2.12: Steps involved in molecular docking

68
 CHAPTER-2

LITERATURE REVIEW

Scheme 6: Biological evaluation

Figure 2.13: Steps involved in anti diabetic biological evaluation

69
 CHAPTER-3
Global Descriptor
Based QSAR
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Descriptor Based QSAR Approach

The objective of present research work has been to elucidate the structural and
physical/chemical requirements for inhibition of GRA and 11β HSD1 which are
promising targets for the development of anti-diabetic drugs. In order to achieve the
set goals battery of chemometric techniques (MLR, PLS, NN) has been used to
construct predictive quantitative structure activity relationships models. All the
developed models were rigorously validated to ensure high predictability.

Materials and Methods

Series Selection
Development of a successful QSAR models depends on the quality of structural and
biological activity data. For proper selection of the data, following guidelines have
been laid down:
1. Minimum 20 compounds is recommended
2. Same biological assay method from the same lab is preferred.
3. The biological activity data should be broad and uniform.
4. Biological activity data should be in real numbers.
On the basis of above mentioned guidelines, three series of compounds with GRA and
11β HSD1inhibitory activity was selected.
Series 1: β alanine, isoserine and thiazole based series developed by Madsen, P and
others were selected and merged to have a complete series of 70 compounds of
GRAs.
Series 2: Pyrrolidine carboxamide derivatives synthesized by Cheng, H and others.
Series 3: Bicyclo[2.2.2]octyltriazole derivatives synthesized by Maletic, M and
others.

Data Set Preparation

The diverse structures of all the three mentioned series were sketched using
CHEMDRAW and their biological activity values were changed into corresponding
negative log IC50 values. The chemical structures of all the series were imported into
the TSAR window via.mol files. The structures were converted to their respective 3D
structures by CORINA make 3D, a method developed by Gasteiger, Rudolph and
Sadowski.311 All the structures were subjected to COSMIC energey calculation which
70
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

calculates molecular energies by summing bond length, bond angle, torsion angle,
Van der Waals and coulombic terms for all appropriate sets of atoms.. Charges were
evaluated by charge 2 option.312

Defining Substituents
Since substituents, i.e., atoms or group of atoms present at different positions of the
parent structure affect the biological activity, the substitutions of each series were
defined using “define substituent option available within TSAR”.
1. Series 1 comprised of three classes of compounds i.e. β alanine, isoserine and
thiazole. In TSAR package, it is necessary to have same nucleus and same
number of substitution for all the compounds. Out of 70 compounds, 66 had
common substructure of 4-(aminomethyl)-N-methylbenzamide, with three
substituents which were defined (Table 3.1). Four compounds, a61, c24, c27
and c28 were excluded from the series.
2. Series 2 comprised of 30 pyrrolidine carboxamide derivatives as 11β
HSD1inhibitors. 28 compounds appeared with common substructure of 2-
(formylamino)adamantane. 2-(formylamino) adamantane with two substitutes
which were defied (Table 3.1).
3. Series 3 consisted of bicyclo[2.2.2]octyltriazole derivatives as 11β
HSD1inhibitors. 1-methylbicyclo[2.2.2.]octane was considered as the main
nucleus and around it two substitutes were defined (Table 3.1).

Descriptor Calculation
TSAR offers calculation of number of descriptors which includes molecular surface
area, molecular volume, moments of inertia, ellipsoidal volume, Verloop parameters,
lipole moments, dipole moments, molecular mass, molecular connectivity indices,
molecular shape indices, Wiener index, electrotopological state indices, log P, number
of defined atoms (carbon, nitrogen, etc.), rings (aromatic and aliphatic) and groups
(methyl, hydroxyl, etc.). VAMP parameters (electrostatic parameters) like electronic
energy, nuclear repulsion energy, total energy, accessible surface area, atomic charge,
heat of formation, total dipole, polarizability, and dipole components can also be
calculated using TSAR. In the initial approach more than 250 molecular descriptors
for the whole molecule as well as substituent’s belonging to each series were
calculated.313
71
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.1: Lead structures and the substitution pattern of all three series used in QSAR
study

Substituents used in the original series Re-defined substituents

O O

HO N R1
H

N R2

O
Series 1( β O O

alanine,
HO N R1
H
isoserine and H O
R1 N N

thiazole ) R3

O
R3 N R1
(GRA) H
O O N
R2

HO N R1
H

N N R2

R3

Series 2 H
N H
N N R1
(Pyrrolidine R1

carboxamide) O R2 O

(11β HSD1
inhibitors)

N
Series 3
(Bicyclo[2.2.2] N N

R1
octyltriazole) R2

(11β HSD1 X
n N

inhibitors)
N N

Data Reduction
Out of large number of descriptors only few are significantly correlated with the
activity and many of the descriptors are inter-correlated which causes negative effect
on interpretability of the final model. Reduction of irrelevant descriptor is an

72
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

important step to obtain the statistically optimized and meaningful QSAR model. The
correlation matrix was generated to observe the inter-correlation between the
descriptors and correlation between descriptors and biological activity. The
descriptors with high inter-correlation were examined for their correlation with
biological activity and the descriptors with low correlation were discarded. This
restricted false prediction of the QSAR model, as high co-linearity between
descriptors can cause statistical instability and over prediction which makes the
interpretation of mechanism difficult. Descriptors having distributions of values
clumped about a few distinct values were also removed since these do not explain
continuous variation of activity.

Training and Test Set

The preparation of training set and test set is an important step in determining the
quality of the model. All the three data sets were divided into training set and test set
in such a way that both contains structurally diverse, active and inactive compounds.
the training set is used for the development of the model whereas the test set is
employed to evaluate the prognostic ability of the model. During the process of model
development, some compounds were found not to fit in training set. These were
detected as outliers since their residual value was more than two orders of magnitude.
Owing to this, these compounds were finally omitted from training set. Numbers of
training and test set models considered for all the three series are shown in Table 3.2:

Table 3.2: Training and test set data for all three series under investigation

Number of Number of
Number of
Total number compounds in compounds
Series compounds in
of compounds the training identified as
the test set
set outliers

Series 1( β alanine, isoserine and

66 52 9 05
thiazole )

Series 2 (Pyrrolidine carboxamide

28 17 08 03
Inhibitors of 11β HSD1)

Series 3 (Bicyclo[2.2.2] octyltriazole

32 22 8 02
Inhibitors of 11β HSD1)

73
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Statistical Analysis
TSAR provides a wide range of statistical methods which can be used in the
development predictive QSAR models. The QSAR techniques used in TSAR includes
multiple linear regression (MLR), partial least square regression (PLS) and neural
network analysis (ANN). The output is displayed in the form of a model highlighting
substituent points that are strongly correlated with the pharmacological properties
under investigation. The TSAR methodology assumes that a proper distribution of
chosen descriptors facilitates all the information required to understand the
mechanism of action.314

Multivariate Linear Regression (MLR) Analysis

Multiple linear regressions (MLR) establish the relationship between the biological
activities as a linear function of all the structural parameter (descriptors).315 The MLR
models were developed for all the three series and the significance of the models were
determined by the correlation coefficient, which is estimated on the basis of the
training set. Models with correlation coefficient (r2) value (correlation between
dependent and independent variables) of more than 0.8, cross validation r2cv of more
than 0.6, f value higher than 20, which states the degree of statistical confidence and
the s value (standard error of estimate) less than 0.4; were considered to be a
statistically relevant. Since a test set determines the predictivity of the developed
model, test set prediction was exercised.316 A sound model should have the value of
correlation coefficient r2 of test set greater than 0.6 and less than r2 of the training set.

Partial Least Square (PLS) Analysis

Partial Least Squares (PLS) is a linear regression method suitable for conquering the
problems in MLR related to multicollinear or over-abundant descriptors. It has been
suggested as an option to enlarge the information content in each model and avoid the
danger of over fitting. As an approach to check the robustness, chance correlation and
the prediction ability of the developed MLR models, partial least square (PLS)
analysis was performed on the same training set of compounds and descriptors. The
result of PLS was evaluated on the basis of correlation coefficient r2, r2cv of training
and test set.317

74
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Forward Feed Neural Network (FFNN Analysis)

FFNN analysis is based on the architecture of the network of neurons. The design of
this analysis is based on the number of input and output layers.318 This specific
analysis includes unidirectional flow of information, i.e., the information flows during
the prediction from the input descriptors, through a set of layers, to the output of the
network. It has been observed that many a times, neural network may turn out to be
more predictive, particularly for large and diverse datasets. In such analysis,
dependent variables are predicted from a set of explanatory variables through a
trained net. The input layer in the net receives input data and transmits to the
processing neurons with the help of nodes. The functioning is similar to biological
neurons. The neural net in TSAR uses an identity activation function. A weighting
value of 0.03 is applied to all connections. Then a Monte Carlo algorithm is run to
select a better set of starting weights within the constrain limits. The descriptors
selected in MLR and PLS linear models were used to build FFNN models. Models
with different net configurations were generated to improve the RMS error and the
predictive power of the model. Although the number of variable were the same in the
linear and the FFNN model, there were more adjustable parameters in the FFNN
model, since each connection is considered as adjustable.
The soundness of the devloped FFNN models were evaluated by the value of the
correlation coefficient and r2cv. The convergence of the data during the calculation
was observed and the dependencies of the output variable for each of the input
variable, as a series of dependence plots, were analyzed.

Validation of Model
Validation is a vital feature of any QSAR modeling. Reliability and relevance of the
QSAR model is established by this specific process. It judges the predictive ability of
a model on a different data and helps in determining the complexity of an equation. A
model developed by TSAR can be validated by following methods.
1. Cross Validation (r2cv): It is the most common method of validating a QSAR
model. This method includes repeating the regression many times on subsets
of data by leaving out once in turn and r2 is calculated using the predicted
values of the left out molecule. Cross-validation test (r2cv,) value for very
robust model is usually smaller than the overall r2 for a QSAR equation and

75
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

should be greater than 0.50.319 Cross validation was carried out for all the
developed models and the values were analyzed.
2. Test Set Validation: This method includes comparison of the predicted and
observed activities of the test set of compounds. Structure generation,
optimization and charge derivation for the test set compounds were done in the
same way as that of the training set compounds. The predicted activities of the
test set were calculated using the model produced by the training set.
Predictivity of the developed models was assessed on the basis of r2 value. In
general, r2 test value should be greater than 0.6 and smaller than r2 of the
training.320
3. Overfitting: A model can be suffer with over-fitting if it includes more
descriptors than required. In order to avoid over-fitting, small number of
descriptors were used for model development.321
4. Outliers: Compounds having unpredicted biological activity and are incapable
to fit in a QSAR model are known as outliers. Removing outliers from the
main data set can settle the problem. Calculation of residual values and
standardized residuals helped to find the outliers in data set. The compounds
with higher residual values were deleted as outliers.

Results and discussions

Series 1: β alanine, isoserine and thiazole (GRA)
Dataset of 66 β alanine, isoserine and thiazole derivatives with GRA activity
(IC50nM) used in present QSAR study (MLR, PLS, and FFNN) is shown in Table 3.3.
Whole data set of 66 compounds was divided into 52 training set and 9 test set
compounds. The design of training set and test set was done by keeping in mind that
the training set must contain most active and least active compound. Five compounds
were identified as outliers and were deleted.

76
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.3: Structures and biological activities of 66 glucagon receptor antagonists

β alanine, isoserine and thiazole derivatives

R3 N R1
H
N
R2

Basic nucleus

Comp. Configuration R1(Substitution1) R2(Substitution1) R3(Substitution3) IC50 (nM)

S
O
a24 - O 116
HO
Cl

H
N O
F
a25 Trans O 55
O F
HO
F

H
N O
F
a26 Cis O 47
O F
F HO

H
N O
a27 Trans O
1045
N HO

H
N Cl O
a28 Trans 226
O
Cl HO

H
N
O
a29 Trans O 35
HO

F
H
N F
F O
a30 Trans O 56
HO
F F F

77
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

H
N O
a31 O N 2429
O HO

H
N O
a32 140
O
O HO

F F

F
O O
H
a33 N 215
HO
O
N

Br
H
N O
a34 O
F 153
O F HO
F

H O
N
a35 1088
O HO

F F
F
O
H
a36 N 182
O HO
O

N
O
H
a37 N 56
F
F
O
O
HO
F

H Cl
N O
a38 O 80
HO
Cl

H
a39 N O
141
O
OCF3 HO

O N H O
N
a44 782
O
OCF3 HO

78
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

O H
N O
a45 1984
O
OCF3 HO

H
N O
a46 59
O
OCF3 HO

H
N O
a47 122
O
OCF3 HO

H O
a48 O N
528
O
OCF3 HO

O H O
O N
a49 1207
O
OCF3 HO

H
N O
a50 96
O
OCF3 HO

H
N O
a51 137
O HO
OCF3

H
N O
a52 571
O
OCF3 HO

H
N O
a53 O + O 167
N
O
HO

O
F
a54 F 262
S F HO

79
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

O
a55 179
HO

O
F
a56 F 61
O F
HO

Cl
O
a57 27
HO
Cl

H
N O
a62 1700
O
OCF3 HO

H
N
H
N
a65* HO 2960
O
OCF3 O

Me
Me Me H
N O
b11 36
O HO
OCF3

H
N Cl O
b13 O 12
HO
Cl

H O
N Cl

b15* R O HO 3.9

Cl OH

H
N Cl
O

b16 S O HO 51
Cl OH

80
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Me
H
Me Me N CF3
O
b17 Trans O 16
CF3
HO

Me H
Me Me N CF3 O

b18 R, trans O HO 8

CF3 OH

Me
Me Me H
N CF3
O

b19 S,trans O HO 82
CF3 OH

H O
N Cl

b20 R O HO 693
Cl OH

Me
Me Me H O
N Me

b21 O HO 27
Me
OH

O
H
N Br
b22 HO 8
O

OH

H O
N CF3

b23 O HO 7
Br OH

H
O
N CF3

b24 O
HO 14
OMe
OH

81
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

H O
N CF3

b25 O HO 9
OMe OH

Me
Me Me H O
N CF3

b26 Trans O HO 10
Me OH

H
O
N
b27 HO 166
O
OH

H
O
N Cl
b28 HO 14
O
OH

H O
N
b29 HO 6
O
SCF3
OH

H
O
N CF3
b30 HO 9
O
CN
OH

O
H
N
b31 HO 18
O
OH

H O
N Cl

b32 O
HO 4
Cl
OH

H F F O
N
O
b33 O F HO 6
O
F
OH

82
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

H
N CF3
O

b34 O HO 13
F OH

O
H
N SMe
b35 HO 9
O
OH

H
O
N SO2CF3
b36 HO 5
O
OCF3
OH

OCF3

N
O
c6 59
S HO
H

CF3

N
O
c7 64
S
HO
H

OCF3
O
N
c8 96
S
HO
H

Cl
Cl O
c10 N
54
S
H
HO

CF3
O
N
c11 48
S
HO
H

H3C CH3 CF3

O
CH3 N
c12 93
S
H
HO

83
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Cl CF3

N
O
c16 53
S HO
H

Cl OCF3
O
N
c17 53
S
HO
H

CF3 Cl

N
O
c18 93
S HO
H

OCF3 CF3
O
N
c19 51
S
HO
H

MLR Analysis
The linear regression analysis was executed on the dataset having GRA activity as
dependent variable and reduced pool of descriptors as the independent variable. The
three descriptors namely Verloop L, Dipole moment X component and Log P; having
minimum inter-correlations among them were left after data reduction (Table 3.4).

Table 3.4: Correlation matrix of the independent variables used in the final model illustrating
the degree of correlation

X1: Verloop L X2:Dipole Moment X X3: Log P

Descriptors Y: -log Ic50
(Subst.3) Component (Subst.2)
(Whole Molecule)
X1: Verloop L
1 -0.0999 0.16488 -0.70929
(Subst.3)

X2: Dipole
Moment X
0.16488 1 -0.10984 0.36158
Component
(Subst.2)

X3: Log P
0.16488 -0.10984 1 0.26705
(Whole Molecule)

Y: -log Ic50 -0.70929 0.36158 0.26705 1

During the whole processes of model development, five molecules (a36, a45, b11,
b13 and c12) were detected as outliers having standardized residuals greater than

84
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

±1.7. Removal of these five molecules yielded the model having the |r2-r2cv| of just
0.04 exhibiting good predictive ability of the model (Equation 1 and 2).
Original Data: Y = = -1.147*X1 + 0.101*X2 + 0.344*X3 + 1.143 Eq. (1)
Standardized Data: Y = -0.56*S1 + 0.25*S2 + 0.32*S3 - 1.827 Eq. (2)
Where, X1 = Verloop L (Subst. 3)
X2 = Dipole Moment X Component (Subst. 2)
X3 = Log P (Whole Molecule)
r = 0.87, r2 = 0.76, r2cv = 0.72, F = 52.88, s = 0.37
Residual sum of square = 6.921, Predictive sum of square = 8.119

r2 of 0.76 explains 76% variance in biological activity, which is a measure of good fit
by the regression equation. The model was cross-validated by leave one out method
resulting in r2cv value of 0.72 which is an excellent statistical validation of the
developed model. All other statistical value such as t-value, jackknife SE and
Covariance SE as shown in Table 3.5 also confirmed the significance of each and
every descriptor for GRA activity. The exclusion of any of the descriptor from the
model resulted in a very poor model.

Table 3.5: Statistical data of the independent variable illustrating the significance in terms of
statistical parameters

Descriptor Jacknife SE Covariance SE t-value t-probability

Verloop L (Subst.3) 0.11383 0.10976 -10.4849 5.2513

Dipole Moment X
Component 0.021454 0.021499 4.6931 2.2729
(Subst.2)

Log P (Whole 0.064656 0.058892 5.7658 5.7012

Molecule)

The 9 test set compounds which were left a side during model development were used
to check the external predictive ability of QSAR model. A QSAR model with r2 value
of more than 0.6 for the test set is considered to be good and in present case r2 value
for test set compounds was found to be 0.74, which clearly indicates that the
developed model is highly predictive and statistically significant.

85
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

PLS Analysis
To further confirm the robustness and prognostic ability of generated model PLS
analysis was executed on the same data set. The results of the PLS as shown in
Equation 3 also were evaluated on the basis of r2 and statistical significance of the
model.

Y = -1.145*X1+ 0.102*X2+ 0.345*X3+ 1.129 Eq. (3)

Statistical significance (1.10), residual sum of square (12.27), predictive sum of

squares (13.84), E-statistics (0.89), r2cv (0.73), fraction of variance r2 (0.76)
The statistical output of developed PLS model in terms of r2cv (0.73) clearly shows
that all the three descriptors identified during MLR model development are highly
significant and are contributing heavily towards GRA activity of the compounds
under study.

NN Analysis
The multiple-layer NN functionality, under a supervised training by the back
propagation, was used. The number of neurons in the hidden layer and the number of
rows in the training set were balanced to retrieve the best predictive power for the
neural network. In the present case the neural network run with 1 hidden nodes and
70% of training set and 30% of the test set generated the best correlation coefficient
(r2) value. The statistics obtained for the NN treatment were N = 52, input columns
(descriptors) = 3, net configuration = 3-1-1 (3 input nodes, 1 processing node, 1
output node), with test rms = 0.11, best rms = 0.086 and r2 = 0.76 for training and r2 =
0.73 for test. It is noteworthy that NN with same set of descriptors produced model
with equal statistical strength to that of MLR and PLS models. The QSAR model
obtained exhibited negative dependency on the Verloop L (Subst. 3); positive
dependencies on Dipole moment X component (Subst. 2) and Log P (Whole
molecule). The dependency plots are depicted in figure (Figure 3.4-3.6). The actual
and estimated value obtained from MLR, PLS and FFNN analysis of the training and
test set of compounds are shown in Table 3.6-3.7 and their related graphs are shown
in (Figure 3.1-3.3).

86
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.6: Actual and predicted activity data obtained from MLR, PLS and NN analysis of
training set compounds

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

1 a24 -2.06446 -2.14017 -2.13998 -2.00156

2 a30 -1.74819 -2.03367 -2.0312 -1.917

3 a31 -3.38543 -3.30584 -3.3089 -3.41749

4 a32 -2.14613 -2.47897 -2.47724 -2.61452

5 a33 -2.33244 -2.27268 -2.27116 -2.23474

6 a34 -2.18469 -1.85767 -1.85435 -1.72193

7 a35 -3.03663 -2.67848 -2.67724 -2.90047

8 a37 -1.74819 -1.81246 -1.80787 -1.75799

9 a38 -1.90309 -2.27963 -2.27764 -2.29425

10 a39 -2.14922 -2.30411 -2.30202 -2.34473

11 a44 -2.89321 -2.91345 -2.91301 -3.1734

12 a47 -2.08636 -1.70928 -1.71265 -1.49501

13 a48 -2.72263 -2.76156 -2.76262 -2.83115

14 a49 -3.08171 -2.29391 -2.29383 -2.45932

15 a50 -1.98227 -2.34724 -2.34711 -2.27614

16 a51 -2.13672 -1.8395 -1.83682 -1.64588

17 a52 -2.75664 -2.60874 -2.6074 -2.82565

18 a53 -2.22272 -2.4227 -2.42429 -2.44898

19 a54 -2.4183 -2.27804 -2.27555 -2.32942

20 a55 -2.25285 -2.48188 -2.48051 -2.58446

21 a56 -1.78533 -1.95784 -1.9552 -1.81092

22 a57 -1.43136 -1.41635 -1.41555 -1.32214

23 a62 -3.23045 -2.69111 -2.69413 -2.65478

24 a65 -3.47129 -3.21851 -3.22098 -3.29112

87
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

25 b15 -0.59106 -1.59895 -1.60125 -1.4537

26 b16 -1.70757 -1.43326 -1.43577 -1.3085

27 b18 -0.90309 -1.44202 -1.44439 -1.25404

28 b19 -1.91381 -1.23308 -1.23576 -1.10635

29 b20 -2.84073 -2.67382 -2.67393 -2.75654

30 b21 -1.43136 -1.73664 -1.74029 -1.56933

31 b22 -0.90309 -1.65715 -1.66096 -1.46087

32 b23 -0.8451 -1.28533 -1.28785 -1.15831

33 b24 -1.14613 -1.29739 -1.29776 -1.26776

34 b25 -0.95424 -1.11504 -1.11477 -1.12258

35 b26 -1 -1.18742 -1.18848 -1.12332

36 b27 -2.22011 -1.59769 -1.59878 -1.56616

37 b28 -1.14613 -1.33009 -1.33052 -1.29981

38 b29 -0.77815 -0.83988 -0.83852 -0.9266

39 b30 -0.95424 -0.92645 -0.92507 -1.01938

40 b31 -1.25527 -0.77953 -0.77962 -0.89393

41 b32 -0.60206 -1.02955 -1.02975 -1.04307

42 b33 -0.77815 -0.8008 -0.80015 -0.91337

43 b34 -1.11394 -1.18128 -1.18148 -1.16183

44 b36 -0.69897 -0.47993 -0.47749 -0.79001

45 c6 -1.77085 -1.66819 -1.66691 -1.3399

46 c7 -1.80618 -1.26926 -1.27065 -1.10791

47 c8 -1.98227 -1.95505 -1.95427 -1.67506

48 c10 -1.73239 -1.15693 -1.15769 -1.02124

49 c11 -1.68124 -2.12369 -2.12313 -1.8796

50 c16 -1.72428 -1.31074 -1.31242 -1.14836

88
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

51 c18 -1.96848 -1.96578 -1.96632 -1.74472

52 c19 -1.70757 -1.9927 -1.99201 -1.70784

Table 3.7: Actual and predicted activity data obtained from MLR, PLS and NN analysis of test
set compounds

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

1 a25 -1.74036 -2.75524 -2.75886 -2.69886

2 a26 -1.6721 -2.66704 -2.66952 -2.61811

3 a27 -3.01912 -3.42608 -3.43099 -3.48158

4 a28 -2.35411 -2.81244 -2.81587 -2.8267

5 a29 -1.54407 -2.23843 -2.23635 -2.22534

6 a46 -1.77085 -2.79171 -2.79311 -2.85297

7 b17 -1.20412 -2.34373 -2.34385 -2.20567

8 b35 -0.95424 -1.60064 -1.60151 -1.52741

9 c17 -1.72428 -2.06315 -2.06249 -1.82196

Figure 3.1: Plot of actual activity versus estimated activity of training and test set obtained
from MLR analysis
89
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.2: Plot of actual activity versus estimated activity of training and test set obtained
from PLS analysis

Figure 3.3: Plot of actual activity versus estimated activity of training and test set obtained
from NN analysis

90
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.4: Dependency graph illustrating correlation between the Verloop L (subst.3) used
to train the neural network architecture versus the actual activity data

Figure 3.5: Dependency graph illustrating correlation between the Dipole moment X
component (subst.2) used to train the neural network architecture versus the actual activity
data

91
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.6: Dependency graph illustrating correlation between the Log P (whole molecule)
used to train the neural network architecture versus the actual activity data

Interpretation of Descriptors Used in MLR, PLS and NN Model

From the QSAR Equation 1, 2, 3 and Figure 3.4-3.6, it is clear that the descriptors
Dipole moment X component (Subst.2) and Log P (Whole molecule) are positively
and Verloop L (Subst.3) is negatively correlated with glucagon receptor anatagonist,
which means that the glucagon receptor anatagonistic activity will increase with the
increase in Dipole moment X component ( substitution 2) and Log P values of whole
molecule respectively and the glucagon receptor anatagonistic activity will decrease
with an increase in value of Verloop L for substitution 3.

Verloop L (Subst. 3)
It is a steric parameter that describes the bulkiness of the compound and is based on
the standard substitution of bond lengths and angles of the compounds.322 Verloop L,
the length parameter, is defined as the length of the substituent along the axis of its
substitution with the parent. From the regression equation and dependency plot it is
clear that Verloop L parameter is negatively correlated with the biological activity
indicating that decrease in the length of substituent 3 will increase the GRA activity
because high length of substituent orients the molecule in such a way that there is
minimum binding of ligand to receptor. It can be explained with the help of the
92
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

chemistry of substituent 3 of all compounds under consideration. There are three

different functional groups present as substituent 3; acetic acid, 2-hydroxyacetic acid
and N-hydroxyformamide. A strong acid always has shorter bond length, i.e, more the
s character shorter is the bond length. Thus, 2-hydroxyacetic acid, the most acidic
substituent will have the least bond length and thus smaller Verloop L, which is the
case of most active compound b15. Similarly, N-hydroxyformamide, present in the
least active compound a65 is least acidic due to presence of amide group and thus has
higher bond length and thus strong Verloop L character. The acidity of acetic acid lies
in between these two and hence the length parameter also has moderate effect. The
shift in the value of Verloop L from lowest to highest with the decrease in the
biological activity is tabulated in Table 3.8.

Table 3.8: Correlation of biological activity of active and inactive molecules with Verloop L
descriptors

Biological
Name of Verloop L Substituent on R3
activity IC50
compound (Subst.3) position
(µM)
b15 3.9 3.96829 2-hydroxyacetic acid
Active
compounds b32 4 3.87144 2-hydroxyacetic acid

b36 5 3.92826 2-hydroxyacetic acid

a31 140 4.15298 Acetic acid

Inactive
a45 1984 4.93876 Acetic acid
compounds
a65 2960 5.10859 N-hydroxyformamide

Dipole Moment X Component (Subst. 2)

It is an electrostatic descriptor related to the resultant vectors of molecular charge
distribution in X axis and is affected by the presence of electronegative group. It
explains the charge distribution in the molecule by indicating the polar nature and
orientation behavior of a molecule in an electrostatic field. A positive correlation of
it’s with biological activity reveals that an increase in electronegativity or polarity at
substituent 2 will enhance the activity. This can be explained by considering the most
active b15 and least active a65 compounds. Compound b15 and a65 differs at
substituent 2 by the presence of two highly electronegative Cl groups at 3 and 5 with a

93
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

comparatively weak electronegative OCF3 group at 4 position respectively in both the

compounds. Similarly, other active compounds like b32, b36 and least active
compounds like a31, a45 also support the theory. Thus, it clearly indicates
contribution of electronegativity character and hence the effect of dipole moment X
component on the biological activity. The shift in the value of Dipole moment X
component from highest to lowest with the increase in the biological activity is
tabulated in Table 3.9.

Table 3.9: Correlation of biological activity of active and inactive molecules with Dipole
moment X component descriptors

Dipole
Biological
Name of moment X Substituent on R2
activity IC50
compound component position
(µM)
(Subst.2)
H
N Cl

b15 3.9 5.89083 O

Cl

Active H
N Cl
compounds b32 4 2.44226 O

Cl

H
N SO2CF3
b36 5 5.41165
O
OCF3

H
N

a31 140 -4.25323 O N

H
Inactive N
a45 1984 -2.65035
O
compounds OCF3

H
N
a65 2960 -4.6015
O
OCF3

Log P (Whole Molecule)

It is a partition coefficient parameter which describes the partition of a bioactive
solute between polar and non polar solvents. A positive correlation of Log P with the
biological activity reveals the importance of hydrophobic character of whole
molecule. An increase in hydrophobicity of whole molecule will increase the
biological activity and vice versa. High log P value of all the active compounds like

94
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

b15, b32 and b36 and small log P value of the least active compounds like a31 and
a45 supports the theory (Table 3.10).

Table 3.11: Correlation of biological activity of active and inactive molecules with Log P
descriptors
Biological
Name of Log P
activity IC50
compound (Whole molecule)
(µM)

b15 3.9 5.427

Active
compounds b32 4 5.8643

b36 5 6.7739

a31 140 4.7843

Inactive
compounds a45 1984 3.1558

Figure 3.7: Effect of various descriptors on different substituents of glucagon receptor

antagonist

Conclusion
Specific molecular features important for binding of 66 antagonists to the glucagon
receptor have been identified by the reported 2D QSAR models. The equations of
MLR, PLS and dependency plots obtained from NN provide highly valuable

95
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

information about the importance of structural features encoded by physical/chemical

molecular descriptors, such as Verloop L, Dipole moment X component, and Log P
for anti diabetic activity. The present study will be useful in designing more
promising novel GRA containing lead structure.

Series 2: Pyrrolidine Carboxamide (11β HSD1 Inhibitors)

Dataset of 28 pyrrolidine carboxamide derivatives with 11β HSD1 inhibitory activity
(IC50nM) used in present QSAR study (MLR, PLS, and FFNN) is shown in Table
3.11. The data set of 28 compounds was divided into 17 training and 8 test set
compounds. The design of training set and test set was done by keeping in mind that
the training set must contain most active and least active compound. Three
compounds were identified as outliers and were deleted.

Table 3.11: Structures and biological activities of 28 11β HSD1 inhibitors

Pyrrolidine carboxamide derivatives

H
N R1

R2 O

Basic Nucleus

Comp. R1 R2 IC50(nm)

4 H 4.15
N

7 H 2.2

11 N
H 22

12 N H 1.2

96
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

13 N H 3.5

14 N H 0.03

15 N H 0.03

16 N H 0.03

17 N H 0.32

18 N H 0.32
CF3

19 N H 0.32
CF3

Cl
20 N H 0.32

CN
21 N H 0.03

22 N H 0.32
CN

23 N H 20
OH

97
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

24 N H 5
N

25 N H 10
O

OH
26 N H 0.34

OH
N
27 H 0.03

OH

35 H 34
N

OH

36 H 0.8
N

OH

37 H 0.173
N

OH

38 H 1
N

OH

39 H 0.03
N F

98
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

OH

40 H 30
N O

H
H
41 H 19.4
N

42 N
OH 23

43 N OH 31.3

MLR Analysis
The linear regression analysis was executed on the compounds showing 11β HSD1
inhibitory activity as dependent variable and reduced pool of descriptors as the
independent variable. The two descriptors namely Dipole moment X component and
Log P; having minimum inter correlations among them and maximum correlation
with the biological activity were left after data reduction (Table 3.12).

Table 3.12: Correlation matrix of the independent variable used in the final model illustrating
the degree of correlation

X1:Dipole Moment X
Descriptors Component (Whole X2: Log P Y: -log Ic50
Molecule) (Whole Molecule)

X1: Dipole
Moment X
1 -0.170075 -0.612715
Component (Whole
Molecule)

X2: Log P
(Whole Molecule)
-0.170075 1 0.821642

Y: -log Ic50 -0.612715 0.821642 1

During the whole processes of model development, three molecules (15, 17 and 24)
were detected as outliers having residual value greater than ±1. Removal of these
99
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

three molecules yielded the model having the |r2-r2cv| of just 0.03 exhibiting good
predictive ability of the model. Equation 4 and 5 shows good correlation between 11β
HSD1 inhibitors and selected set of two descriptors. Small difference between r2 and
r2cv defines the good prediction ability of the model.
Original Data: Y = -0.211*X1 + 1.418*X2 - 3.091 Eq.4
Standardized Data: Y = -0.558*S1 + 0.847*S2 + 0.179 Eq.5
Where, X1 = Dipole Moment X Component (Whole molecule)
X2 = Log P (Whole Molecule)
r = 0.95, r2 = 0.91, r2cv = 0.87, F = 67.04, s = 0.38
Residual sum of square = 1.9877, Predictive sum of square = 2.76722

r2 of 0.91 explains 91% variance in biological activity, which is a measure of good fit
by the regression equation. The model was cross-validated by leave one out method
resulting in r2cv value of 0.87 which is an excellent statistical validation of the
developed model. All other statistical value such as t-value, jackknife SE and
Covariance SE (Table 3.13) also confirmed the significance of each and every
descriptor for 11β HSD1 inhibitory activity. The exclusion of any of the descriptor
from the model resulted in a very poor model.

Table 3.13: Statistical data of the independent variable illustrating the significance in terms
of statistical parameters

Descriptor Jacknife SE Covariance SE t-value t-probability

Dipole Moment X
Component (Whole 0.0347095 0.0362151 -5.84084 4.28677
Molecule)

Log P (Whole
0.155433 0.160054 8.85972 4.09371
Molecule)

The 8 test set compounds were used to check the prediction ability of the developed
QSAR models. These compounds were treated in a manner analogous to the
compounds in the training set. A QSAR model having r2 value of more than 0.6 for
the test set is considered to be a sound model. The r2 value for test set compounds was
found to be 0.84, which indicate that the developed model is highly predictive and
statistically significant.

100
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

PLS Analysis
To further confirm the robustness and predictive ability of generated model PLS
analysis was executed on the same data set. For a well-defined problem, both MLR
and PLS should have comparable results. The results of the PLS (Equation 6) also
were evaluated on the basis of r2 and statistical significance of the model. The
statistical output of developed PLS model in terms of r2cv (0.88) clearly shows that all
the two descriptors identified during model development are highly significant and
are contributing heavily towards 11β HSD1 inhibitory activity of the compounds
under study. Moreover statistical significance value of 1.09 further testified the
model.

Y = -0.211*X1+ 1.418*X2- 3.091 Eq.6

Statistical significance (1.09), residual sum of square (1.51), predictive sum of

squares (1.84), E-statistics (1.041), r2cv (0.88), fraction of variance r2 (0.90).

NN Analysis
The multiple-layer NN functionality, under a supervised training by the back
propagation, was used. The number of neurons in the hidden layer and the number of
rows in the training set were balanced to retrieve the best predictive power for the
neural network. In the present case, the neural network run with 1 hidden nodes and
75% of training set and 25% of the test set generated the best value of correlation
coefficient (r2).The statistics obtained for the NN treatment were N = 17, input
columns (descriptors) = 2, net configuration = 2-1-1 (2 input nodes, 1 processing
node, 1 output node), test rms = 0.12, best rms = 0.11 and r2 = 0.88 for training and r2
= 0.82 for test. With the same descriptors as used for the MLR and PLS models, the
NN approach showed almost similar results in terms of predictions abilities. The
model exhibited negative dependency on Dipole moment X component (Whole
molecule) and positive dependency on Log P (Whole molecule). The dependency
plots are depicted in figure (Figure 3.11-3.12). The actual and estimated value
obtained from MLR, PLS and FFNN analysis for the training and test set of
compounds are shown in Table 3.14-3.15 and their related graphs are shown in
(Figure 3.8-3.10).

101
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.14: Actual and predicted activity data obtained from MLR, PLS and NN analysis of
training set compounds

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

1 4 -0.61805 -0.706449 -0.706449 -0.712778

2 11 -1.3242 -1.00108 -1.00108 -0.907767

3 12 -0.07918 0.271998 0.271998 0.463881

4 13 -0.54407 -0.0398059 -0.0398059 0.028041

5 14 1.52288 1.21212 1.21212 1.17063

6 16 1.52288 1.35837 1.35837 1.33296

7 21 1.52288 1.35912 1.35912 1.22877

8 22 0.49485 0.794409 0.794409 0.719967

9 26 0.468521 0.736228 0.736228 1.11993

10 27 1.52288 1.36819 1.36819 1.36999

11 35 -1.5315 -1.80526 -1.80526 -1.4278

12 37 0.761954 1.06275 1.06275 1.27573

13 39 1.52288 1.47157 1.47157 1.37597

14 41 -1.2878 -0.979553 -0.979553 -0.993089

15 43 -1.49554 -1.34736 -1.34736 -1.23708

16 36 0.09691 -0.687822 -0.687822 -0.71313

17 19 0.49485 -0.016285 -0.016285 0.228182

Table 3.15: Actual and predicted activity data obtained from MLR, PLS and NN analysis of test
set compounds

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

1 7 0.34242 1.22147 1.22147 1.20164

2 20 0.49485 1.61399 1.61399 1.22619

3 25 -1 -0.187372 -0.187372 0.10685

102
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

4 38 0 1.02864 1.02864 1.19366

5 40 -1.47712 -0.870754 -0.870754 -0.438893

6 42 -1.36173 -2.54055 -2.54055 -1.65935

7 18 0.49485 1.18649 1.18649 1.27882

8 23 -1.30103 -2.10736 -2.10736 -1.57784

Figure 3.8: Plot of actual activity versus estimated activity of training and test set obtained
from MLR analysis

Figure 3.9: Plot of actual activity versus estimated activity of training and test set obtained
from PLS analysis
103
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.10: Plot of actual activity versus estimated activity of training and test set obtained
from NN analysis

Figure 3.11: Dependency graph illustrating correlation between the Dipole moment X
component (whole molecule) used to train the neural network architecture versus the actual
activity data

104
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.12: Dependency graph illustrating correlation between the Log P (whole molecule)
used to train the neural network architecture versus the actual activity data

Interpretation of Descriptors Used in MLR, PLS and NN Model

From the QSAR Equation 4, 5, 6 and Figure 3.11-3.12, it is clear that the descriptors
Dipole moment X component (Whole molecule) is negatively correlated and Log P
(Whole molecule) is positively correlated with 11β HSD1 inhibitory activity, which
means that the 11β HSD1 inhibitory activity will decrease with the increase in Dipole
moment X component of whole molecule and will increase with the increase in Log P
values of whole molecule respectively.

Dipole Moment X Component (Whole Molecule)

It is an electrostatic descriptor related to the resultant vectors of molecular charge
distribution in X axis and is a measure of the polarity of the molecule. It explains the
charge distribution in the molecule by indicating the polar nature and orientation
behavior of a molecule in an electrostatic field. Symmetry of the compounds plays a
major determination factor of the polarity and hence the dipole moment. A negative
correlation of it’s with biological activity reveals that a decrease in polarity of the
whole molecule will enhance the activity. The compounds 14, 15, 16, 21, 27 and 39
having low value of dipole moment X component are most active while compound 35
and 43 with high dipole moment X component are least active (Table 3.16).

105
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.16: Correlation of biological activity of active and inactive molecules with Dipole
moment X component descriptors

Name of Biological activity Dipole moment X component

compound IC50 (µM) (Whole molecule)

14 0.03 0.754892

15 0.03 -2.59325

Active 0.03 0.992423

16
compounds
21 0.03 -4.17145

27 0.03 -3.22158

39 0.03 0.754892

43 31.3 2.62094
Inactive
compounds 35 34 1.72068

Log P (Whole molecule)

A positive correlation of Log P with the biological activity reveals the importance of
hydrophobic character of whole molecule. An increase in hydrophobicity of whole
molecule will increase the biological activity and vice versa. The compounds 14, 15,
16, 21, 27 and 39 having high value of log P are most active while compound 35 and
43 with low log P are least active (Table 3.17).

106
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.17: Correlation of biological activity of active and inactive molecules with Dipole
moment X component descriptors

Name of Biological activity Log P

compound IC50 (µM) (Whole molecule)

14 0.03 3.1472

15 0.03 2.8586

Active 0.03 2.7509

16
compounds
21 0.03 3.2863

27 0.03 2.5224

39 0.03 2.737

43 31.3 1.6206
Inactive
compounds 35 34 1.1634

Figure 3.13: Effect of various descriptors on the whole molecule of 11β HSD1 inhibitory
activity

Conclusion
Specific molecular features important for binding of 28 11β HSD1 inhibitors have
been identified by the reported 2D QSAR models. The equations of MLR, PLS and
dependency plots obtained from NN provide highly valuable information about the

107
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

importance of structural features encoded by physical/chemical molecular descriptors,

such as Dipole moment X component, and Log P for anti diabetic activity. The
present study will be useful in designing more promising novel 11β HSD1 inhibitors
containing lead structure.

Series 3: Bicyclo[2.2.2] octyltriazole (11β HSD1 inhibitors)

Dataset of 32 Bicyclo[2.2.2] octyltriazole derivatives with 11β HSD1 inhibitory
activity (IC50nM) used in present QSAR studies (MLR, PLS, and FFNN) is shown in
Table 3.18. Whole data set of 32 compounds was divided into 22 training and 8 test
set compounds.

Table 3.18: Structures and biological activities of the 32 11β HSD1 inhibitors

Bicyclo[2.2.2]octyltriazole derivatives

R1
R2

Basic Nucleus

Comp. R1 R2 IC50(nm)

1 n-butyl N 4
N N

2 Isobutanol N 35
N N

CF3

3 n-butyl N 1.8
N N

CF3

5 O N 1.2
N N

108
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

CF3

13 MeO N 2
N N

CF3

14 OCH2CH3 N 4
N N

CF3
Cl
15 N 1.1
O
N N

CF3
SO2CH3
16 N 1
O
N N

CF3

17 NHSO2CH3 N 30
N N

CF3

18 NHSO2CH2CH3 N 20
N N

CF3

19 NHSO2(CH2)2CH3 N 8.1
N N

CF3

20 CH2SCH2CH3 N 3.3
N N

CF3

21 CH2SOCH2CH3 N 4.1
N N

CF3

22 CH2SO2CH2CH3 N 7
N N

CF3

23 CH2SO2CH3 N 11
N N

109
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

CF3

24 CH2SO2CH2CH2CH3 N 5.1
N N

CF3

25 CH2SO2CH(CH3)2 N 4.1
N N

CF3

26 CH2SO2C(CH3)3 N 5.4
N N

CF3
O
27 S N 4.4
O
N N

CF3
F O
28 S N 5.5
O
N N

CF3

29 CH2SO2CF3 N 6.1
N N

CF3

30 CH2SO2CH2CF3 N 3.7
N N

CF3

31 SO2CH2CH3 N 25
N N

CF3

32 CH2CH2SO2CH2CH3 N 3.2
N N

CF3

33 CH2CH2CH2SO2CH2CH3 N 1.6
N N

Cl

34 CH2SO2CH2CH3 N 7.5
N N

110
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Cl

35 CH2SO2CH2CH3 N 900
N N

Cl
N
36 CH2SO2CH2CH3 2800
N N

Cl
OH
37 CH2SO2CH2CH3 N 2.8
N N

Br

38 CH2SO2CH2CH3 N 10
N N

39 CH2SO2CH2CH3 N 20
N N

40 CH2SO2CH2CH3 N 140
N N

MLR Analysis
The linear regression analysis was performed using the 11β HSD1 inhibitory activity
as dependent variable and reduced pool of descriptors as the independent variable.
The three descriptors namely Inertial moment length, Log P and VAMP polarization
XX; having minimum inter correlations among them and maximum correlation with
the biological activity were left after data reduction (Table 3.19).
During the whole processes of model development, two molecules (35 and 37) were
detected as outliers having residual value greater than ±0.65. Removal of these two
molecules yielded the model having the |r2-r2cv| of just 0.09 signifying the statistical
strength of the model. Equation 7 and 8 shows correlation between 11β HSD1
inhibitors and selected set of two descriptors
Original Data: Y = -0.163*X1 + 0.271*X2 + 0.037*X3 - 3.095 Eq.7
Standardized Data: Y = -0.491*S1 + 0.250*S2 + 0.192*S3 - 0.928 Eq.8
Where, X1 = Inertia moment 1 length (Subst.2)
X2 = Log P (Whole Molecule)
111
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

X3 = VAMP Polarization XX (Whole molecule)

r = 0.94, r2 = 0.89, r2cv = 0.79, F = 46.87, s = 0.27
Residual sum of square = 1.36056, Predictive sum of square = 2.48005

Table 3.19: Correlation matrix of the independent variable used in the final model illustrating
the degree of correlation

X1:Inertial
X2: Log P X3:VAMP Y:
Moment
Descriptors (Whole Polarization XX
Length -log Ic50
Molecule) (Whole Molecule)
(Subst.2)

X1:Inertial
Moment
1 -0.265945 -0.271725 -0.810608
Length
(Subst.2)

X2: Log P
(Whole -0.265945 1 0.485858 0.625431
Molecule)

X3:VAMP
Polarization
-0.271725 0.485858 1 0.587932
XX (Whole
Molecule)

Y: -log Ic50 -0.810608 0.625431 0.587932 1

r2 of 0.94 explains 94% variance in biological activity, which is a measure of good fit
by the regression equation. The model was cross-validated by leave one out method
resulting in r2cv value of 0.89 which is an excellent statistical validation of the
developed model. All other statistical value such as t-value, jackknife SE and
Covariance SE (Table 3.20) also confirmed the significance of each and every
descriptor for 11β HSD1 inhibitory activity. The exclusion of any of the descriptor
from the model resulted in a very poor model.

Table 3.20: Statistical data of the independent variable illustrating the significance in terms
of statistical parameters

Descriptor Jacknife SE Covariance SE t-value t-probability

Inertial Moment
0.0336276 0.0210388 -7.74538 3.87257
Length (Subst.2)

Log P
0.0599812 0.0751529 3.61263 0.00199047
(Whole Molecule)

VAMP Polarization
XX (Whole 0.0143 0.0133546 2.76035 0.0128846
Molecule)

112
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

The 8 test set compounds were used to check the prediction ability of QSAR model.
All the test compounds were treated in a manner analogous to the compounds in the
training set. A QSAR model having r2 value of more than 0.6 for the test set is
considered to be a sound model. The r2 value for test set compounds was 0.86, which
clearly indicates that the developed model is highly predictive and statistically
significant.

PLS Analysis
To further confirm the robustness and predictive ability of generated model, PLS
analysis was performed using the same data set. For a well-defined problem, both
MLR and PLS should have comparable results. The results of the PLS as shown in
Equation 9 were also evaluated on the basis of r2 and statistical significance of the
model. The statistical output of developed PLS model in terms of r2cv (0.81) clearly
shows that all the three descriptors identified during model development are highly
significant and are contributing heavily towards 11β HSD1 inhibitory activity of the
compounds under study. Moreover statistical significance value of 1.15 further
testified the model.

Y = -0.163*X1+ 0.263*X2+ 0.038*X3- 3.113 Eq. 9

Statistical significance (1.15), residual sum of square (2.38), predictive sum of
squares (4.02), E-statistics (0.47), r2cv (0.81), fraction of variance r2 (0.89)

NN Analysis
The multiple-layer NN functionality, under a supervised training by the back
propagation, was used. The number of neurons in the hidden layer and the number of
rows in the training set were balanced to retrieve the best predictive power for the
neural network. In the present case the neural network run with 1 hidden nodes and
70% of training set and 30% of the test set generated the best value of correlation
coefficient (r2).The statistics obtained for the NN treatment were N = 22, input
columns (descriptors) = 3, net configuration = 3-1-1 (3 input nodes, 1 processing
node, 1 output node), test rms = 0.08, best rms = 0.05 and r2 = 0.91 for training and r2
= 0.75 for test. With the same descriptors as used for the MLR and PLS models, the
NN approach exhibited almost similar predictions abilities. The QSAR model
obtained showed negative dependency on Inertia moment 1 length (Substituent 2) and

113
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

positive dependencies on Log P (Whole molecule) and VAMP Polarization XX

(Whole molecule). The dependency plots are depicted in figure (Figure 3.17-3.19).
The actual and estimated value obtained from MLR, PLS and FFNN analysis of the
training and test set of compounds are shown in Table 3.21-3.22 and their related
graphs are shown in (Figure 3.14-3.16).

Table 3.21: Actual and predicted activity data obtained from MLR, PLS and NN analysis of
training set compounds

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

1 1 -0.60206 -0.51201 -0.51724 -0.43294

2 3 -0.25527 -0.32674 -0.34428 -0.37563

3 14 -0.60206 -0.96936 -0.97368 -0.90564

4 15 -0.04139 -0.20528 -0.21775 -0.30486

5 16 0 -0.11285 -0.09494 -0.23569

6 17 -1.4771 -1.34453 -1.33948 -1.59685

7 18 -1.301 -1.21713 -1.21388 -1.27269

8 21 -0.61278 -0.85909 -0.8504 -0.48917

9 23 -1.04139 -1.0439 -1.0396 -0.81741

10 24 -0.70757 -0.80358 -0.80531 -0.57365

11 26 -0.73239 -0.70385 -0.70216 -0.44056

12 27 -0.64345 -0.17212 -0.16645 -0.25464

13 28 -0.74036 -0.33262 -0.33572 -0.30399

14 29 -0.78533 -0.6769 -0.69221 -0.65538

15 31 -1.3979 -1.14153 -1.14272 -1.23031

16 32 -0.50515 -0.81046 -0.81118 -0.5677

17 33 -0.20412 -0.46536 -0.46031 -0.30981

18 34 -0.87506 -1.23555 -1.23402 -1.24366

19 36 -3.4472 -3.47091 -3.47568 -3.49709

114
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

20 38 -1 -0.95044 -0.94455 -0.58761

21 39 -1.301 -1.45898 -1.45386 -1.34717

22 40 -2.1461 -1.6055 -1.60326 -2.12728

Table 3.22: Actual and predicted activity data obtained from MLR, PLS and NN analysis of test
set compounds

Predicted Activity (-log IC50)

Actual Activity
S. No. Compound
(-log IC50)
MLR PLS NN

1 2 -1.5441 -1.218 -1.21565 -1.08742

2 5 -0.07918 -0.62682 -0.63504 -0.49019

3 13 -0.30103 -0.60943 -0.61327 -0.42972

4 19 -0.90849 -0.93829 -0.93335 -0.63256

5 20 -0.51851 -0.61483 -0.61869 -0.43295

6 22 -0.8451 -1.02385 -1.0246 -0.896

7 25 -0.61278 -0.81063 -0.8116 -0.57046

8 30 -0.5682 -0.67598 -0.67617 -0.44089

Figure 3.14: Plot of actual activity versus estimated activity of training and test set obtained
from MLR analysis
115
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.15: Plot of actual activity versus estimated activity of training and test set obtained
from PLS analysis

Figure 3.16: Plot of actual activity versus estimated activity of training and test set obtained
from NN analysis

116
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.17: Dependency graph illustrating correlation between the Inertia moment1 length
(Subst.2) used to train the neural network architecture versus the actual activity data

Figure 3.18: Dependency graph illustrating correlation between the Log P (whole molecule)
used to train the neural network architecture versus the actual activity data

117
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Figure 3.19: Dependency graph illustrating correlation between the VAMP Polarization XX
(whole molecule) used to train the neural network architecture versus the actual activity data

Interpretition of Descriptors Used in MLR, PLS and NN Model

From the QSAR Equation 7, 8, 9 and Figure 3.17-3.19, it is clear that the descriptors
Inertia moment1 length (Subst.2) is negatively correlated; Log P (Whole molecule)
and VAMP polarization XX (Whole molecule) are positively correlated with 11β
HSD1 inhibitory activity, which means that the 11β HSD1 inhibitory activity will
increase with the decrease in Inertia moment1 length at substituent 2 and will increase
with the increase in Log P and VAMP polarization XX values of whole molecule
respectively.

Inertia Moment 1 Length (Subst. 2)

It is a steric parameter which represents the strength and orientation behaviors of any
compound in an electrostatic field. It defines the biological interaction between ligand
and compound by describing the steric function to a specific axis in the form of
ellipsoidal volume. The negative correlation of the inertia moment 1 length at
substitution 2 shows that presence of complex, non-linear and branched substituents
will enhance the biological activity. It can be explained by the behavior of active
compounds 16, 15 and 5 and inactive compounds 36, 35 and 40. The active
118
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

compounds have trifluoromethyl benzene ring attached to triazole; whereas the

inactive compounds 36, 35 and 40 have halobenzene rings attached to triazole ring at
substitution 2. The complexity and branched characteristic of trifluoromethyl benzene
in comparison to a less complex and unbranched halobenzene is responsible for the
higher activity of the compounds 16, 15 and 5. Further, low values of inertia moment
1 length for the active compounds and high values for the inactive compounds
confirm the fact (Table 3.23).

Table 3.23: Correlation of biological activity of active and inactive molecules with Inertia
moment 1 length descriptors

Biological
Name of Inertia moment 1 Substituent on R2
activity IC50
compound length (Subst.2) position
(µM)
CF 3

16 1 5.74142 N

N N

CF3
Active
15 1.1 5.74547 N
compounds N N

CF 3

5 1.2 5.72429 N

N N

40 140 7.69677 N

N N

Cl
Inactive
35 900 11.8791 N
compounds
N N

Cl
36 2800 19.6661 N

N N

Log P (Whole Molecule)

A positive correlation of Log P with the biological activity reveals the importance of
hydrophobic character of whole molecule. An increase in hydrophobicity of whole
molecule will increase the biological activity and vice versa. The compounds 16, 15
and 5 having high value of log P are most active whereas compounds 36, 35 and 40
with low log P are least active (Table 3.24).

119
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.24: Correlation of biological activity of active and inactive molecules with Log P
descriptors

Biological
Name of Log P
activity IC50
compound (Whole molecule)
(µM)

16 1 5.9216
Active
compounds 15 1.1 7.2946

5 1.2 6.2234

40 140 4.3435
Inactive 900 4.722
35
compounds
36 2800 4.722

VAMP Polarization XX (Whole Molecule)

VAMP Polarization XX is a spatial descriptor which calculates the electronic
properties of a compound and projects polarization towards XX planes. The
polarizability is directly proportional to the total number of valence electrons on each
atom. It is a measure of weak intermolecular interactions and represent chemical
reactivity index of a compound at XX plane. A positive correlation of VAMP
polarization XX with the biological activity reveals the direct link of chemical
reactivity index with the biological activity. The compounds 16, 15 and 5 having high
value of VAMP polarization XX are most active whereas compounds 36, 35 and 40
with low VAMP polarization XX are least active (Table 3.25).

120
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

Table 3.25: Correlation of biological activity of active and inactive molecules with VAMP
polarization XX descriptors

Biological
Name of VAMP Polarization XX
activity IC50
compound (Whole molecule)
(µM)

16 1 62.6568
Active
compounds 15 1.1 50.0554

5 1.2 46.4158

40 140 42.432
Inactive 900 42.103
35
compounds
36 2800 41.9509

Figure 3.20: Effect of various descriptors on the whole molecule of 11β HSD1 inhibitory
activity

Conclusion
Specific molecular features important for binding of 32 11β HSD1 inhibitors have
been identified by the reported 2D QSAR models. The equations of MLR, PLS and

121
 CHAPTER-3

GLOBAL DESCRIPTOR BASED QSAR

dependency plots obtained from NN provide highly valuable information about the
importance of structural features encoded by physical/chemical molecular descriptors,
such as Inertia moment 1 length, Log P and VAMP polarization XX for anti diabetic
activity. The present study will be useful in designing more promising novel 11β
HSD1 inhibitors containing lead structure.

122
 CHAPTER- 4
Ligand Based
Pharmacophore (LBP)
Modeling
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

3D QSAR Based Pharmacophore Modeling

A pharmacophore model is a three dimensional spatial arrangements of the chemical

features involved in the drug-receptor interaction. A pharmacophore can be used; as a
query for screening of chemical compound databases, for designing compounds with
specific desired features (lead optimization), and for evaluating similarity and
diversity of compounds using pharmacophore fingerprints. It can also be used to align
compounds based on the 3D arrangement of chemical features or to develop
predictive 3D QSAR models (Figure 4.1).

Figure 4.1: Pharmacophore modeling workflow

123
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Materials and Methods

Pharmacophore modeling is the most commonly used technique for classification and
identification of essential features from a group of molecules comprising active and
inactive compounds. In order to construct predictive pharmacophore, three
structurally diverse compounds datasets of GRA and 11β HSD1 inhibitor were
collected. The compounds were sketched using ISIS Draw 2.5 and energy minimized
using CATALYST software to the closest local minimum by using the CHARMM
force field. Since conformation generation is a prerequisite for three dimensional
pharmacophore modeling, 250 conformations for each compound were generated
using 'Poling' algorithm with an energy threshold of 20 kcal/mol above minimum
global energy value.

Selection of Training Set and Test Set323

Development of a sound pharmacophore model depends on the selection of training
set compounds. A set of rules have been framed for proper selection of structural and
biological data for pharmacophore modeling:
1. The training set compounds should have diversity in their structure and activity
values.
2. The activity data of the compounds should be generated from comparable binding
assays.
3. The activity range should be at least 3.5–5 orders of magnitude.
4. The compounds should be selected in order to add novel perspective to the model
while avoiding redundancy and bias both in terms of structural features and activity
range.
5. The most active and least active compounds should be included so that they can
give an insight on the active and poor features of a pharmacophore.
On the basis of above mentioned cardinal rule all the compounds of the three series
were divided into training set and test set. Numbers of compounds considered in the
training and test set for all the three series are shown in Table 4.1.

124
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Table 4.1: Training and test set data for all three series under investigation

Number of Number of
Number of
Total number compounds in compounds
Series compounds in
of compounds the training identified as
the test set
set outliers
Series 1( β alanine, isoserine and
70 56 13 01
thiazole antagonist of GRA)
Series 2 (Pyrrolidine carboxamide
30 20 10 00
Inhibitors of 11β HSD1)
Series 3 (Bicyclo[2.2.2]
octyltriazole Inhibitors of 11β 32 24 8 00
HSD1)

Dataset Preparation and Pharmacophore Generation

Three dimensional pharmacophore modeling was performed on CATALYST 4.9
platform (Accelrys Inc., San Diego, CA). The important feature of this software is the
HypoGen algorithm which identifies the hypotheses identical to the “active”
compounds in the training set and at the same time subtracts the hypothesis identical
to the “inactives”. During the process of pharmacophore generation the parameters
such as function weight, mapping coefficient, activity uncertainty were kept 0.2, 0,
297 pm, and 3 respectively. An uncertainty Δ indicates the activity value of any
compound lying somewhere in the range of “activity divided by Δ” to “activity
multiplied by Δ”. Using different combinations of the the most relevant features and
the aforementioned parameter values various pharmacophore hypothesis were
generated and evaluated on the basis of statistical outputs and fitness.

Statistical Assessment of the Generated Hypotheses

The generated pharmacophore model was evaluated on the basis of statistical results.
The important statistical parameters considered in pharmacophore study included root
mean square (RMS), correlation coefficient, fixed cost, null cost, total cost,
configuration cost, error cost and weight cost. A fixed cost value is the lowest
possible value and denotes the perfect fit of the model.288 A null cost represents a zero
hypothesis with no features. The “total cost” is the summation of three cost
parameters: weight, configuration and error. The error cost is directly proportional to

125
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

the root mean-square (RMS) difference between the estimated and the actual affinity
for the training set increases. The RMS value predicts the quality of prediction of the
model. The weight cost is directly proportional to the difference between the actual
and ideal weights of the features. The ideal value of the weight is reported to be 2
because higher weight values leads to fitting of abnormal conformations of the
compounds to the features. The third function, configuration cost characterizes
complexity of the model. The configuration cost is a fixed cost that represents the
complexity of the hypothesis space to be optimized. It should not exceed a maximum
value of 17 for a significant pharmacophore model. A difference of 20 bits or more
between null cost and total cost signifies the good correlation of the model. An ideal
hypothesis is characterized by low RMS value, high correlation coefficient (r) and the
highest difference between null and total cost. In addition to these statistical
parameters, the soundness and predictability of the generated pharmacophore models
were also evaluated by fisher randomization test, internal and external test set
validation.

Validation of Pharmacophore Model

Fischer’s Randomization Test

The Fisher’s randomization test is a technique which is used to validate the
correlation coefficient obtained between chemical structures and biological activity.
In this test, the training set compounds are scrambled to create pharmacophore
hypotheses with same features and parameters used in the development of the original
pharmacophore hypothesis. If the randomized data set results similar or better
statistical outputs, the original hypothesis is considered to be obtained by chance. 19
spreadsheets were created for a 95% confidence level, 49 for 98% and 99 for 99%
confidence level and the results were analyzed.

Internal Test Set Validation

A sound and robust pharmacophore model should not only be able to predict the
activity of the training set compounds accurately but should also precisely predict the
activity of test set compounds within the domain. Three different internal test sets
containing diverse structure and activities were used to evaluate the prediction ability
of all the developed models. The predicted activity was calculated on the basis of

126
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

geometric fit of the entire test set compounds mapped onto the best pharmacophore.
The geometric fit value is calculated on the basis of number of pharmacophore
features successfully superimposed to the corresponding functional groups, the weight
of the corresponding hypothesis features spheres, the distance between the center of a
particular pharmacophoric sphere (feature centroid) and the center of the
corresponding superimposed chemical moiety of the fitted compound.324 The squared
correlation coefficient (r2) values were determined and evaluated. Moreover the
mapping pattern of the most active compound was compared with least active
compound and the results were analyzed.

External Test Set Validation

A true pharmacophore should be capable of predicting activity of the compounds
outside the domain. Hence the developed pharmacophore models were validated
using structurally diverse external test sets. During the process the chosen external
test set compounds were mapped onto the generated pharmacophore model. The
difference between estimated and actual activity for all the compounds were critically
evaluated.

Result and Discussion

Series 1: β alanine, isoserine and thiazole Derivatives as GRA

Pharmacophore Generation
A series of 70 structurally diverse compounds having GRA activity from 3.9 nM to
2960 nM was used for development of pharmacophore model (Table-4.2). Out of 70
compounds a training set of 57 compounds was used to build the pharmacophore
model, while 13 compounds were kept as internal test set to validate the predictivity
of the generated model. Compound a24 showed an error ratio, so it was removed from
the dataset (training set) during the course of pharmacophore model development.

127
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Table 4.2: Structure and biological activity of β-lanine, isoserine and thiazole core
derivatives with glucagon receptor antagonistic activity

Glucagon receptor antagonist

O O O
HO
HO N N
H H H
N N O N N
F F
F F
O O
O F O F

a61, IC50 = 1000 a62, IC50 = 1700

O O
O
H HO N
N H H
HO N OMe N N Cl
H H
O N N
F O
F
O
O F Cl

a65, IC50 = 2960 b20, IC50 = 693

β alanine derivatives

O O

HO N R1
H
N R2

β alanine moiety O

a24-a57, b11, b13, b17

Comp. Configuration R1 R2 IC50 (nM)

a24 - O 116

Cl

H
N
F
a25 Trans O 55
O F
F

128
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

H
N
F
a26 Cis O 47
O F
F

H
N
a27 Trans O
1045
N

H
N Cl
a28 Trans 226
O
Cl

H
N

a29 Trans O 35

F
H
N F
F
a30 Trans O 56

F F F

H
N

a31 O N 2429
O

H
N
a32 140
O
O

F F

F
O
H
a33 N 215
O
N

Br
H
N
a34 O
F 153
O F
F

H
N
a35 1088
O

129
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

F F
F
H
a36 N 182
O
O

H
a37 N
F 56
F
O
O F

H Cl
N

a38 O 80
Cl

H
a39 N
141
O
OCF3

O N H
N
a44 782
O
OCF3

O H
N
a45 1984
O
OCF3

H
N
a46 59
O
OCF3

H
N
a47 122
O
OCF3

H
a48 O N
528
O
OCF3

130
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

O H
O N
a49 1207
O
OCF3

H
N
a50 96
O
OCF3

H
N
a51 137
O
OCF3

H
N
a52 571
O
OCF3

H
N

a53 O + O 167
N
O

F
a54 F 262
S F

a55 179

F
a56 F 61
O F

Cl

a57 27
Cl

131
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Me
Me Me H
N
b11 36
O
OCF3

H
N Cl

b13 O 12
Cl

Me
H
Me Me N CF3

b17 Trans O 16
CF3

Isoserine derivative

O O

HO N R1
H
OH N R2

Isoserine moiety O

b15-b16, b18-b19, b21-b36

Comp. Configuration R1 R2 IC50 (nM)

H
N Cl

b15* R O 3.9

Cl

H
N Cl

b16 S O 51
Cl

Me H
Me Me N CF3

b18 R, trans O 8

CF3

132
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Me
Me Me H
N CF3

b19 S,trans O 82
CF3

H
N Cl

b20 R O 693
Cl

Me
Me Me H
N Me

b21 O 27
Me

H
N Br
b22 8
O

H
N CF3

b23 O 7
Br

H
N CF3

b24 O 14
OMe

H
N CF3

b25 O 9
OMe

Me
Me Me H
N CF3

b26 Trans O 10
Me

133
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

H
N
b27 166
O

H
N Cl
b28 14
O

H
N
b29 6
O
SCF3

H
N CF3
b30 9
O
CN

H
N
b31 18
O

H
N Cl

b32 O 4
Cl

H F F
N
O
b33 O F
6
O
F

H
N CF3

b34 O 13
F

134
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

H
N SMe
b35 9
O

H
N SO2CF3
b36 5
O
OCF3

β alanine moiety with Thiazole core derivatives

O O

HO N R1
H
X N
n R2
β alanine moiety S

Thiazole core

c6-c19, c24, c27-c28

Comp. n X R1 R2 IC50 (nM)

OCF3

c6 1 N 59

CF3

c7 1 N 64

OCF3

c8 1 N 96

Cl
Cl
c10 1 N 54

135
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

CF3

c11 1 N 48

H3C CH3 CF3

CH3
c12 1 N 93

CF3
Cl

c16 1 N 53

Cl OCF3

c17 1 N 53

CF3 Cl

c18 1 N 93

OCF3 CF3

c19 1 N 51

Cl

c24 0 N 34

CF3
Cl

c27 1 C 38

Cl OCF3

c28 1 C 25

136
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Pharmacophore Model Development and Statistical Evaluation

Among various generated models (Table 4.3) hypothesis 1 comprising of one HBD,
one HY, one NI and one RA with cost difference of 35.24, correlation coefficient of
0.94, root mean square (RMS) of 1.00, squared correlation coefficient value of 0.89
and configuration of 15.41 was considered as best.
The HypoGen algorithm was used to develop pharmacophore models with chemical
features required for GRA activity. Various pharmacophore models were developed
and assessed on the basis of their cost functions represented in bits unit. Out of many
runs a set of ten hypotheses with fixed cost of 205.21 bits (well separated from the
null hypothesis cost of 259.35 bits) was considered. Among chosen set of ten
hypothesis the top-ranked pharmacophore model (Hypothesis1) had the best
predictive power and statistical significance described by the high correlation
coefficient ( r = 0.89, r2 = 0.80), low root mean- square deviation ( 0.71), weight
(0.92) and error costs (202.66) and a cost difference of 39.84. The configuration cost
was 15.92, which indicates that no over fitting of the data has been exercised. The
cost difference between total and fixed costs for the best hypothesis was only 14.3
bits, indicating the high probability of the true correlation of the data. Thus,
hypothesis1 with four features, viz., one H-bond donor (HBD), one negative ionizable
(NI), one hydrophobe (HY) and one ring aromatic (RA) was considered as the best
pharmacophore model for GRA activity. The cost values, correlation coefficients (r),
RMSD, and features for the top ten hypotheses are listed in Table 4.3.

Table 4.3: Results of the top 10 pharmacophore hypotheses generated by the HypoGen
algorithm

Correlation
Hypothesis Total Cost Cost Difference RMSD Features
(r)

1 219.511 39.836 0.714 0.89 HBD, HY,NI,RA

2 224.185 35.162 0.820 0.85 HBD, HY,NI,RA

3 228.008 31.339 0.900 0.82 HBD, HY,NI,RA

4 228.467 30.88 0.910 0.82 HBD, HY,NI,RA

5 228.56 30.787 0.913 0.82 HBD, HY,NI,RA

6 228.827 30.52 0.916 0.82 HBD, HY,NI,RA

137
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Correlation
Hypothesis Total Cost Cost Difference RMSD Features
(r)

7 229.054 30.293 0.921 0.81 HBD, HY,NI,RA

8 229.582 29.765 0.931 0.81 HBD, HY,NI,RA

9 229,696 29.651 0.916 0.82 HBD, HY,NI,RA

10 259.347 0 1.592 0.81 HBD, HY,NI,RA

Hypothesis1, identified as the best hypothesis, was used to estimate the activity of the
training set molecules. All the training set compounds were classified by their activity
as highly active (<100 nM, +++), moderately active (100-300 nM, ++) and inactive
(>300 nM, +).. Among 56 training set compounds, two active compounds (a30, a37)
were predicted as moderate and inactive respectively; three moderate compounds
(a53, a36, a28) were predicted to be active and inactives respectively; and two
inactive compounds (b20, a45) were predicted as moderate by Hypothesis1.
Consequently, for 49 of 56 training set compounds, the predicted IC50 (nM) values
were within the same activity scale as the experimental values in the training set.
Table 4.4 and Figure 4.2 represent the actual and estimated GRA activity of the 56
training set molecules based on the best hypothesis

Table 4.4: Actual and estimated values of the training set based on the pharmacophore
hypothesis

Actual Estimated Fit Error Activity Estimated

Compound
IC50 IC50 value factor scale Activity scale

a24 116 58.021 5.85 2.1 ++ +++

a28 226 315.795 5.1 1.4 ++ +

a30 56 211.654 5.27 3.8 +++ ++

a31 2429 527.375 4.91 5 + +

a32 140 203.255 5.28 1.5 ++ ++

a35 1088 421.252 4.98 2.6 + +

a36 182 601.772 4.79 3.6 ++ +

138
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Actual Estimated Fit Error Activity Estimated

Compound
IC50 IC50 value factor scale Activity scale

a37 56 340.264 5.07 5.9 +++ +

a39 141 227.122 5.24 1.6 ++ ++

a44 782 384.177 5.02 2.1 + +

a45 1984 289.049 5.11 6.4 + ++

a47 122 281.218 5.16 2.3 ++ ++

a48 528 342.504 5.05 1.5 + +

a49 1207 621.771 4.8 1.9 + +

a51 137 261.74 5.18 1.9 ++ ++

a52 571 896.157 4.63 1.6 + +

a53 167 58.283 5.82 2.8 ++ +++

a54 262 268.528 5.17 1 ++ ++

a55 179 218.257 5.27 1.2 ++ ++

a56 61 60.172 5.8 1 +++ +++

a57 27 45.151 5.94 3.2 +++ +++

a61 1000 1065.04 4.56 1.1 + +

a62 1700 700.417 4.71 2.2 + +

a65 2960 2493.08 4.14 1 + +

b11 36 47.334 5.92 1.3 +++ +++

b15 3.9 3.983 6.98 1.1 +++ +++

b16 51 48.682 5.89 1 +++ +++

b17 16 64.326 5.79 4.1 +++ +++

139
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Actual Estimated Fit Error Activity Estimated

Compound
IC50 IC50 value factor scale Activity scale

b18 8 9.123 6.64 1.1 +++ +++

b19 82 11.233 6.53 7.1 +++ +++

b20 693 114.476 5.52 5.9 + ++

b21 27 84.805 5.66 1.7 +++ +++

b22 8 8.839 6.63 1.2 +++ +++

b23 7 13.36 6.46 2 +++ +++

b24 14 6.588 6.78 2.1 +++ +++

b25 9 11.802 6.67 1.1 +++ +++

b26 10 7.761 6.69 1.2 +++ +++

b27 166 55.09 5.86 3 ++ +++

b28 14 23.294 6.22 1.7 +++ +++

b29 6 5.958 6.84 1 +++ +++

b30 9 8.22 6.57 1.2 +++ +++

b31 18 43.394 5.99 2.3 +++ +++

b32 4 5.946 6.81 1.5 +++ +++

b33 6 8.159 6.68 1.4 +++ +++

b35 9 11.465 6.54 1.3 +++ +++

c10 54 51.46 5.87 1 +++ +++

c12 93 78.983 5.71 1.2 +++ +++

c16 53 53.627 5.87 1 +++ +++

c17 53 49.917 5.89 1 +++ +++

140
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Actual Estimated Fit Error Activity Estimated

Compound
IC50 IC50 value factor scale Activity scale

c18 93 64.785 5.79 1.4 +++ +++

c19 51 51.657 5.89 1 +++ +++

c24 34 70.63 5.75 2.1 +++ +++

c27 38 67.762 5.76 1.8 +++ +++

c6 59 49.32 5.89 1.2 +++ +++

c7 64 63.593 5.79 1 +++ +++

c8 96 47.764 5.91 2 +++ +++

Figure 4.2: Plot of correlation coefficient between actual and estimated activity value of
training set

Pharmacophore Mapping
The obtained pharmacophoric features and their interfeature distances are shown in
Figure 4.3 (a) and 4.3(b). Mapping of the most active compound b15 on the
pharmacophore (Figure 4.3 (c)) reveals that the carboxyl group is mapped to negative
ionizable group, hydroxyl groups to hydrogen bond donor, benzene ring adjacent to
the bulky group to the hydrophobic and benzene ring near the electron withdrawing

141
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

region to ring aromatic group, whereas the inactive compound a65 (Figure 4.3 (d))
missed negative ionizable feature due to absence of carboxyl group.

Figure 4.3: Pharmacophore features (Hypothesis 1) and pharmacophore mappings of GRA:

(a) Four pharmacophore features: one HBD (pink color), one NI (dark blue color), one HY
(light blue color) and one RA (orange color) (b) Interfeature distances between the
pharmacophore features (c) Pharmacophore mapping with most active compound, b15 of
training set (d) Pharmacophore mapping with least active compound, a65 of training set

Validation of Pharmacophore Model

In order to check the accuracy of the model, various validation approaches were
adopted.
i) Internal Test Set Validation: The pharmacophore hypothesis was internally
validated by using the test set comprising of 13 molecules with GRA activity. The test
set compounds were treated alike the training set compounds. A squared correlation
coefficient of 0.73 as shown in Figure 4.4 indicated a good correlation between the
142
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

actual and estimated activities of the test set compounds. The scored estimated
activities and actual value of test set compounds are shown in Table 4.5. The
agreement between actual and predicted activity of test set compound testifies the
soundness of hypothesis 1.

Table 4.5: Actual and estimated values of the test set based on the pharmacophore hypothesis

Actual Estimated Fit Activity Estimated

Compound
IC50 IC50 value scale Activity scale

a25 55 459.746 4.937 +++ +

a26 47 164.763 5.383 +++ ++

a29 35 139.538 5.455 +++ +++

a33 215 405.708 4.992 ++ +

a34 153 295.43 5.13 ++ ++

a38 80 466.187 4.931 +++ +

a46 59 222.964 5.252 +++ ++

a50 96 274.132 5.162 +++ ++

b13 12 45.151 5.945 +++ +++

b34 13 9.617 6.617 +++ +++

b36 5 16.668 6.378 +++ +++

c11 48 48.127 5.918 +++ +++

c28 25 78.069 5.708 +++ +++

143
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.4: Plot of correlation coefficient between actual and estimated activity value test set

ii) Cat Scramble Validation: 99 spreadsheets were generated in order to have a 99%
confidence level (Figure 4.5). The data of cross validation clearly indicates that all
values generated after randomization produced hypotheses with no significant value.
Out of 99 runs, all trials had a correlation value less than 0.89, and also RMS
deviation and total cost were very high, which is not desirable for a good hypothesis.
Thus Cat Scramble validation also provided strong confidence in the generated model.

Figure 4.5: Graph of 99% Fischer’s randomization test of hypothesis1

iii) External Test Set Validation: An external test set of 9 compounds comprising
thiophene derivatives and alkylidine hydrazide derivatives with activity range of 89
nm to 60000 nm showed a squared correlation coefficient (r2) of 0.65 (Figure 4.6).

144
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.6: Plot of correlation coefficient between actual and estimated activity value of
external test set

As an additional external validation step some clinical trial candidates like BAY-27-
9955 and LY2409021 were mapped onto the developed model. BAY-27-9955 showed
only two feature mapping with a fit value of 4.02. The hydroxyl moiety was mapped
by HBD and methyl group by HY (Figure 4.7 (a)). This may be the reason why BAY-
27-9955 was failed in clinical trials. LY2409021 showed two features mapping with
a fit value of 5.238. The amino group was mapped by HBD and carboxyl moiety was
mapped by NI (Figure 4.7 (b)).

Figure 4.7: Pharmacophore mapping of clinical trial drugs: (a) BAY-27-9955 showed
mapping with HBD and HY (b) LY2409021 showed mapping with HBD and NI
145
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Conclusion
The present study has been focused on the development of pharmacophore model to
reveal structural and physical/chemical requirement for glucagon receptor
antagonistic activity. Best pharmacophore model generated using 56 molecules of
training set comprised of one hydrogen bond donor (HBD), one negative ionizable
(NI), one hydrophobic (HY) and one ring aromatic (RA) feature. The robustness of
the model was characterized by a high correlation coefficient of the training set r
value of 0.89 and test set r2 0.73. The developed model was validated by three
methods i.e. internal test set validation, Fischer’s validation and external test set
validation. All validation process suggested that the developed model is a good and
reliable pharmacophore model with high statistical significance.

Series 2: Pyrrolidine Carboxamide as 11β HSD1 Inhibitors

Pharmacophore Generation
A series of 30 structurally diverse compounds having having 11β HSD1 inhibitory
activity from 0.03 nM to 290 nM was used for development of pharmacophore model
(Table-4.6). Out of 30 compounds a training set of 20 compounds was used to build
the pharmacophore model, while 10 compounds were kept as internal test set to
validate the predictivity of the generated model.

Table 4.6: Structure and biological activity of Pyrrolidine carboxamide derivatives with 11β
HSD1 inhibitory activity

Pyrrolidine carboxamide derivatives

H
N R1 N
N
N N
R2 O O
O O

4, 7, 11-27, 35-43 5 6

Comp. R1 R2 IC50(nm)

4 H 4.15
N

146
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

7 H 2.2

11 N
H 22

12 N H 1.2

13 N H 3.5

14 N H 0.03

15 N H 0.03

16 N H 0.03

17 N H 0.32

18 N H 0.32
CF3

19 N H 0.32
CF3

Cl
20 N H 0.32

147
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

CN
21 N H 0.03

22 N H 0.32
CN

23 N H 20
OH

24 N H 5
N

25 N H 10
O

OH
26 N H 0.34

OH
N
27 H 0.03

OH

35 H 34
N

OH

36 H 0.8
N

OH

37 H 0.173
N

148
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

OH

38 H 1
N

OH

39 H 0.03
N F

OH

40 H 30
N O

H
H
41 H 19.4
N

42 N
OH 23

43 N OH 31.3

Pharmacophore Model Development and Statistical Evaluation

Among various generated models (Table 4.7) hypothesis 1 comprising of one HBD,
one PI, two HYs and 4 excluded volumes (EV) with cost difference of 47.89,
correlation coefficient of 0.92, root mean square (RMS) of 1.02, squared correlation
coefficient value of 0.85 and configuration of 10.95 was considered as best.
The HypoGen algorithm was used to develop pharmacophore models with chemical
features required for 11β HSD1 inhibitory activity. Various pharmacophore models
were developed and assessed on the basis of their cost functions represented in bits
unit. Out of many runs a set of ten hypotheses with fixed cost of 79.15 bits (well
separated from the null hypothesis cost of 138.68 bits) was considered. Among
chosen set of ten hypothesis the top-ranked pharmacophore model (Hypothesis1) had
the best predictive power and statistical significance described by the high correlation
149
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

coefficient (r = 0.92, r2 = 0.85), root mean- square deviation (1.02), weight (2.01) and
error costs (77.82) and a cost difference of 47.89. The configuration cost was 10.95,
which indicates that no over fitting of the data has been exercised. The cost difference
between total and fixed costs for the best hypothesis was only 11.6 bits, indicating the
high probability of the true correlation of the data. Thus, hypothesis1 with four
features, viz., one H-bond donor (HBD), one positive ionizable (PI) and two
hydrophobes (HY1 and HY2) was considered as the best pharmacophore model for
11β HSD1 inhibitory activity. The cost values, correlation coefficients (r), RMSD,
and features for the top ten hypotheses are listed in Table 4.7.
Table 4.7: Results of the top 10 pharmacophore hypotheses generated by the hypoGen
algorithm

Hypothesis Total Cost Cost Difference RMSD Correlation (r) Features

1 90.79 47.89 1.02 0.92 HBD, 2 HY,PI, 4 EV

2 99.75 38.93 1.41 0.84 HBD, 2 HY,PI, 4EV

3 103.16 35.52 1.53 0.82 HBD, 2 HY,PI, 6EV

4 103.53 35.15 1.55 0.81 HBD, 2 HY,PI, 3EV

5 103.77 34.91 1.56 0.80 HBD, 2 HY,PI, 3EV

6 104.12 34.56 1.56 0.81 HBD, 2 HY,PI, 5EV

7 106.56 32.12 1.64 0.78 HBD, 2 HY,PI, 3EV

8 107.40 31.28 1.66 0.78 HBD, 2 HY,PI, 1EV

9 108.19 30.49 1.69 0.77 HBD, 2 HY,PI, 1 EV

10 108.22 30.46 1.70 0.77 HBD, 2 HY,PI, 1 EV

Hypothesis1, identified as the best hypothesis, was used to estimate the activity of the
training set molecules. All the training set compounds were classified by their activity
as highly active (<1 nM, +++), moderately active (1-20 nM, ++) and inactive (>20
nM, +). Among 20 training set compounds, only three inactive compounds (35, 40,
43) were predicted as moderate by Hypothesis1. Consequently, for 49 of 56 training
set compounds, the predicted IC50 (nM) values were within the same activity scale as
the experimental values in the training set. Table 4.8 and Figure 4.8 represent the

150
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

actual and estimated 11β HSD1 inhibitory activity of the 20 training set molecules
based on the best hypothesis.

Table 4.8: Actual and estimated values of the training set based on the pharmacophore
hypothesis

Actual Estimated Fit Activity Estimated

Compound
IC50 IC50 value scale Activity scale

4 4.15 2.65 7.117 ++ ++

5 261 295.913 5.069 + +

6 290 307.001 5.053 + +

7 2.2 6.514 6.726 ++ ++

13 3.5 11.513 6.479 ++ ++

15 0.03 0.238 8.164 +++ +++

17 0.32 0.171 8.306 +++ +++

20 0.32 0.14 8.395 +++ +++

21 0.03 0.114 8.482 +++ +++

23 20 5.658 6.787 ++ ++

24 5 9.396 6.567 ++ ++

25 10 7.269 6.679 ++ ++

26 0.34 0.672 7.713 +++ +++

27 0.03 0.093 8.57 +++ +++

35 34 12.891 6.43 + ++

37 0.173 0.041 8.925 +++ +++

39 0.03 0.163 8.328 +++ +++

40 30 3.094 7.049 + ++

41 19.4 6.26 6.743 ++ ++

43 31.3 12.02 6.46 + ++

151
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.8: Plot of correlation coefficient between actual and estimated activity value of
training set

Pharmacophore Mapping
The obtained pharmacophoric features and their inter feature distances are shown in
Figure 4.9 (a) and 4.9 (b). Most active compounds 14, 15, 21, 27 and 39 showed full
four feature mapping to the pharmacophore (Figure 4.10 (a)-(e)). Adamantane ring of
all the active compounds showed mapping with HY1 whereas amino groups was
mapped onto HBD and pyrrollidine moieties were mapped by PI features. The
pyrrollidine ring substitutions of all the active compounds were mapped over HY2
features. Inactive compound 6 showed only two feature mapping (Figure 4.11).
cyclohexane ring on HY1whereas cycloheptane ring on HY2.

Figure 4.9: Pharmacophore features (Hypothesis 1): (a) Four pharmacophore features: one
HBD (pink color), one PI (red color), two HY1 and HY2 (light blue colors) and four excluded
volumes (grey color) (b) Interfeature distances between the pharmacophore features

152
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.10: Pharmacophore mapping with active compounds: (a) Compound no. 14, (b)
Compound no. 15, (c) Compound no. 22, (d) Compound no. 27 and (e) Compound no. 39

Figure 4.11: Pharmacophore mapping with inactive compound no. 6

Validation of Pharmacophore Model

In order to check the accuracy of the model, various validation approaches were
adopted.
i) Internal Test Set Validation: The pharmacophore hypothesis was internally
validated by using the test set comprising of 10 molecules with 11β HSD1 inhibitory
activity. The test set compounds were treated alike the training set compounds. A
squared correlation coefficient of 0.79 as shown in Figure 4.12 indicated a good
correlation between the actual and estimated activities of the test set compounds. The
153
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

scored estimated activities and actual value of test set compounds are shown in Table
4.9. The agreement between actual and predicted activity of test set compound
testifies the soundness of hypothesis 1.

Table 4.9: Actual and estimated values of the test set based on the pharmacophore hypothesis

Actual Estimated Fit Activity Estimated

Compound
IC50 IC50 value scale Activity scale

11 22 5.001 6.841 + ++

12 1.2 1.025 7.529 ++ ++

14 0.03 0.071 8.686 +++ +++

16 0.03 0.076 8.657 +++ +++

18 0.32 0.384 7.956 +++ +++

19 0.32 0.128 8.433 +++ +++

22 0.32 0.621 7.747 +++ +++

36 0.8 0.669 7.715 +++ +++

38 1 0.119 8.464 +++ +++

42 23 37.007 5.972 + +

Figure 4.12: Plot of correlation coefficient between actual and estimated activity value of test
set
154
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

ii) Cat Scramble Validation: 99 spreadsheets were generated in order to have a 99%
confidence level (Figure 4.13). The data of cross validation clearly indicates that all
values generated after randomization produced hypotheses with no significant value.
Out of 99 runs, all trials had a correlation value less than 0.92, and also RMS
deviation and total cost were very high, which is not desirable for a good hypothesis.
Thus Cat Scramble validation also provided strong confidence in the generated model.

Figure 4.13: Graph of 99% Fischer’s randomization test of hypothesis1

iii) External Test Set Validation: An external test set of 9 compounds comprising
disubstituted cyclohexylbenzamide derivatives with activity range of 0.5 nm to 14 nm
showed a squared correlation coefficient (r2) of 0.62 (Figure 4.14).

Figure 4.14: Plot of correlation coefficient between actual and estimated activity value of
external test set

155
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

As an additional external validation step partial 11β HSD1 inhibitor carbenoxolone

was mapped onto the developed model. Carbenoxolone showed two features mapping
with a fit value of 5.161. The cyclohexane rings were mapped by two HYs (Figure
4.15). It confirms partial binding of carbenoxolone to the 11β HSD1.

Figure 4.15: Pharmacophore mapping of carbenoxolone

Conclusion
The present study has been focused on the development of pharmacophore model to
reveal structural and physical/chemical requirement for 11β HSD1 inhibitory activity.
Best pharmacophore model generated using 30 molecules of training set comprised of
one hydrogen bond donor (HBD), one positive ionizable (PI) and two hydrophobes
(HY1 and HY2) features. The robustness of the model was characterized by a high
correlation coefficient of the training set r2 value of 0.85 and test set r2 0.79. The
developed model was validated by three methods i.e. internal test set validation,
Fischer’s validation and external test set validation. All validation process suggested
that the developed model is a good and reliable pharmacophore model with high
statistical significance.

Series 3: Bicyclo[2.2.2]octyltriazole Derivatives as 11β HSD1 Inhibitors

Pharmacophore Generation
A series of 32 structurally diverse compounds having having 11β HSD1 inhibitory
activity from 1 nM to 2800 nM was used for development of pharmacophore model
(Table-4.10). Out of 32 compounds a training set of 24 compounds was used to build
the pharmacophore model, while 8 compounds were kept as internal test set to
validate the predictivity of the generated model.
156
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Table 4.10: Structure and biological activity of Bicyclo[2.2.2]octyltriazole derivatives with

11β HSD1 inhibitory activity

Bicyclo[2.2.2]octyltriazole derivatives

R1
R2

1-40

Comp. R1 R2 IC50(nm)

1 n-butyl N 4
N N

2 Isobutanol N 35
N N

CF3

3 n-butyl N 1.8
N N

CF3

5 O N 1.2
N N

CF3

13 MeO N 2
N N

CF3

14 OCH2CH3 N 4
N N

CF3
Cl
15 N 1.1
O
N N

CF3
SO2CH3
16 N 1
O
N N

157
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

CF3

17 NHSO2CH3 N 30
N N

CF3

18 NHSO2CH2CH3 N 20
N N

CF3

19 NHSO2(CH2)2CH3 N 8.1
N N

CF3

20 CH2SCH2CH3 N 3.3
N N

CF3

21 CH2SOCH2CH3 N 4.1
N N

CF3

22 CH2SO2CH2CH3 N 7
N N

CF3

23 CH2SO2CH3 N 11
N N

CF3

24 CH2SO2CH2CH2CH3 N 5.1
N N

CF3

25 CH2SO2CH(CH3)2 N 4.1
N N

CF3

26 CH2SO2C(CH3)3 N 5.4
N N

CF3
O
27 S N 4.4
O
N N

158
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

CF3
F O
28 S N 5.5
O
N N

CF3

29 CH2SO2CF3 N 6.1
N N

CF3

30 CH2SO2CH2CF3 N 3.7
N N

CF3

31 SO2CH2CH3 N 25
N N

CF3

32 CH2CH2SO2CH2CH3 N 3.2
N N

CF3

33 CH2CH2CH2SO2CH2CH3 N 1.6
N N

Cl

34 CH2SO2CH2CH3 N 7.5
N N

Cl

35 CH2SO2CH2CH3 N 900
N N

Cl
N
36 CH2SO2CH2CH3 2800
N N

Cl
OH
37 CH2SO2CH2CH3 N 2.8
N N

Br

38 CH2SO2CH2CH3 N 10
N N

159
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

39 CH2SO2CH2CH3 N 20
N N

40 CH2SO2CH2CH3 N 140
N N

Pharmacophore Model Development and Statistical Evaluation

Among various generated models (Table 4.11) hypothesis 1 comprising of one
Hydrogen bond acceptor lipid (HBA_Li), three Hydrophobes (HY1, HY2, HY3) and
one ring aromatic (RA) with cost difference of 14.48, correlation coefficient of 0.89,
root mean square (RMS) of 0.80, squared correlation coefficient value of 0.79 and
configuration of 13.42 was considered as best.
The HypoGen algorithm was used to develop pharmacophore models with chemical
features required for 11β HSD1 inhibitory activity. Various pharmacophore models
were developed and assessed on the basis of their cost functions represented in bits
unit. Out of many runs a set of ten hypotheses with fixed cost of 95.06 bits (well
separated from the null hypothesis cost of 117.50 bits) was considered. Among
chosen set of ten hypothesis the top-ranked pharmacophore model (Hypothesis1) had
the best predictive power and statistical significance described by the high correlation
coefficient (r = 0.89, r2 = 0.79), root mean- square deviation (0.80), weight (1.19) and
error costs (88.40) and a cost difference of 14.48. The configuration cost was 13.42,
which indicates that no over fitting of the data has been exercised. The cost difference
between total and fixed costs for the best hypothesis was only 7.95 bits, indicating the
high probability of the true correlation of the data. Thus, hypothesis1 with five
features, viz., one hydrogen bond acceptor lipid (HBA_Li), three hydrophobes (HY1,
HY2, HY3) and one ring aromatic (RA) was considered as the best pharmacophore
model for 11β HSD1 inhibitory activity. The cost values, correlation coefficients (r),
RMSD, and features for the top ten hypotheses are listed in Table 4.11.

160
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Table 4.11: Results of the top 10 pharmacophore hypotheses generated by the hypoGen
algorithm

Hypothesis Total Cost Cost Difference RMSD Correlation (r) Features

HBA_Li, HY1, HY2,

1 103.02 14.48 0.80 0.89
HY3, RA

HBA_Li, HY1, HY2,

2 103.06 14.44 0.80 0.89
HY3, RA

HBA_Li, HY1, HY2,

3 103.16 14.34 0.81 0.88
HY3, RA

HBA_Li, HY1, HY2,

4 103.66 13.84 0.84 0.87
HY3, RA

HBA_Li, HY1, HY2,

5 103.68 13.82 0.83 0.87
HY3, RA

HBA_Li, HY1, HY2,

6 104.95 12.55 0.90 0.85
HY3, RA

HBA_Li, HY1, HY2,

7 105.57 11.93 0.93 0.84
HY3, RA

HBA_Li, HY1, HY2,

8 105.64 11.86 0.93 0.84
HY3, RA

HBA_Li, HY1, HY2,

9 105.76 11.74 0.94 0.84
HY3, RA

HBA_Li, HY1, HY2,

10 105.81 11.69 0.94 0.84
HY3, RA

Hypothesis1, identified as the best hypothesis, was used to estimate the activity of the
training set molecules. All the training set compounds were classified by their activity
as highly active (<10 nM, +++), moderately active (10-30 nM, ++) and inactive (>30
nM, +). Among 24 training set compounds, one active compound (1) was predicted as
moderate; and five moderately active compounds (17, 18, 23, 31, 39) were predicted
as active by Hypothesis1. Consequently, for 19 of 24 training set compounds, the
predicted IC50 (nM) values were within the same activity scale as the experimental
values in the training set. Table 4.12 and Figure 4.16 represent the actual and
estimated 11β HSD1 inhibitory activity of the 24 training set molecules based on the
best hypothesis.

161
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Table 4.12: Actual and estimated values of the training set based on the pharmacophore
hypothesis

Actual Estimated Fit Activity Estimated

Compound
IC50 IC50 value scale Activity scale

1 4 14.038 8.053 +++ ++

3 1.8 5.963 8.425 +++ +++

5 1.2 6.167 8.41 +++ +++

13 2 6.744 8.371 +++ +++

14 4 6.872 8.363 +++ +++

15 1.1 0.667 9.376 +++ +++

16 1 1.528 9.016 +++ +++

17 30 6.576 8.382 ++ +++

18 20 5.937 8.426 ++ +++

19 8.1 5.713 8.443 +++ +++

20 3.3 7.219 8.342 +++ +++

21 4.1 5.69 8.445 +++ +++

22 7 7.062 8.351 +++ +++

23 11 5.954 8.425 ++ +++

24 5.1 6.327 8.399 +++ +++

26 5.4 5.663 8.447 +++ +++

27 4.4 2.466 8.808 +++ +++

31 25 5.958 8.425 ++ +++

32 3.2 6.123 8.413 +++ +++

35 900 376.42 6.624 + +

36 2800 989.39 6.205 + +

38 10 7.927 8.301 +++ +++

39 20 8.145 8.289 ++ +++

40 140 216.75 6.864 + +

162
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.16: Plot of correlation coefficient between actual and estimated activity value of
training set

Pharmacophore Mapping
The obtained pharmacophoric features and their inter feature distances are shown in
Figure 4.17 (a) and 4.17 (b). Most active compounds 16 showed full five features
mapping to the pharmacophore (Figure 4.17 (c)). Triazole ring showed mapping with
HBA_Li, carbon trifluoride moiety was mapped over HY1, benzene ring attached to
carbon trifluoride on HY2 and bicycle ring onto the HY3; whereas benzene ring
attached to sulfonyl group showed mapping over RA feature. Inactive compound 36
showed only three feature mapping (Figure 4.17 (d)). Methyl ring attached to the
triazole ring on HY1, the bicycle ring onto HY2 and HBA_Li showed mapping with
oxygen of sulfonyl group.

Validation of Pharmacophore Model

In order to check the accuracy of the model, various validation approaches were
adopted.
i) Internal Test Set Validation: The pharmacophore hypothesis was internally
validated by using the test set comprising of 8 molecules with 11β HSD1 inhibitory
activity. The test set compounds were treated alike the training set compounds. A
squared correlation coefficient of 0.72 as shown in Figure 4.18 indicated a good
correlation between the actual and estimated activities of the test set compounds. The
scored estimated activities and actual value of test set compounds are shown in Table

163
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.17: Pharmacophore features (Hypothesis 1) and pharmacophore mappings of 11β

HSD1 inhibitor: (a) Five pharmacophore features: one HBA_Li (green color), three HY1,
HY2 and HY3 (light blue color) and one RA (orange color) (b) Interfeature distances between
the pharmacophore features (c) Pharmacophore mapping with most active compound, 16 of
training set (d) Pharmacophore mapping with least active compound 36 of training set.

4.13. The agreement between actual and predicted activity of test set compound
testifies the soundness of hypothesis 1.

Table 4.13: Actual and estimated values of the test set based on the pharmacophore
hypothesis

Actual Estimated Fit Activity Estimated

Compound
IC50 IC50 value scale Activity scale

2 35 861.551 6.265 + +

25 4.1 5.772 8.439 +++ +++

28 5.5 6.595 8.381 +++ +++

164
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Actual Estimated Fit Activity Estimated

Compound
IC50 IC50 value scale Activity scale

29 6.1 6.05 8.418 +++ +++

30 3.7 5.309 8.475 +++ +++

33 1.6 5.893 8.43 +++ +++

34 7.5 6.71 8.373 +++ +++

37 2.8 6.971 8.357 +++ +++

Figure 4.18: Plot of correlation coefficient between actual and estimated activity value of test
set

ii) Cat Scramble Validation: 99 spreadsheets were generated in order to have a 99%
confidence level (Figure 4.19). The data of cross validation clearly indicates that all
values generated after randomization produced hypotheses with no significant value.
Out of 99 runs, all trials had a correlation value less than 0.89, and also RMS
deviation and total cost were very high, which is not desirable for a good hypothesis.
Thus Cat Scramble validation also provided strong confidence in the generated model.

165
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.19: Graph of 99% Fischer’s randomization test of hypothesis1

iii) External Test Set Validation: An external test set of 33 compounds comprising
4-Methyl-5-phenyl triazoles derivatives with activity range of 1 nm to 650 nm showed
a squared correlation coefficient (r2) of 0.67 (Figure 4.20).

Figure 4.20: Plot of correlation coefficient between actual and estimated activity value of
external test set

As an additional external validation step partial 11β HSD1 inhibitor carbenoxolone

was mapped onto the developed model. Carbenoxolone showed four features mapping
with a fit value of 7.663. The cyclohexane rings were mapped by HY1, HY2 and
HY3; oxo group attached to the cyclohexane ring was mapped by HBA_Li (Figure
4.21). It confirms partial binding of carbenoxolone to the 11β HSD1.
166
 CHAPTER-4

LIGAND BASED PHARMACOPHORE (LBP) MODELING

Figure 4.21: Pharmacophore mapping of carbenoxolone

Conclusion
The present study has been focused on the development of pharmacophore model to
reveal structural and physical/chemical requirement for 11β HSD1 inhibitory activity.
Best pharmacophore model generated using 24 molecules of training set comprised of
one hydrogen bond acceptor_lipid (HBA_Li), three hydrophobes (HY1, HY2 and
HY3) and one ring aromatic (RA). The robustness of the model was characterized by
a high correlation coefficient of the training set r2 value of 0.85 and test set r2 0.79.
The developed model was validated by three methods i.e. internal test set validation,
Fischer’s validation and external test set validation. All validation process suggested
that the developed model is a good and reliable pharmacophore model with high
statistical significance.

167
 CHAPTER-5
Virtual Screening
&
Structure Based
Pharmacophore (SBP)
Modeling
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Virtual Screening Based Identification of New Chemical Entities

Virtual Screening
Virtual screening represents a fast & cost effective tool for identification of novel lead
by screening chemical compound libraries. Virtual screening method has emerged as
powerful technique in drug discovery and now a day most of the research based
pharmaceutical companies are relying on computational tools for lead discovery and
optimization. Lipinski’s rule, estimated activity and fit values are used as
computational filters to obtain the most significant compounds from the large
compound databases.

Material and Method

Pharmacophore Based Database Searching

It is a well known fact that a statistically healthy pharmacophore comprising of all the
chemical functionalities essential for the desired activity can be used for mining of
chemical compound databases to retrieve novel chemical entities.

In view of this we have attempted to use in-house developed pharmacophore as 3D

query to identify novel GRA and 11β HSD1 inhibitors. In Discovery Studio the
virtual screening is based on two type of search option, fast and best flexible search.
A best flexible search databases method was employed for database screening in order
to identify putative compounds which spatially map with the corresponding features
in the pharmacophoric query. In present study NCI database was screened which is a
collection of 3D structures for over 400,000 compounds, The NCI database is
maintained by the Developmental Therapeutics Program Division of Cancer
Treatment, National Cancer Institute, Rockville, MD. After initial round of compound
identification, different computational filters like, estimated activity and fit values and
Lipinski’s rule-of five were applied to identify most potent compounds with drug like
properties.

Estimated Activity and Fit Values

A fit value shows the mapping of the chemical features of the molecule over the
pharmacophore features. Higher the fit value better is the mapping of the molecule
onto the pharmacophore. Therefore fit values were considered as filter for screening
168
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

of potential hits. In addition to this the database compounds were also screened on the
basis of the estimated values. Only the hits which had the estimated activity very
close to the most active compounds of the series taken to develop the pharmacophore
model were retained.

Drug Likeness Screening

Most of the potent compounds are not developed into drugs due to poor
pharmacokinetic profile. Hence all the identified hits were checked for the
pharmacokinetic profile by applying Lipinski’s rule-of five, which calculates the
absorption and intestinal permeability of the compound. A compound with Log P less
than 5, molecular weight less than 500, number of H-bond donors less than 5, number
H-bond acceptors less than 10 and number of rotatable bonds less than 10, will show
good absorption. Compounds violating more than one of these five rules were
screened out.

Result and Discussion

The validated pharmacophore models were used for pharmacophore-based virtual
screening (VS) to identify promising and selective GRA and 11β HSD1 inhibitors. A
sequential VS procedure was applied, wherein the pharmacophore-based VS were
followed by estimated activity and fit value pre-filtration, and further drug likeness
screening to reduce the number of hits in each different screening step. By this
process we have retrieved a good number of hits as shown in Table 5.1.

Table 5.1: Number of compounds identified through pharmacophore based sequential virtual
screening

Database Number of
Series
library compounds retrieved

Series 1( β alanine, isoserine and thiazole as NCI 3

GRA )- Pharmacophore

Series 2 (Pyrrolidine carboxamide NCI 1

Inhibitors of 11β HSD1) – Pharmacophore

Series 3 (Bicyclo[2.2.2] octyltriazole NCI 2

Inhibitors of 11β HSD1) – Pharmacophore

169
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Series 1 Pharmacophore Based Virtual Screening

Needless to mention that utility of validated pharmacophore model lies in virtual
screening and since the developed pharmacophore model showed all the signs of its
soundness and universality, it was used to screen NCI data base comprising of
(260,071) structurally diverse compound using best flexible database search option.
NCI chemical compound database screening led to the retrieval of 295 hits. All the
hits were checked for their fit value, estimated activity and Lipinski’s violation. This
process resulted in selection of three promising compounds namely SKDGRA1,
SKDGRA2 and SKDGRA3 shown in (Table 5.2) exhibiting perfect four feature
mapping with good fit value ranging from 7.09- 6.60 respectively and zero Lipinski’s
violation.
Table 5.2: NCI hits obtained from β alanine, isoserine and thiazole derivatives based
pharmacophore

Estimated
Name of hits Fit value Mapping of hits compounds
value(µM)

3.227
SKDGRA1 7.091

5.539
SKDGRA2 6.857

9.799
SKDGRA3 6.609

170
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Series 2 Pharmacophore Based Virtual Screening

The validated pharmacophore model based on pyrrolidine carboxamide derivatives
was used to screen NCI data base of structurally diverse compound using best flexible
database search option. The retrieved hits were filtered according to their fit and
estimated value which led to retention of 17 out of 300 hits. Lipinski’s rule of five
was applied to the selected hits to check their druggable properties which further
reduced the list of hits to 1 with fit value of 7.893 as shown in (Table 5.3).

Table 5.3: NCI hits obtained from pyrrolidine carboxamide based pharmacophore

Estimated
Name of hits Fit value Mapping of hits compounds
value(µM)

0.443
SKDHSD1 7.893

Series 3 Pharmacophore Based Virtual Screening

The validated pharmacophore model based on bicyclo[2.2.2] octyltriazole derivatives
was used to screen NCI data base of structurally diverse compound using best flexible
database search option. The retrieved hits were filtered according to their fit and
estimated value which led to retention of 298 out of 300 hits. Lipinski’s rule of five
was applied to the selected hits to check their druggable properties which further
reduced the list of hits to 2 with fit values ranging from 8.449-8.228 as shown in
(Table 5.4).

Table 5.4: NCI hits obtained from bicyclo[2.2.2] octyltriazole based pharmacophore

Estimated
Name of hits Fit value Mapping of hits compounds
value(µM)

5.626
NSC 13385 8.449

171
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Estimated
Name of hits Fit value Mapping of hits compounds
value(µM)

9.366
NSC 16653 8.228

Structure Based Pharmacophore Generation

Structure based pharmacophores generation has emerged as a powerful technique in

the area of drug design. The interest structure based drug design has aroused due to
the rapid development of high resolution protein structures. The structure based
pharmacophore helps in understanding the protein– ligand interaction within the
protein structural framework, and helps in designing of ligands for target protein.

Material and method

Protein Structure Preparation

Recently discovered three-dimensional structure of GRA (4L6R) and 11β HSD1
(2BEL) was obtained from PDB (Figure 5.1a & 5.1b) and used for receptor structure
based pharmacophore generation. The protein structures were checked for valency
and missing hydrogens were added.

Binding Site Identification and Interaction Generation

The prepared protein structures were subjected to binding site identification. All the
structure based pharmacophore modeling studies were performed on Accelery’s
Discovery Studio which offers grid-based methods to develop pharmacophore
models. The receptor binding site was defined using a sphere of radius 9.0 Ǻ, to
include the important protein residues involved in binding interaction with ligands.
The density of lipophilic sites and polar sites parameter value were kept to 10.
Accelrys software package provides Ludi interaction maps for hydrogen bond donor,
hydrogen bond acceptor and hydrophobic interactions; hence these feature types were
used to generate structure based pharmacophore models.

172
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Figure 5.1: Three dimensional structure of glucagon receptor and 11β HSD1 enzyme

Clustering and Pharmacophore Generation

The interaction map exhibits a large number of chemical features. Thus selection of
essential pharmacophore features from the interaction map is quite a tedious task. To
overcome this problem, neighboring features of the same type were grouped to the
same clusters. The features closest to the geometric centers of cluster were selected to
represent the clusters, by omitting the root of features. To further reduce the number
of features multiple 3D queries with lesser numbers of features were generated from
the interaction map by considering all possible combinations. Finally, two clusters
each for hydrogen bond acceptor, hydrogen bond donor and a hydrophobe were used
to construct final pharmacophore model. The final models were subjected to non-
feature atom exclusion. The exclusion constraint features is an object that represents
an unoccupied volume in space, within a circumference.325 A pharmacophore with an
excluded volume only matches if number of atoms penetrates the excluded area.

Results and Discussions

The obtained hypotheses contained six features namely: two hydrogen bond donors,
two hydrogen bond acceptors and two hydrophobe describing the interaction between
the protein and a ligand. The inter-atomic distances of the GR and 11β HSD1

pharmacophores are depicted in Figure 5.2 and Figure 5.3. The developed six
featured structures based pharmacophore models were validated using some well

173
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

known clinical trial/clinical drug candidates with GRA and 11β HSD1 inhibitory
activity.

Figure 5.2: The inter-feature distances of the glucagon receptor pharmacophore

Figure 5.3: The inter-feature distances of the 11β HSD1 enzyme pharmacophore

Mapping of Clinical Trials Drugs (GRA)

In case of GR the developed six featured pharmacophore model was validated using
two clinical trial drugs i.e., BAY-27-9955 and LY2409021. Both the chemical entities

174
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

were mapped onto the pharmacophore. BAY-27-9955 showed mapping with 2HY and
1HBD features whereas LY2409021, exhibited mapping with 2HBA, 1HY and 1
HBD. The fit values were found to be 0.8 and 1.867 respectively for BAY-27-9955
and LY2409021 (Table 5.5).

Table 5.5: Pharmacophore mapping, fit value and lipinski’s violation of clinical trial drugs
(BAY-27-9955 and LY2409021)

Lipinski
Fit
Name Violation Features Mapped
Value

BAY-27-9955 0 0.8

LY2409021 0 1.867

Mapping of NCI Hits (GRA)

In order to check the validity of the NCI hits (SKDGRA1, SKDGRA2 and
SKDGRA3) with GR antagonist activity retrieved through ligand based virtual
screening, they were mapped over the six feature structure based pharmacophore
model. SKDGRA1 exhibited five features mapping (2HBA, 2HBD and 1HY);
SKDGRA2 showed four feature mapping (2HBA, 1HBD and 1HY) and SKDGRA3
also exhibited four feature mapping (2HBA, 1HBD and 1HY). The fit values of the
hits were found to be 1.647, 0.999 and 1.561 respectively for SKDGRA1, SKDGRA2
and SKDGRA3 (Table 5.6).

175
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Table 5.6: Pharmacophore mapping, fit value and lipinski’s violation of NCI retrieved hits
(SKDGRA1, SKDGRA2 and SKDGRA3)
Lipinski
Name Violation Fit Value Features Mapped

SKDGRA1 0 1.647

SKDGRA2 0 0.999

SKDGRA3 0 1.561

The mapping of the retrieved hits onto the SBP provided a fair insight into the binding
of the hits with the receptor. SKDGRA1 showed good mapping with the SBP, missing
only one feature. It is interesting to note that, SKDGRA1 mapped onto HY and HBD
176
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

at the same position of LBP mapping (Figure 5.2). The other hits SKDGRA2 and
SKDGRA3 also showed good mapping with the SBP, missing only two features.
SKDGRA2 mapped onto HBD at the same position of LBP mapping (Figure 5.2) and
SKDGRA3 mapped onto HY at the same position of LBP mapping (Figure 5.2).
Good mapping of SKDGRA1, SKDGRA2 and SKDGRA3 onto the SBP and
commonality in mapping with the LBP confirmed the binding potential of the hits
onto the receptor.

Mapping of Reference Compound (11β HSD1 Inhibitor)

The 11β HSD1 based six featured pharmacophore model was validated using one
partial inhibitor of 11β HSD1 i.e., carbenoxolone. Carbenoxolone was mapped onto
the pharmacophore and out of 6 features 2HBA and 1HY were mapped. The fit value
of the carbenoxolne was found to be 1.987 (Table 5.7).

Table 5.7: Pharmacophore mapping, fit value and lipinski’s violation of partial 11β HSD1
inhibitor (carbenoxolone)

Lipinski
Fit
Name Violation Features Mapped
Value

Carbenoxolone 0 1.987

Mapping of NCI Hits (11β HSD1 Inhibitor)

Three 11β HSD1 inhibitors (NCI hits- SKDHSD1, SKDHSD2 and SKDHSD3)
retrieved through ligand based virtual screening were mapped over the six feature
structure based pharmacophore model. SKDHSD1 exhibited 3 feature mapping
(1HBA, 1HBD and 1HY); SKDHSD2 also showed 3 feature mapping (1HBA, and
2HY) and SKDHSD3 exhibited four feature mapping (2HBA, and 2HY). The fit
values were found to be 1.634, 2.079 and 1.45 respectively for SKDHSD1,
SKDHSD2 and SKDHSD3 (Table 5.8).

177
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Table 5.8: Pharmacophore mapping, fit value and lipinski’s violation of NCI retrieved hits
(SKDHSD1, SKDHSD2 and SKDHSD3)
Lipinski
Fit
Name Violation Features Mapped
Value

SKDHSD1 0 1.634

SKDHSD2 0 2.079

SKDHSD3 0 1.45

SKDHSD1 and SKDHSD2 showed fair mapping with the SBP (three features).
SKDHSD1 showed mapping onto HY at the same position of LBP mapping (Figure
5.3) while SKDHSD2 showed mapping onto HY and HBA at the same position of
LBP mapping (Figure 5.4). SKDHSD3 showed good mapping with the SBP (four
features). It showed mapping onto HY and HBA at the same position of LBP mapping
(Figure 5.4). Good mapping of SKDHSD1, SKDHSD2 and SKDHSD3 onto the SBP
and commonality in mapping with the LBP established the binding potential of the
hits onto the receptor.

178
 CHAPTER-5

VIRTUAL SCREENING & STRUCTURE BASED PHARMACOPHORE (SBP) MODELING

Conclusion
3D SBP models for GRA and 11β HSD1 inhibitor were developed and used for
validation of NCI hits retrieved through ligand based virtual screening. All the
retrieved hits mapped well to the SBP and also showed commonality in mapping with
LBP at the same position. This provides a confirmation of the possible binding of all
the retrieved hits obtained from LBP based virtual screening.

179
 CHAPTER- 6
Molecular Docking
&
Experimental
Validation
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

Molecular Docking
Molecular docking is a powerful in silico technique often used to visualize the mode
of binding of the ligand to its target protein. The vital information provided by the
molecular docking on drug-receptor interaction can be used in the designing of the
novel ligands with improved efficacy and reduced side effects. Now a days molecular
docking is exercised either before or after binding experiments and in vitro or in vivo
activity assays, to get insight into the precise binding geometry of the compounds
under investigation.326 Molecular docking consist of two essential components;
generation of promising ligand binding orientations known as sampling and prediction
of the binding strength for a specific ligand orientations known as scoring. The best
orientation with the lowest energy score is considered as the preferred binding mode.
Because of precision in predicting the correct binding interactions and orientations (in
some cases at a very high accuracy with reference to existing crystal structure of the
complex studied), molecular docking is being widely used in modern days drug
design projects.

Material and Methods

With an aim to understand the binding interaction between the target proteins
(glucagon receptor and 11β HSD1 enzyme) and the hits obtained by virtual screening,
molecular docking studies were performed using “Libdocker” protocol.

Target Identification
Crystal structures of the recently discovered human glucagon receptor (4L6R-Figure
6.1) and 11β HSD1 in complex with NADP and carbenoxolone (2BEL- Figure 6.2)
was obtained from protein data bank and used for molecular docking studies.
Receptor and Ligand Preparation
The protein structures (glucagon receptor and 11β HSD1 enzyme) obtained from PDB
were checked for valency and the missing hydrogen were added. All the water
molecules from the protein hierarchy were removed and the structures were split into
the protein part and crystal ligand part. In both the cases the protein part was defined
as receptor molecules. The active site of receptor was identified using a sphere whose
location and radius was adjusted to 9.0 A˚, so that the active sites and the key residues

180
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

of the proteins can be involved in interaction with ligands. The CharmM force field
was applied and the energy was minimized. The hits obtained from pharmacophores

Figure 6.1: Crystal structure of human glucagon receptor (4L6R). Labeled amino acids
(black color) represent the active site in crystal structure.

Figure 6.2: Crystal structure of 11β HSD1 enzyme (2BEL). Labeled amino acids (black
color) represent the active site in crystal structure.

181
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

based virtual screening were energy minimized and used for docking.

Receptor Ligand Docking

The prepared receptors were given an input in ‘input receptor molecule’ parameter in
the LibDOCKER protocol parameter explorer. NCI hits retrieved from the ligand
based virtual screening were given an input using ‘input ligands’ option. The correct
binding mode (poses) of each molecule was elucidated by performing a number of
trials and the poses that were energetically best were retained. The search was stopped
when a certain number of trials have been carried out and/or a sufficient number of
poses have been found for a molecule. The upper limit for pose generation was set to
300. This process helped in finding the correct orientation and, as most ligand
molecules are flexible, the correct conformations of the docked molecule. The
LibDOCKER energy (protein-ligand interaction energies) of best poses docked into
the receptor of all the NCI hits were observed and analyzed.

Result and Discussion

Molecular Docking of NCI Hits (GRA)

The three hits obtained from NCI database were docked into the active site of
protein using LibDock software implemented in DS. Total 300 poses were generated
for each compound. Out of 300, top five poses were evaluated for each molecule on
the basis of LibDock score. SKDGRA1 docked (Figure 6.3) well with a LibDock
score of 127.65. It showed 7 strong hydrogen bond interactions with Gln232,
Leu386, Tyr149, Phe365, Asn238 and Arg308. SKDGRA1 also exhibited 10 weak
Van der Waal interactions with Lys381, Asn238, Phe365, Val191, Leu386, Lys187,
Asn238, Gln232, Leu207 and Leu382. The second lead compound NSC 92543
showed Van der Waal interactions with Ile315, Ser389, Glu362, Leu307, Tyr239,
Lys187 and Val 311 (Figure 6.4) with a LibDock score of 106.262. The third lead
compound NSC 167411 (Figure 6.5) with an promising LibDock score of 126.279
showed strong hydrogen bonding interactions with Glu362, Ser389 and Gln232 and
weak Van der Waal interactions with Phe365, Glu362, Tyr239 and Ile235.
On the basis of molecular docking study results it can be concluded that all the three
compounds have the ability to interact with GR active site amino acid residues

182
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

(Gln232, Leu386, Phe365, Asn238, Lys187), hence the these compounds can be
further developed as potential GR antagonists.

Figure 6.3: Binding pattern of NCI compounds SKDGRA1

Figure 6.4: Binding pattern of NCI compounds SKDGRA2

183
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

Figure 6.5: Binding pattern of NCI compounds SKDGRA3.

Molecular Docking of NCI Hits ( 11β HSD1 Inhibitors)

The three hits obtained from NCI database were docked into the active site of

protein (2BEL) using LibDock software implemented in DS. 100 docking poses

were generated for SKDHSD1 and 247 and 260 docking poses were generated each

for SKDHSD2 and SKDHSD3. Top five poses were evaluated for each compound on

the basis of LibDock score. SKDHSD1 docked well with a Lib Dock score of 81.234

(Figure 6.6). It showed 5 strong hydrogen bond interactions with Asn119, Tyr183,

Lys187, Lys44 and Gly41. SKDHSD1 also exhibited 7 weak Van der Waal

interactions with Gly45, Ile46, Leu215, Ile121 and His120. SKDHSD2 showed 2

strong hydrogen bond interactions with Gly41, and Ser169. It also exhibited 7 weak

Van der Waal interactions with Ser170, Asn119, His120, Ser43, Lys44, Thr220 and

Ile46 (Figure 6.7) with a LibDock score of 106.916. SKDHSD3 showed only one

strong hydrogen bond interaction with Ser169 and 10 weak Van der Waal interactions

(Figure 6.8) with Ser170, Lys187, Asn119, Ile121, Tyr183, Gly41, Ser43, Leu215,

Gly45 and Ile46 with a LibDock score of 103.432.

184

C
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

The results of present docking study reveals that all the three 11β HSD1 Inhibitors

have the potential to interact with crucial active site amino acid residues (Ile46,

Lys187, Asn119, Ser169, Ser170 and Tyr183). In view of potential of identified hits

we envisaged to experimentally evaluate them for anti-diabetic activity.

Figure 6.6: Binding pattern of NCI compounds SKDHSD1

Figure 6.7: Binding pattern of NCI compounds SKDHSD2

185
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

Figure 6.8: Binding pattern of NCI compounds SKDHSD3

Biological Evaluation of Identified Lead Compounds

In view of good estimated activity, fit value, LibDock score and zero Lipinski

violation two representative hits (SKDHSD2 and SKDHSD3) were subjected to

tanimoto similarity indices protocol available in Discovery Studio in order to check

the novelty of the hits. The hits were compared with other ligands crystallized with

11β HSD1 inhibitors. SKDHSD2, SKDHSD3 showed low tanimoto similarity indices

of 0.136 and 0.108 respectively. Since both the compounds exhibited novelty they

were procured and subjected to experimental validation using an in-vivo

streptozotocin (STZ) induced diabetes model as reported previously by Kumar

Dharmendra et al.327

Experimental

Animals
Adult male wistar mice were used for the biological evaluation of SKDHSD2 and
SKDHSD3. Animals were purchased from Chaudhary Charan Singh University,
Hissar, Haryana. All the animals were approximately 9-10 week’s old and weighed
27-35 grams at the beginning of the experiment. Mice were housed in Central animal
186
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

house of Banasthali Vidyapith with a light cycle of 12 hours light/ 12 hours dark and
allowed free choice access to standard rodent chow and water at temperature (22°C)-
controlled environment. All the experimental protocols were approved by IAEC (Reg.
Number: BU/BT/179/2011-12) and throughout the experiment the guidelines of
Committee for the Purpose of Control and Supervision of Experiments on Animals
(CPCSEA) and Institutional Animal Ethics Committee were followed. All animals
were cared in compliance with the Principles of Laboratory Animal Care and the
Guide.

STZ Diabetes Induction

STZ (STZ), a specific β-cell cytotoxin, is commonly used in rodent models for
inducing diabetes. The cytotoxic effect of STZ involves mitochondrial and β-cell
dysfunction with increased oxidative stress, ATP depletion, DNA alkylation, lipid
peroxidation and apoptosis.328 After 14 hours of fasting, mice were injected with STZ
(60 mg/kg intra-peritoneally), which was prepared by dissolving STZ in 0.1M citric
acid buffer (pH 4.5).

Pre-Treatment Methodology
15 adult male wistar mice were divided into the following five groups (n = 3 per
group):
I. Normal control (5% DMSO)
II. STZ (40mg/kg)
III. Carbenoxolone (50mg/kg)
IV. SKDHSD2 (5mg/kg) and
V. SKDHSD3 (5mg/kg)
Group III, IV and V were treated for 3 days with carbenoxolone (50mg/kg),
SKDHSD2 (5mg/kg) and SKDHSD3 (5mg/kg) respectively before STZ diabetes
induction. Plasma samples were collected on 0th, 1st, 3rd and 7th day post STZ
injection, through the tail vein of mice and were immediately used for the estimation
of plasma glucose with a glucometer.

Post Treatment Methodology

15 adult male wistar mice were divided into the following five groups (n = 3 per
group):
187
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

I. Normal control (5% DMSO)

II. STZ (40mg/kg)
III. Carbenoxolone (50mg/kg)
IV. SKDHSD2 (5mg/kg) and
V. SKDHSD3 (5mg/kg)
Groups II-V mice were treated with STZ (60mg/kg). After five days, mice with
fasting plasma glucose level >150mg/dl were considered diabetic. Groups III-V were
selected for further treatment and were treated for 3 days with carbenoxolone
(50mg/kg), SKDHSD2 (5mg/kg) and SKDHSD3 (5mg/kg) respectively. Plasma
samples were collected on 0th, 1st and 3rd day and were immediately used for the
estimation of plasma glucose with a glucometer.

Statistical Analysis
All the experimental values were expressed as mean ± S.E.M and evaluated by
ANOVA.

Results and Discussions

Pre Treatment
The hypoglycemic effect of pretreatment with SKDHSD2 and SKDHSD3 on the
plasma glucose level is shown in table 6.1. SKDHSD2 and SKDHSD3 were found to
improve plasma glucose level in pre treated STZ induced mice. Plasma glucose levels
were found to be significantly reduced (P<0.05) in carbenoxolone and SKDHSD2 and
SKDHSD3 treated groups (122.33 mg/dl ± 2.90 vs. 110 mg/dl ± 1.73 and 116 mg/dl ±
1.52, respectively) in comparison to the STZ induced control group (Figure 6.9).

Table 6.1: Effect of SKDHSD2 and SKDHSD3 on blood glucose level of previously treated
STZ induced albino wistar mice

Blood glucose mg/dl

Treatment
0 day 1st day 3rd day
Control 113.66±2.33 122±10.39 123±5.56
STZ induced diabetic control (35mg/kg) 139±1.52 163±1.73 196.66±4.80
Carbenoxolone +STZ (50mg/kg) 132.33±1.20 119.33±0.88 122.33±2.90
SKDHSD2 + STZ (5mg/kg) 136±2.08 109.33±9.52 110±1.73
SKDHSD3 + STZ (5mg/kg) 129±2.08 110.66±3.75 116±1.52

188
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

Figure 6.9: Effect of SKDHSD2 and SKDHSD3 on blood glucose level of previously treated
STZ induced albino wistar mice

Post Treatment
The hypoglycemic effect of SKDHSD2 and SKDHSD3 on the plasma glucose level is
shown in table 6.2. Plasma glucose levels were found to be reduced (P<0.05) in
carbenoxolone and SKDHSD2 and SKDHSD3 treated groups (153 mg/dl ± 1.15 vs.
134.33 mg/dl ± 2.40 and 145.33 mg/dl ± 1.78, respectively) in comparison to the STZ
induced control groups (Figure 6.10).
The result clearly shows that SKDHSD2 and SKDHSD3 are more potent inhibitor of
11β HSD1 than standard drug carbenoxolone. Undoubtedly the In-silico and In-vivo
work flow adopted in present study has yielded wonderful results.

Table 6.2: Effect of SKDHSD2 and SKDHSD3 on blood glucose level of STZ induced
albino wistar mice

Blood glucose mg/dl

Treatment
0 day 1st day 3rd day
Control 113.66±2.33 122±10.39 123±5.56
STZ induced diabetic control (60mg/kg) 187.33±2.60 215±4.72 246±5.13
Carbenoxolone (50mg/kg) 186.33±2.02 180±1.73 153±1.15
SKDHSD2 (5mg/kg) 179.33±4.33 159.66±3.75 134.33±2.40
SKDHSD3 (5mg/kg) 186.33±3.06 160.33±0.72 145.33±1.78

189
 CHAPTER-6

MOLECULAR DOCKING & EXPERIMENTAL VALIDATION

Figure 6.10: Effect of SKDHSD2 and SKDHSD3 on blood glucose level of STZ induced
albino wistar mice

Conclusion
In brief, two promising hits based on ligand pharmacophore based virtual screening
were identified having anti diabetic activity. The order of activity for test drugs and
carbenoxolone in both the models were found to be SKDHSD2 > SKDHSD3 >
carbenoxolone, which is in agreement to pharmacophore based virtual screening
results. It is noticeable that SKDHSD2 and SKDHSD3 produced the response at a
dose of 5mg/KG whereas a comparable response of carbenoxolone was observed at a
dose of 50 mg/KG, which proves that the identified hits are more potent than standard
drug. Eventually, anti diabetic assay validated the finding of virtual screening based
on ligand pharmacophore model. The results clearly show that In-silico methods are
powerful source for identification of novel chemical entity.

190
SUMMARY
 SUMMARY

Summary

The present thesis entitled “In Silico Ligand Based Pharmacophore Design of GRA
and 11β-HSDI Inhibitors: An Approach to Database Mining and Identification of
New Lead Compounds”, enunciates exploration of essential structural requirements
for development of GRA and 11β HSDI inhibitor and recognition of promising lead
compounds having glucagon receptor antagonistic and 11β HSDI inhibitory activity.
In order to accomplish the aforesaid objectives, several steps like proper selection of
the compound datasets; statistical evaluation of the correlation between compound
dataset and 2D global descriptor; ligand pharmacophore model development and its
interpretation and validation; identification of lead compounds via virtual screening,
structure based pharmacophore model validation for the lead compounds; molecular
docking, and pharmacological in vivo biological evaluation of the lead compounds
were executed.

Establishment of a quantitative relationship between the physical/chemical structures

of the compounds and their specific pharmacological activity by means of 2D and 3D
QSAR is a powerful technique in modern drug designing. The physical/chemical
properties of a compound or 2D descriptors determine the fate of a compound towards
its pharmacological activity on the other hand a 3D ligand based pharmacophore
model conveys information in the form of 3D spatial arrangements and orientation of
chemical features necessary for desired pharmacological activity. Various statistical
parameters and validation techniques are used to confirm the authenticity of the
model. Further there validated models can be used as 3D query to search lead
compounds from the pool of compound libraries.

This whole objective of present work has been divided into three stages. The first
stage is comprised of the development of 2D QSAR/Hansch models for GRA and 11β
HSD1 inhibitors with best statistics. The second stage constitutes development of
three dimensional spatial arrangements and orientation model known as
pharmacophore model for GRA and 11β HSD1 inhibitors. The model with the best
statistical parameters, i.e., highest correlation, best RMS fit and proper difference
between total cost and null cost was subjected for mapping and validation studies.
After validation the model was used as pharmacophore query to identify lead

191
 SUMMARY

compounds from compound libraries, the process called as virtual screening. The last
stage was the confirmation of the identified lead compounds as promising GRA and
11β HSD1 inhibitors has been exercised through structure based pharmacophore
model development and its mapping onto the lead compounds; molecular docking and
in vivo anti-diabetic evaluation of the lead compounds.

Chapter 1
The first part of the chapter 1 provides information on the glucagon and 11β HSD1
enzyme as promising targets, in the fight with type II diabetes mellitus. Moreover this
part also provides insight into three dimensional structures of glucagon and 11β
HSD1 enzymes, working strategies as anti-diabetic target, available drugs in the
market, need of development of new drugs as GRA and 11β HSD1 inhibitors and
their fate. Second part of chapter 1 explains the various drug design techniques such
as ligand based drug design and structure based drug design. The chapter gives a
generalized introduction of CADD and QSAR and its application in designing new
chemical entities. This chapter also describes the brief methodology of 2DQSAR,
ligand based pharmacophore modeling, virtual screening, SBDD and molecular
docking.

Chapter 2
This chapter contains a compilation of reports on GRA and 11β HSD1 inhibitors
synthesized and bio-assayed till date. Apart from this the QSAR studies executed till
date on GRA and 11β HSD1 inhibitors has also been reviewed. This chapter also
highlights the problems underlying behind the development of GRA and 11β HSD1
inhibitors as anti-diabetic targets and the scheme of work which has been followed in
present research endeavour.

Chapter 3
This chapter depicts the 2D global descriptor based QSAR model development by
establishing relationship between most significant descriptors and one GRA series and
two 11β HSD1 inhibitory series i.e. β alanine, isoserine and thiazole (GRA);
Pyrrolidine carboxamide Inhibitors (11β HSD1 inhibitor); and Bicyclo[2.2.2]
octyltriazole Inhibitors (11β HSD1 inhibitor).

192
 SUMMARY

QSAR analysis (MLR, PLS and FFNN) of Series 1 antagonists (β alanine, isoserine
and thiazole) clearly ilustrates that the glucagon receptor antagonistic activity depends
on Verloop L, Dipole moment X and Log P parameters as determined by MLR
analysis and further validated by PLS and FFNN methods. Validation of models by
internal test set clearly shows the robust nature of the model. Multivariate analysis of
Series 2 (Pyrrolidine carboxamide inhibitors) as 11β HSD1 inhibitor demonstrated
dependency on dipole moment X component and log P towards the inhibition of 11β
HSD1 enzyme. The statistical outputs (r, r2, r2CV) of MLR, PLS and FFNN analysis
confirmed the robustness of the developed models. The multivariate analysis of Series
3 (Bicyclo[2.2.2] octyltriazole inhibitors as 11β HSD1 inhibitor) exhibited the
importance of Inertia moment length, Log P and VAMP Polarization XX as important
parameters for the desired 11β HSD1 inhibitory activity. The statistical outputs of
MLR, PLS and FFNN analysis revealed the predictability of the generated model. The
generated models can assist the rational design of promising drug entities with
enhanced GRA and 11β HSD1 inhibitory activity.

Chapter 4
Chapter 4 reports the 3D ligand based pharmacophore models (LBP) developed using
the structure activity based hypothesis generation approach (HypoGen). The LBP
models provides spatial orientations and arrangements of chemical features required
for interaction of receptor with the ligands.
3D LBP modeling of Series 1 (β alanine, isoserine and thiazole) entails the
significance of four pharmacophoric features i.e. one hydrogen bond donor (HBD),
one hydrophobe (HY), one negative ionizable (NI) and one ring aromatic (RA). The
statistical significance of the developed model was evaluated on the basis of
correlation coefficient, cost values and RMS. The developed model was cross
validated to avoid chance correlation by Cat scramble validation, internal and external
test set prediction. The external test set of 9 compounds resulted a squared correlation
coefficient value of 0.65 indicating the predictive ability of the model. As an
additional validation step some clinical trial drugs BAY-27-9955 and LY409021 were
mapped onto the generated pharmacophore model. BAY-27-9955 showed mapping
with two features (HBD and HY) and LY409021 mapped with three features (HBD,
HY and NI).

193
 SUMMARY

3D LBP model development of Series 2 Pyrrolidine carboxamide inhibitors (11β

HSD1 inhibitor) revealed the significance of one hydrogen bond donor (HBD), two
hydrophobic (HY1 and HY2), one positive ionizable (PI) features and four excluded
volumes . The robustness of the model was characterized by a high correlation
coefficient of the training set r value of 0.85 and test set r2 0.79. Finally, 3D LBP
model development of Series 3 Bicyclo[2.2.2]octyltriazole inhibitors (11β HSD1
inhibitor) disclosed the significance of one hydrogen bond acceptor_lipid (HBA_Li),
three hydrophobic (HY1, HY2 and HY3) and ring aromatic (RA) features. The
statistical fitness of the model was assessed on the basis of correlation coefficient
values for the training set and test sets, i.e., 0.85 and 0.79 respectively. The developed
model was validated by aforementioned three validation methods. Carbenoxolone
was mapped onto the developed pharmacophore models and two features (HY1
and HY2) mapping on series 2 model while four features (HBA_Li, HY1, HY2
and HY3) mapping on series 3 model was observed.

Chapter 5
This chapter depicts the approach adopted to identify promising novel scaffolds with
glucagon receptor antagonistic activity and 11β HSD1 inhibitory property. The
validated LBP models were used as a 3D search query to screen NCI database. A
sequential VS procedure was applied, wherein the pharmacophore-based VS was
followed by estimated activity and fit value pre-filtration, and further drug likeness
screening to reduce the number of hits in each different screening step. The LBP
model of Series1, β alanine, isoserine and thiazole derivatives as GRA, was used to
screen NCI database to retrieve 3 potent hits, SKDGRA1, SKDGRA2 and SKDGRA3
with fits value ranging from 7.09-6.60. Similarly the LBP models of Series 2
(Pyrrolidine carboxamide inhibitors) and Series 3 (Bicyclo[2.2.2]octyltriazole
inhibitors), 11β HSD1 inhibitors produced one potent hit SKDHSD1 of fit value
7.893 and two potent hits SKDHSD2 and SKDHSD3 of fit values ranging 8.449-
8.228 respectively. Most importantly all the retrieved hits exhibited zero Lipinski
violation rule.
This chapter also explains the GRA and 11β HSD1 structure based pharmacophore
(SBP) model development. A SBP model helps in understanding the protein ligand
interaction within the protein structural framework, and helps in designing of ligands.
194
 SUMMARY

The crystallized structure of glucagon receptor (4L6R) and 11β HSD1 enzyme
(2BEL) was used for the SBP models development. In order to validate the developed
SBP models; clinical trials drugs, BAY-27-9955 and LY409021 were mapped onto
the GRA model. BAY-27-9955 showed mapping with three features (2HY and
1HBD) and LY409021 mapped with four features (2HBA, 1HY and 1 HBD). The
retrieved hits SKDGRA1, SKDGRA2 and SKDGRA3 were also mapped onto the
SBP model. SKDGRA1 showed five feature mapping (2HBA, 2HBD and 1HY);
SKDGRA2 showed four features (2HBA, 1HBD and 1HY) mapping and SKDGRA3
also showed four features (2HBA, 1HBD and 1HY) mapping. Similarly,
carbenoxolone was also mapped onto the 11β HSD1 structure based model.
Carbenoxolone showed three features (2HBA and 1HY) mapping. The retrieved hits
of 11β HSD1 inhibitors, SKDHSD1, SKDHSD2 and SKDHSD3 were also mapped
onto the SBP model. SKDHSD1 showed three features (1HBA, 1HBD and 1HY)
mapping; SKDHSD2 also showed three features (1HBA, and 2HY) mapping and
SKDHSD3 showed four features (2HBA, and 2HY) mapping.

Chapter 6
This chapter deals with molecular docking and anti-diabetic evaluation of retrieved
lead compounds. Molecular docking is a technique to predict the binding mode of the
ligand molecules. Molecular docking consists of two essential components;
generation of promising ligand binding orientations known as sampling and prediction
of the binding strength for a specific ligand orientations known as scoring. In order to
understand the type of interaction between the chosen hits (obtained by the process of
virtual screening) and the targets glucagon receptor and 11β HSD1 enzyme, molecular
docking was performed using “LibDocker” protocol.
The three hits obtained from modeling of Series 1 (β alanine, isoserine and thiazole)
GRAs were docked into the active site of protein. SKDGRA1 docked well with a
LibDock score of 127.65 with 7 strong hydrogen bond interactions and 10 Van der
Waal interactions. SKDGRA2 showed Van der Waal interactions with a LibDock
score of 106.262. NSC 167411 showed three strong hydrogen bonding interactions
and four Van der Waal interactions with an eminent LibDock score of 126.279.

195
 SUMMARY

Hit retrieved from Series 2 (Pyrrolidine carboxamide-11β HSD1 inhibitor),

SKDHSD1 showed seven strong hydrogen bond interactions and four Van der Waal
interactions with LibDock score of 81.234
SKDHSD2 and SKDHSD3 of Series 3 (Bicyclo[2.2.2] octyltriazole-11β HSD1
inhibitors), exhibited the docking score of 106.916 and 103.432 respectively.
SKDHSD2 showed 2 strong hydrogen bond interactions and 7 Van der Waal
interactions. SKDHSD3 showed only one strong hydrogen bond interaction and 10
Van der Waal interactions.
This chapter also deals with experimental validation of the retrieved hits. Since, GRA
and 11β HSD1 are target for the development of anti-diabetic drugs, the chosen hits
were subjected to anti-diabetic evaluation on STZ induced diabetes animal models.
Two types of studies were performed, treatment with drug before inducing diabetes
and post diabetes treatment model. The 11β HSD1 inhibitors, SKDHSD2 and
SKDHSD3 were found to improve plasma glucose level in both the anti-diabetic
studies. Plasma glucose levels were found to be significantly reduced (P<0.05) in case
of SKDHSD2 and SKDHSD3 in comparison to the carbenoxolone. The order of
activity for test drugs and carbenoxolone were found to be SKDHSD2 > SKDHSD3 >
carbenoxolone, which is in agreement to pharmacophore based virtual screening
results.
It is noticeable that SKDHSD2 and SKDHSD3 produced the response at a dose of
5mg/KG whereas a comparable response of carbenoxolone was observed at a dose of
50 mg/KG, which proves that the identified hits are more potent than standard drug.
In summary, through our well defined pharmacophore based virtual screening we
have identified two potent and structurally diverse 11β HSD1 inhibitors with potent
anti-diabetic activity.

196
REFERENCES
 REFERENCES

References

1. Cnop M, Welsh N, Jonas JC, Jorns A, Lenzen S, Eizirik DL (2005) Mechanisms of

pancreatic β-cell death in type 1 and type 2 diabetes, many differences, few
similarities. Diabetes 54:97-107.

2. Menge BA, Gruber L, Jørgensen SM, Deacon CF, Schmidt WE, Veldhuis JD et al
(2011) Loss of inverse relationship between pulsatile insulin and glucagon secretion
in patients with type 2 diabetes. Diabetes 60:2160-2168.

3. Bishop DB, O'Connor PJ, Desai J (2010) Diabetes In: Remington PL, Brownson R,
Wegner MV (ed) Chronic disease epidemiology and control, 3rd edn. American
Public Health Association, Washington, DC, pp 291--334.

4. Hex N, Bartlett C, Wright D, Taylor M, Varley D (2012) Estimating the current and
future costs of type 1 and type 2 diabetes in the UK, including direct health costs and
indirect societal and productivity costs. Diabetic Medicine 29: 855–62.

5. Kaveeshwar SA, Cornwall J (2014) The current state of diabetes mellitus in India.
Australas Med J 7: 45–48.

6. Danaei G, Finucane MM, Lu Y, Singh GM, Cowan MJ, Paciorek CJ et al (2011) National,
regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980:
systematic analysis of health examination surveys and epidemiological studies with 370
country-years and 2.7 million participants. Lancet 378:31-40.

7. Muoio DM, Newgard CB (2008) Molecular and metabolic mechanisms of insulin

resistance and β-cell failure in type 2 diabetes. Nat Rev Mol Cell Biol 9:193-205

8. Spellman CW (2010) Pathophysiology of type 2 diabetes: Targeting islet cell

dysfunction. J Am Osteopath Assoc 110: S2-7.

9. Unger RH (1971) Glucagon physiology and pathophysiology. N Engl J Med

285:443-449.

10. Definition and diagnosis of diabetes mellitus and intermediate hyperglycemia:

report of a WHO/IDF consultation. Geneva: World Health Organization. (2006) pp
21.

197
 REFERENCES

11. Renosky RJ, Hunt B, Hill-Briggs F, Wray L, Ulbrecht JS (2008) Counseling

People Living With Diabetes. J Rehabil 74:31-40.

12. Mayo KE, Miller LJ, Bataille D, Dalle S, Göke B, Thorens B, Drucker DJ (2003)
International union of pharmacology XXXV. The glucagon receptor family.
Pharmacol Rev 55:167-94.

13. Aronoff SL, Berkowitz K, Shreiner B, Want L (2004) Glucose metabolism and
regulation: beyond insulin and glucagon. Diabetes Spectrum 17:183-190.

14. Holst JJ (2010) Glucagon and glucagon-like peptides 1 and 2. Results Probl Cell
Differ 50:121-35.

15. Christensen M, Vedtofte L, Holst JJ, Vilsboll T, Knop FK (2011) Glucose-

dependent insulinotropic polypeptide: a bifunctional glucose-dependent regulator of
glucagon and insulin secretion in humans. Diabetes 60:3103-3109.

16. Quesada I, Tuduri E, Ripoll C, Nada A (2008) Physiology of the pancreatic a-cell
and glucagon secretion: role in glucose homeostasis and diabetes. J Endocrin 199:5–
19.

17. Thorens B (2011) Brain glucose sensing and neural regulation of insulin and
glucagon secretion. Diabetes Obes Metab 1:82-88.

18. D'Alessio D (2011) The role of dysregulated glucagon secretion in type 2

diabetes. Diabetes Obes Metab 13:126-32.

19. Rivera N, Everett-Grueter CA, Edgerton DS, Rodewald T, Neal DW, Nishimura
E et al (2007) A novel glucagon receptor antagonist, NNC 25-0926, blunts hepatic
glucose production in the conscious dog. J Pharmacol Exp Ther 321:743-752.

20. Jamison RA, Stark R, Dong J, Yonemitsu S, Zhang D, Shulman G (2011)

Hyperglucagonemia precedes a decline in insulin secretion and causes hyperglycemia
in chronically glucose-infused rats. Am J Physiol Endocrinol Metab 301:1174–1183.

21. Brubaker PL, Drucker DJ (2002) Structure-function of the glucagon receptor

family of g protein–coupled receptors: the glucagon, GIP, GLP-1, and GLP-2
receptors. Receptors Channels 8:179–188.
198
 REFERENCES

22. Iwanij V, Hur, KC (1985) Direct cross-linking of 125I-labeled glucagon to its

membrane receptor by UV irradiation. Proc Natl Acad Sci U. S. A 82:325-329.

23. Wakelam MJ, Murphy GJ, Hruby VJ, Houslay MD (1986) Activation of two
signal-transduction systems in hepatocytes by glucagon. Nature 323:68–71.

24. Li XC, Zhuo JL (2007) Targeting glucagon receptor signalling in treating

metabolic syndrome and renal injury in Type 2 diabetes: theory versus promise. Clin
Sci (Lond) 113:183–193.

25. Hoare SR (2005) Mechanisms of peptide and nonpeptide ligand binding to class B
G-protein-coupled receptors. Drug Discov Today 10:417–427.

26. Lagerstrom MC, Schioth HB (2008) Structural diversity of G protein-coupled

receptors and significance for drug discovery. Nature Rev Drug Discov 7:339–357.

27. Siu FY, He M, De Graaf C, Han, GY, Yang D, Zhang Z et al (2013) Structure of
the human glucagon class B G-protein-coupled receptor. Nature 499: 444-449.

28. Cho YM, Merchant CE, Kieffer TJ (2012) Targeting the glucagon receptor family
for diabetes and obesity therapy. Pharmacol Ther 135:247–278.

29. Goldfine I D, Roth J, Birnbaumer L (1972) Glucagon receptors in -cells. binding

of 125 I-glucagon and activation of adenylate cyclase. J Biol Chem 247:1211–1218.

30. Bregman MD, Trivedi D, Hruby VJ (1980) Glucagon amino groups. Evaluation of
modifications leading to antagonism and agonism. J Biol Chem 255:11725–11731.

31. Unson CG, Macdonald D, Ray K, Durrah TL, Merrifield RB (1991) Position 9
replacement analogs of glucagon uncouple biological activity and receptor binding. J
Biol Chem 266:2763–2766.

32. Potterat O, Wagner K, Gemmecker G, Mack J, Puder C, Vettermann R (2004) BI-

32169, a bicyclic 19-peptide with strong glucagon receptor antagonist activity from
Streptomyces sp. J Nat Prod 67:1528–1531.

33. Madsen P, Knudsen LB, Wiberg FC, Carr RD (1998) Discovery and structure–
activity relationship of the first non-peptide competitive human glucagon receptor
antagonists. J Med Chem 41:5150–5157.
199
 REFERENCES

34. Cascieri MA, Koch GE, Ber E, Sadowski SJ, Louizides D, de Laszlo SE (1999)
Characterization of a novel, non-peptidyl antagonist of the human glucagon receptor.
J Biol Chem 274:8694-8697.

35. Chang LL, Sidler KL, Cascieri MA, De Laszlo S; Koch G, Li B (2001)
Substituted imidazoles as glucagon receptor antagonists. Bioorg Med Chem Lett 11:
2549–53.

36. Parker JC, McPherson RK, Andrews KM, Levy CB, Dubins JS, Chin JE (2000)
Effects of skyrin, a receptor-selective glucagon antagonist, in rat and human
hepatocytes. Diabetes 49:2079–2086.

37. Ling A, Hong Y, Gonzalez J, Gregor V, Polinsky A, Kuki, A et al (2001)

Identification of alkylidene hydrazides as glucagon receptor antagonists. J Med Chem
44:3141–3149.

38. Petersen K F, Sullivan JT (2001) Effects of a novel glucagon receptor antagonist

(Bay 27-9955) on glucagon-stimulated glucose production in humans. Diabetologia
44:2018–2024.

39. Deeg MA, Abu-Raddad EJ, Tham LS, Fu H, Pinaire JA, Kelly RP (2011) Single
dose of the glucagon receptor antagonist LY2409021 reduces blood glucose in
healthy subjects and patients with type 2 diabetes mellitus. Diabetologia 54:S362.

40. Brekke OH, Sandlie I (2003) Therapeutic antibodies for human diseases at the
dawn of the twenty-first century. Nat Rev Drug Discov 2:52–62.

41. Brand CL, Rolin B, Jorgensen PN, Svendsen I, Kristensen JS, Holst JJ (1994)
Immunoneutralization of endogenous glucagon with monoclonal glucagon antibody
normalizes hyperglycemia in moderately streptozotocin-diabetic rats. Diabetologia
37:985-993.

42. Brand CL, Jorgensen PN, Svendsen I, Holst JJ (1996) Evidence for a major role
for glucagon in regulation of plasma glucose in conscious, nondiabetic, and alloxan-
induced diabetic rabbits. Diabetes 45:1076–1083.

200
 REFERENCES

43. Nader N, Chrousos GP, Kino T (2010) Interactions of the circadian CLOCK
system and the HPA Axis. Trends Endocrinol Metab 21:277–286.

44. Nader N, Ng SS, Lambrou GI, Pervanidou P, Wang Y, Chrousos GP et al (2010)

AMPK regulates metabolic actions of glucocorticoids by phosphorylating the
glucocorticoid receptor through p38 MAPK. Mol Endocrinol 24:1748–1764.

45. Wilcox G (2005) Insulin and Insulin Resistance. Clin Biochem Rev 26:19-39.

46. Adam TC, Epel ES (2007) Stress, eating and the reward system. Physiol Behav
91:449–458

47. Van Rossum EF, Russcher H, Lamberts SW (2005) Genetic polymorphisms and
multifactorial diseases: facts and fallacies revealed by the glucocorticoid receptor
gene. Trends Endocrinol Metab 16:445–450.

48. Vetere A, Choudhary A, Burns SM, Wagner BK (2014) Targeting the pancreatic
β-cell to treat diabetes. Nat Rev Drug Discov 13: 278–289.

49. Dickmeis T (2009) Glucocorticoids and the circadian clock. J Endocrinol 200:3–
22.

50. Adinoff B, Iranmanesh A, Veldhuis J, Fisher L (1998) Disturbances of the Stress

Response: The Role of the HPA Axis During Alcohol Withdrawal and Abstinence.
Alcohol Health Res World 22:67-72.

51. Szablewski L (2011) Glucose homeostasis and insulin resistance. Bentham

Science Publishers.

52. Beaudry JL, Riddell MC (2012) Effects of glucocorticoids and exercise on

pancreatic β-cell function and diabetes development. Diabetes Metab Res Rev 28:
560–573.

53. Klemcke HG, Kumar RS, Yang K, Vallet JL, Christenson RK (2003) 11β-
hydroxysteroid dehydrogenase and glucocorticoid receptor messenger rna expression
in porcine placentae: effects of stage of gestation, breed, and uterine environment.
Biol Reprod 69:1945-1950.

201
 REFERENCES

54. Bates HE, Sirek A, Kiraly MA, Yue JT, Riddell MC, Matthews SG et al (2008)
Adaptation to intermittent stress promotes maintenance of beta-cell compensation:
comparison with food restriction. Am J Physiol Endocrinol Metab 295:947–958.

55. Calabrese A, Zhang M, Serre-Beinier V, Caton D, Mas C, Satin LS et al (2003)

Connexin 36 controls synchronization of Ca2+ oscillations and insulin secretion in
MIN6 cells. Diabetes 52:417-424.

56. Schacke H, Rehwinkel H, Asadullah K, Cato AC (2006) Insight into the

molecular mechanisms of glucocorticoid receptor action promotes identification of
novel, ligands with an improved therapeutic index. Exp Dermatol 15:565–573.

57. Sandeep TC, Walker BR (2001) Pathophysiology of modulation of local

glucocorticoid levels by 11β-hydroxysteroid dehydrogenases. Trends in Endocrin Met
12:446-453.

58. White PC, Mune T, Rogerson FM, Kayes KM, Agarwal AK (1997) 11β-
Hydroxysteroid dehydrogenase and its role in the syndrome of apparent
mineralocorticoid excess, Pediatr Res 41:25–29.

59. Krozowski Z, Li KXZ, Koyama K, Smith RE, Obeyesekere VR, Stein-Oakley A,

Sasano H, Coulter C & Sheppard KE (1999) The type 1 and type II 11β
hydroxysteroid dehydrogenase enzymes. Journal of Steroid Biochemistry and
Molecular Biology 69:391–401.

60. Walker EA, Stewart PM (2003) 11β-hydroxysteroid dehydrogenase: unexpected

connections. Trends Endocrinol Metab 14:334-339.

61. Wang M (2005) The role of glucocorticoid action in the pathophysiology of the
Metabolic Syndrome. Nutrition & Metabolism, 2:3

62. Tomlinson JW, Stewart PM (2007) Modulation of glucocorticoid action and the
treatment of type 2 diabetes. Best practice & Research. Clinical Endocrinology &
Metabolism 21:607-619.

63. Draper N, Stewart PM (2005) 11beta-hydroxysteroid dehydrogenase and the pre-

receptor regulation of corticosteroid hormone action. J Endocrinol 186:251–271.

202
 REFERENCES

64. Hosfield DJ, Wu Y, Skene RJ, Hilgers M, Jennings A, Snell GP, Aertgeerts K
(2005) Conformational flexibility in crystal structures of human 11β hydroxysteroid
dehydrogenase type 1 provide insights into glucocorticoid interconversion and
enzyme regulation. J. Biol. Chem 280:4639-4648.

65. Lipson VV, Zamigajlo LL, Petrova ON (2011) Development of 11βHSD1

inhibitors for the treatment of metabolic syndrome. Ukrainica Bioorganica Acta 2:3-
13.

66. Nixon M, Wake DJ Livingstone DE, Stimson RH, Esteves CL, Seckl JR et al
(2012) Salicylate downregulates 11β-HSD1 expression in adipose tissue in obese
mice and in humans, mediating insulin sensitization. Diabetes 61:790-796.

67. Stewart PM, Tomlinson JW (2009) Selective Inhibitors of 11β-Hydroxysteroid

Dehydrogenase Type 1 for Patients With Metabolic Syndrome:Is the Target Liver,
Fat, or Both? Diabetes 58:14–15.

68. Morris DJ, Souness GW, Latif SA, Hardy MP, Brem AS (2004) Effect of
chenodeoxycholic acid on 11 beta hydroxysteroid dehydrogenase in various target
tissues. Metabolism 53:811-816.

69. Hughes KA, Webster SP, Walker BR (2008) 11Betahydroxysteroid

dehydrogenase type 1 (11βHSD1) inhibitors in type 2 diabetes mellitus and obesity.
Expert Opin Investig Drugs 17:481-496.

70. Barf T, Vallgarda J, Emond R, Haggstrom C, Kurz G, Nygren, A et al (2002a)

Arylsulfonamidothiazoles as a new class of potential antidiabetic drugs. Discovery of
potent and selective inhibitors of the 11 beta-hydroxysteroid dehydrogenase type 1. J
Med Chem 45:3813-3815.

71. Alberts P, Nilsson C, Selen G, Engblom LO, Edling NH, Norling S et al (2003b)
Selective inhibition of 11 beta-hydroxysteroid dehydrogenase type 1 improves hepatic
insulin sensitivity in hyperglycemic mice strains. Endocrinology 144:4755-4762.

72. Alberts P, Engblom L, Edling N, Forsgren M, Klingström G, Larsson C et al

(2002) Selective inhibition of 11β-hydroxysteroid dehydrogenase type 1 decreases
blood glucose concentrations in hyperglycaemic mice. Diabetologia 45:1528–1532
203
 REFERENCES

73. Hermanowski-Vosatka A, Balkovec JM, Cheng K, Chen HY, Hernandez M, Koo

GC et al (2005) 11β-HSD1 inhibition ameliorates metabolic syndrome and prevents
progression of atherosclerosis in mice. J Exp Med 202:517–527.

74. Kaur J, Sharma AK, Sharma D (2014) Overview of drug innovation to

commercialization. J Drug Deliv Therap 4:36-43.

75. Congreve M, Murray CW, Blundell TL (2005). Keynote review: Structural

Biology and Drug Discovery. Drug Discov Today 10:895-907.

76. Herrling PL (2005) The drug discovery process Progress in Drug Research. In
Rudin M (ed) Birkhäuser Verlag, Basel, Switzerland.

77. Marton MJ, Derisi JL, Bennett HA, Iyer VR, Meyer MR et al. (1998) Drug target
validation and identification of secondary drug target effects using DNA microarrays.
Nature Medicine 4:1293-1301.

78. Wise A, Gearing K, Rees S (2002) Target validation of G-protein coupled

receptors. Drug Discov Today 7:235-246.

79. Martis EA, Somani RR (2012) Drug Designing, Discovery and Development
Techniques. In Basnet Purusotam (Ed.) Promising pharmaceuticals. Intech, Europe.

80. Bleicher KH, Bohm HJ, Muller K, Alanine AI (2003) Hit and Lead generation:
Beyond, High-Throughput Screening. Nature Rev Drug Discov 2:369-378.

81. Xue L, Bajorath J (2000). Molecular descriptors in chemoinformatics,

computational combinatorial chemistry, and virtual screening. Comb Chem High
Throughput Screen 3:363-372.

82. Kalyani G, Sharma D, Vaishnav Y, Deshmukh VS (2013) A review on drug

designing, methods, its application and prospects. Int J Pharm Res Develop 5:15-30.

83. Caldwell GW, Ritchie DM, Masucci JA, Hageman W, Yan Z (2001) The new pre-
preclinical paradigm: compound optimization in early and late phase drug discovery.
Curr Top Med Chem 1:353-366.

84. Singh R, Nagori BP, Soni B, Singh JV (2011) A Review: Clinical Trial and Data
Management. Pharmacophore 2:200-209.
204
 REFERENCES

85. Snyder JP (1991) Computer-assisted drug design. Part I. Conditions in the 1980s.
Med Res Rev 11:641-62.

86. Beddell CR, Goodford PJ, Norrington FE, Wilkinson S, Woottonet R (1976)
Compounds designed to fit a site of known structure in human haemoglobin. Br J
Pharmacol 57:201–209.

87. Hansch C, Fujita T (1964) p-σ-π Analysis. A method for the correlation of
biological activity and chemical structure. J Am Chem Soc 86:1616-1626.

88. Abraham DJ (1998) Burger’s Medicinal Chemistry and Drug Discovery, Vol. 1,
John Willey & Sons. pp.533.

89. Kubinyi H (1990) Quantitative structure-activity relationships (QSAR) and

molecular modelling in cancer research. J Cancer Res Clini Oncol 116:529-537.

90. Crum-Brown A, Fraser TR (1868) On the connection between chemical

constitution and physiological action. Part 1. On the physiological action of the
ammonium bases, derived from Strychia, Brucia, Thebaia, Codeia, Morphia and
Nicotia. Trans Roy Soc Edinburg 25:151-203.

91. Richet MC (1893) Noté sur la Rapport entre la Toxicité et les Propriétés
Physiques des Corps. Compt Rend Soc Biol (Paris) 45:775-776

92. Lipnick RL, Overton CE (1986) Narcosis studies and a contribution to general
pharmacology. Trends Pharmacol Sci 7:161-164.

93. Kuzmin VE, Artemenko AG, Polischuk PG, Muratov EN, Hromov AI,
Liahovskiy AV (2005) Hierarchic system of QSAR models (1D-4D) on the base of
simplex representation of molecular structure. J Mol Model 11: 457-467.

94. Hammett LP (1970) Physical Organic Chemistry 2nd Ed. McGraw-Hill, New
York.

95. Taft RW (1952) Polar and steric substituent constants for aliphatic and o-benzoate
groups from rates of esterification and hydrolysis of esters. J Am Chem Soc 74:3120-
3128.

205
 REFERENCES

96. Hansch C, Maloney PP, Fujita T, Muir RM (1962) Correlation of Biological

Activity of Phenoxyacetic Acids with Hammett Substituent Constants and Partition
Coefficients. Nature 194:178-180.

97. Hansch C, Leo A (1995) Fundamentals and Applications in Chemistry and

Biology. In Heller SR (Ed), Exploring QSAR. American Chemical Society,
Washington, DC.

98. Saxena AK, Ram S (1979) Quantitative structure-activity relationships. Progg

Drug Res 23:199-232.

99. Free SM Jr, Wilson JW (1964) A mathematical contribution to structure-activity

studies. J Med Chem 7:395-399.

100. Cammarata A, Bustard TM (1974) Reinvestigation of a nonadditive quantitative

structure-activity relation. J Med Chem 17:981-985.

101. Osman R, Weinstein H, Green JP (1979) Parameters and methods in Quantitative

Structure-Activity Relationships. Comp Assist Drug Design 2:21–77.

102. Vedani A, Dobler M, Lill MA (2006) The challenge of predicting drug toxicity
in silico. Basic Clini Pharmacol Toxicol 99:195–208.

103. Myintz KZ, Xie XQ (2010) Recent advances in fragment-based QSAR and
multi-dimensional QSAR methods. Int J Mol Sci 11:3846-3866.

104. Verma J, Khedkar VM, Coutinho EC (2010) 3D-QSAR in drug design a review.
Curr Top Med Chem 10:95-115.

105. Kuzmin VE, Artemenko AG, Muratov EN, Polischuk PG, Ognichenko LN,
Liahovsky AV (2010) Virtual Screening and Molecular Design Based on Hierarchical
Qsar Technology, Recent Advances in QSAR Studies. Chall Adv Comp Chem Phy
8:127-176.

106. Ferreira MMC (2002) Multivariate QSAR. J Brazil Chem Soc 13:742-753.

107. Cheng F, Sutariya V (2012) Applications of Artificial neural network modeling

in drug discovery. Clin Exp Pharmacol 2:1-2.

206
 REFERENCES

108. Hansch C (1969) A quantitative approach to biochemical structure-activity

relationships. Acc Chem Res 2: 232.

109. Liljefors T, Krogsgaard-Larsen P, Madsen U (2003) Textbook of Drug Design

and Discovery. 3rd Ed CRC Press

110. Athri P (2008) Application of computer-aided drug discovery methodologies

towards the rational design of drugs against infectious diseases. Dissertation, Georgia
State University.

111. Hammett, LP (1935) Some relations between reaction rates and equilibrium
constants. Chem Rev 17:125.

112. Stanton DT (2012) QSAR and QSPR model interpretation using partial least
squares (PLS) analysis. Curr Comp Aid Drug Design 8:107-27.

113. Assafrao D, Mohallem JR (2007) The isotopic dipole moment of HDO. J Phy
40:85-91.

114. Clare BW (1994) Frontier orbital energies in quantitative structure-activity

relationships: A comparison of quantum chemical methods. Theo Chim Acta 87:415-
430.

115. Seward JR, Cronin MT, Schultz TW (2002) The effect of precision of molecular
orbital descriptors on toxicity modeling of selected pyridines. SAR QSAR Environ
Res 13:325-40.

116. Singh PP, Sharma SB, Singh K (2010) Quantum chemical and energy descriptors
based QSAR study of triazine derivatives. J Chem Pharmaceutic Res 2:193-205.

117. Datta D (1991) Taft's substituent constants, σ* and σI, and Huheey's group
electronegativity. J Phy Org Chem 4:96–100.

118. Labute P (2004) Derivation and applications of molecular descriptors based on

approximate surface area. Methods Mol Biol 275:261-78.

119. Harper KC, Bess EN, Sigman MS (2012) Multidimensional steric parameters in
the analysis of asymmetric catalytic reactions. Nature Chem 4:366-374.

207
 REFERENCES

120. Singh P, Singh JP (2011) Topological descriptor based QSAR study of

benzamidine as inhibitor of thrombin. J Chem Pharmaceutic Res 3:296-303.

121. Leo A, Hansch C, Elkins C (1971) Partition coefficients and their uses. Chem
Rev 71:525–616.

122. Reddy KN, Locke MA (1994) Relationships between molecular properties and
log p and soil sorption (koe) of substituted phenylureas: qsar models. Chemosphere
28:1929-1941.

123. Sun H (2004) A universal molecular descriptor system for prediction of LOGp,
LOGs, LOGb, and absorption. J Chem Inf Comp Sci 44:748–757.

124. Buiu AS, Seiman CD, Seiman DD, Borcan C, Mihailescu F (2012) Evaluation of
the pharmacological descriptors related to the induction of antidepressant activity and
its prediction by QSAR/QRAR methods. Mini Rev Med Chem 12:467-476.

125. Hua H (2009) Wiener and Schultz Molecular topological indices of graphs with
specified cut edges. Commun Math Comput Chem 61:643-651.

126. Ghorbani M (2013) Remarks on the balaban index Serdica. J Comput 7:25–34

127. Gupta M, Gupta S, Dureja H, Madan AK (2012) Superaugmented eccentric

distance sum connectivity indices: novel highly discriminating topological descriptors
for QSAR/QSPR. Chem Biol Drug Design 79:38-52.

128. Giri A, Mishra BB, Tripathi P, Agrawal PP, Singh PP (2011) Quantitative
structure-activity relationship of matrix metalloproteinase inhibitors based on
topological descriptors. J Chem Pharmaceutic Res 3:290-305.

129. Kier LB, Hall LH (1990) An electrotopological-state index for atoms in

molecules. Pharmaceutical Research 7:801-807.

130. Lowell HH, Brian M, Lemont BK (1991) The electrotopological state: an atom
index for QSAR. Quantitative Structure Activity Relationships 10: 43-51.

131. Saliner AG, Girones X (2005) Topological quantum similarity measures:

applications in QSAR. J Mol Struc THEOCHEM 727:97–106.

208
 REFERENCES

132. Wold S, Dunn III WJ (1983) Multivariate quantitative structure-activity

relationships (QSAR): conditions for their applicability. J Chem Inf Comput Sci 23:6-
13.

133. Xua J, Zhanga H, Wanga L, Liangab G, Wanga L, Shen X (2011) Artificial

neural network-based QSPR study on absorption maxima of organic dyes for dye-
sensitised solar cells. Mol Simul 37:1-10.

134. Kubinyi H (1993) QSAR: Hansch Analysis and Related Approaches. Willey
Publishers, Weinheim, New York.

135. Catalyst. Accelrys Software, San Diego, CA; http://www.accelrys.com.

136. Mason JS, Good AC, Martin EJ (2001) 3D pharmacophores in drug discovery.
Cur Pharm Des 7:567-597.

137. Milne G W, Nicklaus MC, Wang S (1998) Pharmacophores in drug design and
discovery. SAR QSAR Environ. Res 9:23-38.

138. McCluskey A, Keller PA, Morgan J, Garnera J (2003) Synthesis, molecular

modeling and biological activity of methyl and thiomethyl substituted pyrimidines as
corticotropin releasing hormone type 1 antagonists. Org Biomol Chem (Suppl).

139. Guner OF (2000) Pharmacophore Perception, Development, and Use in Drug

Design, Volume 2, Internat'l University Line.

140. Shui L, Bharatham,N, Bharatham K, Lee KW (2006) Urokinase inhibitor design

based on pharmacophore model derived from diverse classes of inhibitors. Bioinform
Biosys 2:115-122.

141. Stahura FL, Bajorath J (2005) New methodologies for ligand-based virtual
screening. Curr Pharmaceutic Design 11:1189-202.

142. Blundell TL, Patel S (2004) High-throughput X-ray crystallography for drug
discovery. Current Opin Pharmacol 4:490-496.

143. Ziarek JJ, Peterson FC, Lytle BL, Volkman BF (2011) Binding site identification
and structure determination of protein-ligand complexes by NMR. Methods Enzymol
493:241–275.
209
 REFERENCES

144. Huang SY, Zou X (2010) Advances and Challenges in Protein-Ligand Docking.
Int J Mol Sci 11:3016–3034.

145. Kroemer RT (2007) Structure-Based Drug Design: Docking and Scoring.

Current Protein and Peptide Science 8:312-328.

146. Grigoryan AV, Wang H, Cardozo TJ (2012) Can the Energy Gap in the Protein-
Ligand Binding Energy Landscape Be Used as a Descriptor in Virtual Ligand
Screening? Plos One 7:e46532

147. Kitchen DB, Decornez H, Furr JR, Bajorath J (2004) Docking and scoring in
virtual screening for drug discovery: methods and applications. Nature Reviews Drug
Discov 3:935-949.

148. Gurd RS, Flanders KC, Mar DH, Folz RJ, England RD, Coolican SA et al (1982)
Semisynthetic Derivatives of Glucagon: (Des-His')Ne-acetimidoglucagon and Na-
Biotinyl-N'-acetimidoglucagon. Biochemistry 21:4244-4251.

149. Coy DH, Sueiras-Diaz J, Lance VA, Murphy WA (1984) Structure-activity

studies on the N-terminal region of glucagon. J Med Chem 27:310-315.

150. Gysin B, Trivedi D, Johnson DG II, Hruby VJ (1986) Design and synthesis of
glucagon partial agonists and antagonists. Biochemistry 25:8278-8284.

151. Wakelam MJO, Murphy GJ, Hruby VJ, Houslay MD (1986) Activation of two
signal-transduction systems in hepatocytes by glucagon. Nature 323:68-71.

152. Gysin B, Johnson DG, Trivedi D, Hruby VJ (1987) Synthesis of Two Glucagon
Antagonists: Receptor Binding, Adenylate Cyclase, and Effects on Blood Plasma
Glucose Levels. J Med Chem 30:1409-1415.

153. Unson CG, Gurzenda EM, Merrifield RB (1989) Biological activities of des-
His1[Glu9]glucagon amide, a glucagon antagonist. Peptides 10:1171–1177.

154. Post SR, Rubinstein PG, Tager HS (1993) Mechanism of action of des-His1-
[Glu9]glucagon amide, a peptide antagonist of the glucagon receptor system (saponin-
permeabilized hepatocytes/ glucagon-receptor interactions/ adenylyl cydase/ GTP/
Mg2+). Proc Natl Acad Sci USA 90:1662-1666.
210
 REFERENCES

155. Hruby VJ, Azizeh BY, Tine BAV, Sturm NS, Hutzler AM, David C (1995) [des
Hisl, des Phe6, GIu9] glucagon amide: a newly designed "pure" glucagon antagonist,
Bioorg Med Chem Lett 5:1849-1852.

156. Hruby VJ, Azizeh BY, Shenderovich MD, Trivedi D, Li G, Sturm NS (1996)
Topographical Amino Acid Substitution in Position 10 of Glucagon Leads to
Antagonists/Partial Agonists with Greater Binding Differences. J Med Chem 39:2449-
2455

157. Hruby VJ, Azizeh BY Tine BAV, Trivedi D (1997) Pure glucagon antagonists:
biological activities and camp accumulation using phosphodiesterase inhibitors.
Peptides 18:633–641.

158. Azizeh BY, Ahn JM, Caspari R, Shenderovich MD, Trivedi D, Hruby VJ (1997)
The role of phenylalanine at position 6 in glucagon's mechanism of biological
action: multiple replacement analogues of glucagon. J Med Chem 40:2555–2562.

159. Guillom J, Dallemagne P, Pfeiffer B, Renard P, Manechez D, Kervrand A, Raulv

S (1998) Synthesis of new pyrrolo[1,2-a]quinoxalines: potential non-peptide glucagon
receptor antagonists. Eur J Med Chem 33:293-308.

160. Hruby VJ, Sturm NS, Lin Y, Burley SK, Krstenansky JL, Ahn JM (1998)
Structure function studies on positions 17, 18, and 21 replacement analogues of
glucagon: the importance of charged residues and salt bridges in glucagon
biological activity. J Med Chem 41:2693–2700.

161. de Laszlo SE, Hacker C, Li B, Kim D, MacCoss M, Mantlo N (1999) Potent,

orally absorbed glucagon receptor antagonists. Bioorg Med Chem Lett 9:641-646.

162. Cascieri MA, Koch GE, Ber E, Sadowski SJ, Louizides D, de Laszlo SE (1999)
Characterization of a novel, non-peptidyl antagonist of the human glucagon receptor.
J Biol Chem 274:8694–8697.

163. Hruby VJ, Trivedi D, Lin Y, Ahn JM, Siegel M, Mollova NN (2000) Design and
synthesis of conformationally constrained glucagon analogues. J Med Chem 43:1714-
1722.

211
 REFERENCES

164. Sivarajah P, Wheeler MB, Irwina DM (2001) Evolution of receptors for

proglucagon-derived peptides: isolation of frog glucagon receptors. Comp Biochem
Physiol Part B 128:517-527.

165. Hruby VJ, Ahn JM, Medeiros M, Trivedi D (2001) Development of potent
truncated glucagon antagonists J Med Chem 44:1372-1379.

166. Hruby VJ, Ahn JM, Gitu PM , Medeiros M, Swift JR , Trivedi D (2001) A new
approach to search for the bioactive conformation of glucagon: positional
cyclization scanning. J Med Chem 44:3109–3116.

167. Ladouceur GH, Cook JH, Doherty EM, Schoen WR, MacDougall MR,
Livingston JN (2002) Discovery of 5-Hydroxyalkyl-4-phenylpyridines as a new class
of glucagon receptor antagonists. Bioorg Med Chem Lett 12:461–464

168. Ling A, Plewe M, Gonzalez J, Madsen P, Sams CK, Lau J et al (2002) Human
glucagon receptor antagonists based on alkylidene hydrazides. Bioorg Med Chem Lett
12:663–666.

169. Smith RA, Hertzog DL, Osterhout MH, Ladouceur GH, Korpusik M, Bobko MA
et al (2002) Optimization of the 4-aryl group of 4-aryl-pyridine glucagon antagonists:
development of an efficient, alternative synthesis. Bioorg Med Chem Lett 12:1303–
1306.

170. Ladouceur GH, Cook JH, Hertzog DL, Jones JH, Hundertmark T, Korpusik M et
al (2002) Integration of optimized substituent patterns to produce highly potent 4-
aryl-pyridine glucagon receptor antagonists. Bioorg Med Chem Lett 12:3421–3424.

171. Ladouceur GH, Cook JH, Doherty EM, Giebel D, Schoen WR (2002) 4-
Phenylpyridine glucagon receptor antagonists: synthetic approaches to the sterically
hindered chiral hydroxy group. Tetrahedron Lett 43:4455–4458.

172. Madsen P, Ling A, Plewe M, Sams CK, Knudsen LB, Sidelmann UG et al

(2002) Optimization of alkylidene hydrazide based human glucagon receptor
antagonists. Discovery of the highly potent and orally available 3-cyano-4-
hydroxybenzoic acid [1-(2,3,5,6-Tetramethylbenzyl)-1H-indol-4-ylmethylene]
hydrazide. J Med Chem 45:5755-5775.
212
 REFERENCES

173. Unson CG, Wu CR, Jiang Y, Yoo B, Cheung C, Sakmar TP et al (2002) Roles of
specific extracellular domains of the glucagon receptor in ligand binding and
signaling. Biochemistry 41:11795–11803.

174. Sapsea AM, Lawtona S, Rothchilda R, Unson CG et al (2003) An ab initio study

of non-peptide glucagon receptor antagonists. J Mol Struc (Theochem) 638:135–145.

175. Hruby VJ, Ying J, Ahn JM, Jacobsen NE, Brown MF (2003) NMR solution
structure of the glucagon antagonist [deshis1, desphe6, glu9]glucagon amide in the
presence of perdeuterated dodecylphosphocholine micelles. Biochemistry 42:2825–
2835.

176. Kurukulasuriya R, Sorensen BK, Link JT, Patel JR, Jae HS, Winn MX et al
(2004) Biaryl amide glucagon receptor antagonists. Bioorg Med Chem Lett 14:2047–
2050

177. Jiang G, Dallas-Yang Q, Shenb X, Strowski M, Bradyd E, Sapersteind R et al

(2004) Hepatic glucagon receptor binding and glucose-lowering in vivo by peptidyl
and non-peptidyl glucagon receptor antagonists. Eur J Pharmacol 501:225– 234

178. Hsu WH, Yibchok-anuna S, Abu-Bashab EA, Yaoc CY, Panichkriangkraia W

(2004) The role of arginine vasopressin in diabetes-associated increase in glucagon
secretion. Regulatory Peptides 122:157– 162.

179. Potterat O, Wagner K, Gemmecker G, Mack J, Puder C, Vettermann R et al

(2004) BI-32169, a Bicyclic 19-Peptide with Strong Glucagon Receptor Antagonist
Activity from Streptomyces sp. J Nat Prod 67:1528-1531.

180. Shen DM, Zhang F, Brady EJ, Candelore MR, Dallas-Yang Q, Ding VDH et al
(2005) Discovery of novel, potent, and orally active spiro-urea human glucagon
receptor antagonists. Bioorg Med Chem Lett 15:4564–4569

181. Duffy JL, Kirk BA, Konteatis Z, Campbell EL, Liang R, Brady EJ et al (2005)
Discovery and investigation of a novel class of thiophene-derived antagonists of the
human glucagon receptor. Bioorg Med Chem Lett 15:1401–1405.

213
 REFERENCES

182. Bolli GB, Fanelli CG, Porcellati F, Rossetti P (2006) Glucagon: The effects of its
excess and deficiency on insulin action. Nutri Metab Cardiovasc Dis 16:S28-S34.

183. Duffy JL, Cohen SM, Miller C, Kirk BA, Candelore MR, Ding VDH et al (2006)
Direct observation (NMR) of the efficacy of glucagon receptor antagonists in murine
liver expressing the human glucagon receptor. Bioorg Med Chem 14:1506–1517.

184. Liang R, Abrardo L, Brady EJ, Candelore MR, Ding V, Saperstein R et al (2007)
Design and synthesis of conformationally constrained tri-substituted ureas as potent
antagonists of the human glucagon receptor. Bioorg Med Chem Lett 17:587–592.

185. Qureshi SA, Yang X, Yates ML, Candelore MR, Feeney W, Hora D et al (2007)
Cloning and expression of canine glucagon receptor and its use to evaluate glucagon
receptor antagonists in vitro and in vivo. Eur J Pharmacol 555:8–16.

186. Mortensen OH, Dichmann DS, Abrahamsen N, Grunnet N, Nishimura E (2007)

Identification of a novel human glucagon receptor promoter: Regulation by cAMP
and PGC-1α. Gene 393:127–136.

187. Lau J, Behrens C, Sidelmann UG, Knudsen LB, Lundt B, Sams C et al (2007)
New β-alanine derivatives are orally available glucagon receptorantagonist. J Med
Chem 50:113–128.

188. Kim RM, Chang J, Lins AR, Brady Ed, Candelore MR, Dallas-Yang Q et al
(2008) Discovery of potent, orally active benzimidazole glucagon receptor
antagonists. Bioorg Med Chem Lett 18:3701–3705.

189. Kodra JT, Jørgensen AS, Andersen B, Behrens C, Brand CL, Christensen IT et al
(2008) Novel glucagon receptor antagonists with improved selectivity over the
glucose-dependent insulinotropic polypeptide receptor. J Med Chem 51:5387–5396.

190. Madsen P, Kodra JT, Behrens C, Nishimura E, Jeppesen CB, Pridal L et al

(2009) Human glucagon receptor antagonists with thiazole cores. a novel series with
superior pharmacokinetic properties. J Med Chem 52:2989–3000.

214
 REFERENCES

191. Li XC, Zhuo JL (2007) Targeting glucagon receptor signalling in treating

metabolic syndrome and renal injury in Type 2 diabetes: theory versus promise. Clin
Sci (Lond) 113:183–193.

192. Knappe TA, Linne U, Xie X, Marahiel MA (2010) The glucagon receptor
antagonist BI-32169 constitutes a new class of lasso peptides. FEBS Lett 584:785–
789.

193. Rafferty EP, Franklin Z, Irwin N, O'Harte FPM (2010) Stability and efficacy
studies in N-terminally truncated glucagon receptor antagonist peptides. Regulatory
Peptides 164:35–51.

194. Shen DM, Brady EJ, Candelore MR, Dallas-Yang Q, Ding VDH, Feeney WP et
al (2011) Discovery of novel, potent, selective, and orally active human glucagon
receptor antagonists containing a pyrazole core. Bioorg Med Chem Lett 21:76–81.

195. Sinz C, Chang J, Lins AR, Brady E, Candelore M, Dallas-Yang Q et al (2011)

Discovery of cyclic guanidines as potent, orally active, human glucagon receptor
antagonists. Bioorg Med Chem Lett 21:7131–7136.

196. Filipski KJ, Bian J, Ebner DC, Lee ECY, Li JC, Sammons MF et al (2012) A
novel series of glucagon receptor antagonists with reduced molecular weight and
lipophilicity. Bioorg Med Chem Lett 22:415–420.

197. Kieffer TJ, Cho YM, Merchant CE (2012) Targeting the glucagon receptor
family for diabetes and obesity therapy. Pharmacol Therap 135:247–278.

198. Xiong Y, Guo J, Candelore MR, Liang R, Miller C, Dallas-Yang Q et al (2012)

Discovery of a novel glucagon receptor antagonist N-[(4-{(1S)-1-[3-(3, 5-
Dichlorophenyl)-5-(6-methoxynaphthalen-2-yl)-1H-pyrazol-1-yl]ethyl}phenyl)
carbonyl]-β-alanine (MK-0893) for the treatment of type II diabetes. J Med Chem
55:6137−6148.

199. Chung JYL, Steinhuebel D, Krska SW, Hartner FW, Cai C, Rosen J et al (2012)
Asymmetric synthesis of a glucagon receptor antagonist via friedel crafts alkylation of
indole with chiral α-phenyl benzyl cation. Org Process Res Dev 16:1832−1845.

215
 REFERENCES

200. Guzman-Perez A, Pfefferkorn JA, Lee ECY, Stevens BD, Aspnes GE, Bian J et
al (2013) The design and synthesis of a potent glucagon receptor antagonist with
favorable physicochemical and pharmacokinetic properties as a candidate for the
treatment of type 2 diabetes mellitus. Bioorg Med Chem Lett 23:3051–3058.

201. Irwin N, Franklin ZJ, O'Harte FPM (2013) desHis1Glu9-glucagon-[mPEG] and

desHis1Glu9(Lys30PAL)-glucagon: Long-acting peptide-based PEGylated and
acylated glucagon receptor antagonists with potential anti diabetic activity. Eur J
Pharmacol 709:43–51.

202. O’Harte FPM, Franklin ZJ, Rafferty EP, Irwin N (2013) Characterisation of
structurally modified analogues of glucagon as potential glucagon receptor
antagonists. Mol Cell Endocrinol 381:26–34.

203. DeMong D, Dai X, Hwa J, Miller M, Lin SI, Kang L et al (2014) The discovery
of N-((2H-Tetrazol-5-yl)methyl)-4-((R)-1-((5R,8R)-8-(tertbutyl)-3-(3,5-
dichlorophenyl)-2-oxo-1,4-diazaspiro[4.5]dec-3-en-1-yl)-4,4-dimethylpentyl)
benzamide (SCH 900822): a potent and selective glucagon receptor antagonist. J Med
Chem 57:2601−2610.

204. Alberts P, Engblom L, Edling N, Forsgren M, Klingström G, Larsson C et al

(2002) Selective inhibition of 11β-hydroxysteroid dehydrogenase type 1 decreases
blood glucose concentrations in hyperglycaemic mice. Diabetologia 45:1528–1532.

205. Alberts P, Nilsson C, Selen G, Engblom LOM, Edling NHM, Norling S et al

(2003) Selective inhibition of 11β-hydroxysteroid dehydrogenase type 1 improves
hepatic insulin sensitivity in hyperglycemic mice strains. Endocrinology 144:4755–
4762.

206. Andrews RC, Rooyackers O, Walker BR (2003) Effects of the 11β-

Hydroxysteroid dehydrogenase inhibitor carbenoxolone on insulin sensitivity in men
with type 2 diabetes. J Clini Endocrinol Metab 88:285–291.

207. Tomlinson JW, Stewart PM (2005) 11β-Hydroxysteroid dehydrogenase type 1 as

a therapeutic target in the metabolic syndrome. Drug Discov Today: Therap Strat
2:93-96.

216
 REFERENCES

208. Gu X, Dragovic J, Koo GC, Koprak SL, LeGrand C, Mundt SS et al (2005)

Discovery of 4-heteroarylbicyclo[2.2.2]octyltriazoles as potent and selective
inhibitors of 11β-HSD1: Novel therapeutic agents for the treatment of metabolic
syndrome. Bioorg Med Chem Lett 15:5266–5269.

209. Olsona S, Astera SD, Brown K, Carbin L, Grahama DW, Hermanowski-

Vosatkab A et al (2005) Adamantyl triazoles as selective inhibitors of 11β-
hydroxysteroid dehydrogenase type 1. Bioorg Med Chem Lett 15:4359–4362.

210. Xianga J, Ipeka M, Surib V, Massefskia W, Pana N, Gea Y et al (2005)

Synthesis and biological evaluation of sulfonamidooxazoles and β-keto sulfones:
selective inhibitors of 11β-hydroxysteroid dehydrogenase type I. Bioorg Med Chem
Lett 15:2865–2869.

211. Schuster D, Maurer EM, Laggner C, Nashev LG, Wilckens T, Langer T et al

2006 The discovery of new 11β-hydroxysteroid dehydrogenase type 1 inhibitors by
common feature pharmacophore modeling and virtual screening. J Med Chem
49:3454-3466.

212. Sua X, Vickera N, Ganeshapillai D, Smitha A, Purohit A, Reed MJ, Potter BVL
(2006) Benzothiazole derivatives as novel inhibitors of human 11β-hydroxysteroid
dehydrogenase type 1. Mol Cell Endocrinol 248:214–217.

213. Richards S, Sorensen B, H-s Jae, Winn M, Chen Y, Wang J et al (2006)

Discovery of potent and selective inhibitors of 11b-HSD1 for the treatment of
metabolic syndrome. Bioorg Med Chem Lett 16:6241–6245.

214. Kurukulasuriya R, Fung S, Monzon K, Chiou W, Wang J, Stolarik D (2006)

Discovery of orally active butyrolactam 11β-HSD1 inhibitors. Bioorg Med Chem Lett
16:5555–5560.

215. Yeh VSC, Patel JR, Kurukulasuriya R, Fung S, Monzon K, Chiou W (2006)
Synthesis and biological evaluation of heterocycle containing adamantane 11β-HSD1
inhibitors. Bioorg Med Chem Lett 16:5414–5419.

217
 REFERENCES

216. Sutin L, Anderson S, Bergquist L, Castro VM, Danielson E, James S et al (2007)

Oxazolones as potent inhibitors of 11b-hydroxysteroid dehydrogenase type 1. Bioorg
Med Chem Lett 17:4837–4840.

217. Fujisawa Y, Nakagawa Y, Li R-S, Liu Y-J, Ohzeki T (2007) Diabetic pregnancy
in rats leads to impaired glucose metabolism in offspring involving tissue-specific
dysregulation of 11β-hydroxysteroid dehydrogenase type 1 expression. Life Sciences
81:724–731.

218. Vicker N, Sua X, Ganeshapillai D, Smith A, Purohit A, Reed MJ et al (2007)

Novel non-steroidal inhibitors of human 11β-hydroxysteroid dehydrogenase type 1. J
Steroid Biochem Mol Biol 104:123–129.

219. Sua X, Vickera N, Lawrencea H, Smith A, Purohit A, Reed MJ et al (2007)

Inhibition of human and rat 11β-hydroxysteroid dehydrogenase type 1 by 18β-
glycyrrhetinic acid derivatives. J Steroid Biochem Mol Biol 104:312–320.

220. Patel JR, Shuai Q, Dinges J, Winn M, Pliushchev M, Fung S et al (2007)

Discovery of adamantane ethers as inhibitors of 11β-HSD-1: Synthesis and biological
evaluation. Bioorg Med Chem Lett 17:750–755.

221. Webster SP, Ward P, Binnie M, Craigie E, McChonnell KMM, Sooy K et al

(2007) Discovery and biological evaluation of adamantly amide 11β-HSD1 inhibitors.
Bioorg Med Chem Lett 17:2838–2843.

222. Yuan C, Jean DJS, Liu Q, Cai L, Li A, Han N et al (2007) The discovery of 2-
anilinothiazolones as 11β-HSD1 inhibitors. Bioorg Med Chem Lett 17:6056–6061.

223. Flyrena K, Bergquitsa LO, Castroa VM, Fotsch C, Johansson L, Jean DJS Jr. et
al (2007) Piperidine amides as 11β-hydroxysteroid dehydrogenase type 1 inhibitors.
Bioorg Med Chem Lett 17:3421–3425.

224. Bhat G, Younis H, Herrera J, Palacio K, Paschual B, Hur G et al (2008)

Antisense inhibition of 11bhydroxysteroid dehydrogenase type 1 improves diabetes in
a novel cortisone-induced diabetic KK mouse model B. Biochem Biophys Res
Commun 365:740–745.

218
 REFERENCES

225. Uckaya G, Karadurmus N, Kutlu O, Corak A ¸ Kizildag S, Ural AU et al (2008)

Adipose tissue 11-beta-Hydroxysteroid Dehydrogenase Type 1 and Hexose-6-
Phosphate Dehydrogenase gene expressions are increased in patients with type 2
diabetes mellitus. Diab Res Clini Prac 82:135 – 140.

226. Iwasaki Y, Takayasu S, Nishiyama M, Tsugita M, Taguchi T, Asai M et al

(2008) Is the metabolic syndrome an intracellular Cushing state? Effects of multiple
humoral factors on the transcriptional activity of the hepatic glucocorticoid-activating
enzyme (11β-hydroxysteroid dehydrogenase type 1) gene. Mol Cell Endocrinol
285:10–18.

227. McCormick KL, Wang X, Mick GJ (2008) Modification of microsomal 11β-

HSD1 activity by cytosolic compounds: Glutathione and hexose phosphoesters. J
Steroid Biochem Mol Biol 111:18–23.

228. Sundbom M, Kaiser C, Bjorkstrand E, Castro VM, Larsson C et al (2008)

Inhibition of 11βHSD1 with the S-phenylethylaminothiazolone BVT116429 increases
adiponectin concentrations and improves glucose homeostasis in diabetic KKAy
mice. BMC Pharmacol 8:1-10.

229. Fotsch C, Bartberger MD, Bercot EA, Chen M, Cupples R, Emery M et al (2008)
Further studies with the 2-amino-1,3-thiazol-4(5h)-one class of 11β-hydroxysteroid
dehydrogenase type 1 inhibitors: reducing pregnane x receptor activity and exploring
activity in a monkey pharmacodynamic model. J Med Chem 51:7953–7967.

230. Xiang J, Wan Z-K, Li H-Q, Ipek M, Binnum E, Nunez J et al (2008) Piperazine
sulfonamides as potent, selective, and orally available 11β-hydroxysteroid
dehydrogenase type 1 inhibitors with efficacy in the rat cortisone-induced
hyperinsulinemia model. J Med Chem 51:4068–4071.

231. Lee JH, Kang NS and Yoo S-E (2008) Docking-based 3D-QSAR study for 11β-
HSD1 inhibitors. Bioorg Med Chem Lett 18:2479–2490.

232. Moreno-Doiaza H, Villalobos-Molina R, Ortiz-Andrade R, Diaz-Coutin D,

Medina-Franco JL, Webster SP et al (2008) Antidiabetic activity of N-(6-substituted-
1,3- benzothiazol-2-yl)benzenesulfonamides. Bioorg Med Chem Lett 18:2871–2877.

219
 REFERENCES

233. Sun D, Wang Z, Di Y, Jean JC, Labelle M, Ma J et al (2008) discovery and

initial sar of arylsulfonylpiperazine inhibitors of 11β-hydroxysteroid dehydrogenase
type 1 (11β-HSD1). Bioorg Med Chem Lett 18:3513–3516.

234. Zhua Y, Olsona SH, Hermanowski-Vosatkab A, Mundt S, Shah K, Springer M et

al (2008) 4-Methyl-5-phenyl triazoles as selective inhibitors of 11β-hydroxysteroid
dehydrogenase type I. Bioorg Med Chem Lett 18:3405–3411.

235. Astera SD, Grahama DW, Kharbandaa D, Patela G, Ponpipoma M, Santorellia

GM (2008) Bis-aryl triazoles as selective inhibitors of 11β-hydroxysteroid
dehydrogenase type 1. Bioorg Med Chem Lett 18:2799–2804.

236. Yanga H, Doub W, Loub J, Leng Y, Shen J (2008) Discovery of novel inhibitors
of 11β-hydroxysteroid dehydrogenase type 1 by docking and pharmacophore
modeling. Bioorg Med Chem Lett 18:1340–1345.

237. Zhua Y, Olsona SH, Grahama D, Patela G, Hermanowski-Vosatkab A, Mundt S

et al (2008) Phenylcyclobutyl triazoles as selective inhibitors of 11β-hydroxysteroid
dehydrogenase type I. Bioorg Med Chem Lett 18:3412–3416.

238. Wanga H, Ruan Z, Lia JJ, Simpkins LM, Smirka RA, Wua SC et al (2008)
Pyridine amides as potent and selective inhibitors of 11β-hydroxysteroid
dehydrogenase type 1. Bioorganic & Medicinal Chemistry Letters 18:11, 3168–3172.

239. Gumy C, Thurnbichler C, Aubry EM, Balazs Z, Pfisterer P, Baumgartner L et al

(2009) Inhibition of 11β-hydroxysteroid dehydrogenase type 1 by plant extracts used
as traditional antidiabetic medicines. Fitoterapia 80:200–205.

240. Miyamoto Y, Morisaki H, Yamanaka I, Kokubo Y, Masuzaki H, Okayama A et

al (2009) Association study of 11b-hydroxysteroid dehydrogenase type 1 gene
polymorphisms and metabolic syndrome in urban Japanese cohort. Diab Res Clini
Prac 85:132 – 138.

241. Sua X, Vicker N, Trusselle M, Halemb H, Cullerb MD, Potter BVL (2009)
Discovery of novel inhibitors of human 11β-hydroxysteroid dehydrogenase type 1.
Mol Cell Endocrinol 301:169–173.

220
 REFERENCES

242. Classen-Houbena D, Schuster D, Cunhad TD, Odermattd A, Wolberb G, Jordis

U et al (2009) Selective inhibition of 11β-hydroxysteroid dehydrogenase 1 by 18-
glycyrrhetinic acid but not 18-glycyrrhetinic acid. J Steroid Biochem Mol Biol
113:248–252.

243. Srivastava RAK (2009) Fenofibrate ameliorates diabetic and dyslipidemic

profiles in KKAy mice partly via down-regulation of 11β-HSD1, PEPCK and
DGAT2. Comparison of PPARα, PPARγ, and liver x receptor agonists. Eur J
Pharmacol 607:258–263.

244. Stewart PM and Tomlinson JW (2009) Selective Inhibitors of 11β-

Hydroxysteroid Dehydrogenase Type 1 for Patients With Metabolic Syndrome Is the
Target Liver, Fat, or Both? Diabetes 58:1-2.

245. Morgan SA, Sherlock M, Gathercole LL, Lavery GG, Lenaghan C, Bujalska IJ et
al (2009) 11β-hydroxysteroid dehydrogenase type 1 regulates glucocorticoid-induced
insulin resistance in skeletal muscle. Diabetes 58:2506–2515.

246. Dzyakanchuka AA, Balazsa Z, Nasheva LG, Amreinb KE, Odermatt A (2009)
11β-Hydroxysteroid dehydrogenase 1 reductase activity is dependent on a high ratio
of NADPH/NADP+ and is stimulated by extracellular glucose. Mol Cell Endocrinol
301:137–141.

247. Lloyd JD, Helmering J, Cordover D, Bowsman M, Chen M, Hale C et al (2009)

Antidiabetic effects of 11β-HSD1 inhibition in a mouse model of combined diabetes,
dyslipidaemia and atherosclerosis. Diab Obes Metab 11:688–699.

248. Siu M, Johnsona TO, Wanga Y, Naira SK, Taylora WD, Crippsa SJ et al (2009)
N-(Pyridin-2-yl) arylsulfonamide inhibitors of 11β-hydroxysteroid dehydrogenase
type 1: Discovery of PF-915275. Bioorg Med Chem Lett 19:3493–3497.

249. Sun D, Wang Z, Cardozo M, Choi R, DeGraffenreid M, Di Y et al (2009)

Synthesis and optimization of arylsulfonylpiperazines as a novel class of inhibitors of
11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1). Bioorg Med Chem Lett
19:1522–1527.

221
 REFERENCES

250. McMinn DL, Rew Y, Sudom A, Caille S, DeGraffenreid M, He X et al (2009)

Optimization of novel di-substituted cyclohexylbenzamide derivatives as potent 11β-
HSD1 inhibitors. Bioorg Med Chem Lett 19:1446–1450.

251. Rew Y, McMinn DL, Wang Z, He X, Hungate RW, Jiang B et al (2009)

Discovery and optimization of piperidyl benzamide derivatives as a novel class of
11β-HSD1 inhibitors. Bioorg Med Chem Lett 19:1797–1801.

252. Neelamkavil SF, Boyle CD, Chackalamannil S, Greenlee WJ, Zhang L,

Terracina G (2009) The discovery of azepane sulfonamides as potent 11β-HSD1
inhibitors. Bioorg Med Chem Lett 19:4563–4565.

253. Zhang X, Zhou Z, Yang H, Chen J, Feng Y, Du L et al (2009) 4-

(Phenylsulfonamidomethyl)benzamides as potent and selective inhibitors of the 11β-
hydroxysteroid dehydrogenase type 1 with efficacy in diabetic ob/ob mice. Bioorg
Med Chem Lett 19:4455–4458.

254. Rochea D, Carniatoa D, Lerichea C, Lepifrea F, Christmann-Francka S,

Graedlerb U et al (2009) Discovery and structure–activity relationships of
pentanedioic acid diamides as potent inhibitors of 11β-hydroxysteroid dehydrogenase
type I. Bioorg Med Chem Lett 19:2674–2678.

255. Lepifre F, Christmann-Francka S, Roche D, Leriche C, Carniato D, Charon C et

al (2009) Discovery and structure-guided drug design of inhibitors of 11β-
hydroxysteroid-dehydrogenase type I based on a spiro-carboxamide scaffold. Bioorg
Med Chem Lett 19:3682–3685.

256. Karlsson C, Jerna M, Olsson B,. Lystig T , Gummesson A, Storlien L et al

(2010) Differences in associations between 11β HSD1 gene expression and metabolic
parameters in subjects with and without impaired glucose homeostasis. Diab Res Clini
Prac 88:252 -258.

257. Odermatt A, Nashev LG (2010) The glucocorticoid-activating enzyme 11β-

hydroxysteroid dehydrogenase type 1 has broad substrate specificity: Physiological
and toxicological considerations. J Steroid Biochem Mol Biol 119:1–13.258. Stewart

222
 REFERENCES

PM, Gathercole LL (2010) Targeting the pre-receptor metabolism of cortisol as a

novel therapy in obesity and diabetes. J Steroid Biochem Mol Biol 122:21–27.

259. Huber R, Rosenstock J, Banarer S, Fonseca VA, Inzucchi SE, Sun W et al (2010)
The 11-beta-hydroxysteroid dehydrogenase type 1 inhibitor INCB13739 improves
hyperglycemia in patients with type 2 diabetes inadequately controlled by metformin
monotherapy. Diab Care 33:1516–1522.

260. Edgerton DS, Basu R, Ramnanan CJ, Farmer TD, Neal D, Scott M et al (2010)
Effect of 11β-hydroxysteroid dehydrogenase-1 inhibition on hepatic glucose
metabolism in the conscious dog. Am J Physiol Endocrinol Metab 298:1019-1026.

261. Wang M, Veniant MM, Hale C, Hungate RW, Gahm K, Emery MG et al (2010)
Discovery of a potent, orally active 11β-hydroxysteroid dehydrogenase type 1
inhibitor for clinical study: identification of (S)-2-((1S,2S,4R)-Bicyclo[2.2.1]heptan-
2-ylamino)-5-isopropyl-5-methylthiazol-4(5H)-one (AMG 221). J Med Chem 53:
4481–4487.

262. Ge R, Huang Y, Liang G, Li X (2010) 11β-hydroxysteroid dehydrogenase type 1

inhibitors as promising therapeutic drugs for diabetes: status and development. Curr
Med Chem 17:412-422.

263. Yan X, Wang Z, Sudom A, Cardozo M, DeGraffenreid M, Di Y et al (2010) The

synthesis and SAR of novel diarylsulfone 11β-HSD1 inhibitors Bioorg Med Chem
Lett 20:7071–7075.

264. Shah U, Boyle CD, Chackalamannil S, Baker H, Kowalski T, Lee J et al (2010)

Azabicyclic sulfonamides as potent 11β-HSD1 inhibitors. Bioorg Med Chem Lett
20:1551–1554.

265. Cheng H, Hoffman J, Le P, Nair S, Cripps S, Matthews J et al (2010) The

development and SAR of pyrrolidine carboxamide 11β-HSD1 inhibitors. Bioorg Med
Chem Lett 20:2897–2902.

266. Tice CM, Zhao W, Krosky PM, Kruk BA, Berbaum J, Johnson JA et al (2010)
Discovery and optimization of adamantyl carbamate inhibitors of 11β-HSD1. Bioorg
Med Chem Lett 20:6725–6729.
223
 REFERENCES

267. Ahn JH, Kim SH, Ramu R, Kwon SW, Lee SH, Kim CH et al (2010) Discovery
of cyclicsulfonamide derivatives as 11β-hydroxysteroid dehydrogenase 1 inhibitors.
Bioorg Med Chem Lett 20:1065–1069.

268. Tice CM, Zhao W, Xu Z, Cacatian ST, Simpson RD, Ye YJ et al (2010)

Spirocyclic ureas: Orally bioavailable 11β-HSD1 inhibitors identified by computer-
aided drug design. Bioorg Med Chem Lett 20:881–886.

269. Ye XY, Chen SY, Nayeem A, Golla R, Seethala R, Wang M et al (2011) Design,
synthesis, and SAR studies of novel polycyclic acids as potent and selective inhibitors
of human 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD-1). Bioorg Med Chem
Lett 21:6699–6704.

270. Liu Y, Fana Z, Dub H, Zhanga M, Menga Z, Chena L (2011) Direct regulation of
glucose and not insulin on hepatic hexose-6-phosphate dehydrogenase and 11β-
hydroxysteroid dehydrogenase type 1. Mol Cell Endocrinol 333:62–69.

271. Simunkova K, Hampl R, Hill M, Kriz L, Vrbikova J, Kvasnickova H et al (2011)

Evaluation of hepatic 11 β-hydroxysteroid dehydrogenase activity by cortisone
acetate test in young adults with diabetes mellitus type 1. Physiol Res 60:263-270.

272. Tice CM, Xu Z, Zhao W, Cacatian S, Ye YJ, Singh SB et al (2011) Structure-

based design and synthesis of 1,3-oxazinan-2-one inhibitors of 11β-hydroxysteroid
dehydrogenase type 1. J Med Chem 54:6050–6062.

273. Kim KY, Park JS, Rhee SD, Kang NS, Jung WH, Kim HY et al (2011) Anti-
diabetic and anti-adipogenic effects of a novel selective 11β-hydroxysteroid
dehydrogenase type 1 inhibitor, 2-(3-benzoyl)-4-hydroxy-1,1-dioxo-2H-1,2-
benzothiazine-2-yl-1-phenylethanone (KR-66344). Biochem Pharmacol 81:1028-35.

274. Sun D, Wang Z, Caille S, DeGraffenreid M, de Turiso FGL, Hungate R et al

(2011) Synthesis and optimization of novel 4,4-disubstituted cyclohexylbenzamide
derivatives as potent 11β-HSD1 inhibitors. Bioorg Med Chem Lett 21:405–410.

275. Shen J, Xia G, Xue M, Liu L, Yu J, Liu H et al (2011) Potent and novel 11β-
HSD1 inhibitors identified from shape and docking based virtual screening. Bioorg
Med Chem Lett 21:5739–5744.
224
 REFERENCES

276. Wua SC, Yoon D, Chin J, Kirk KV, Seethala R, Golla R et al (2011) Discovery
of 3-hydroxy-4-cyano-isoquinolines as novel, potent, and selective inhibitors of
human 11β-hydroxydehydrogenase 1 (11β-HSD1). Bioorg Med Chem Lett 21:6693–
6698.

277. Maletic M, Leeman A, Szymonifka M, Mundt SS, Zokian HH, Shah K et al

(2011) Bicyclo[2.2.2]octyltriazole inhibitors of 11β-hydoxysteroid dehydrogenase
type 1. Pharmacological agents for the treatment of metabolic syndrome. Bioorg Med
Chem Lett 21:2568–2572.

278. Venier O, Pascal C, Braun A, Namane C, Mougenot P, Crespin O et al (2011)

Pyrrolidine-pyrazole ureas as potent and selective inhibitors of 11β-hydroxysteroid-
dehydrogenase type 1. Bioorg Med Chem Lett 21:2244–2251.

279. Sun W, Maletica M, Mundt SS, Shah K, Zokian H, Lyons K et al (2011)

Substituted phenyl triazoles as selective inhibitors of 11β-Hydroxysteroid
Dehydrogenase Type 1. Bioorg Med Chem Lett 21:2141–2145.

280. Ahn JH, Kim KY, Kwon SW, Kang SK, Lee JH, Bok JH et al (2011) Synthesis
and 11β hydroxysteroid dehydrogenase 1 inhibition of thiazolidine derivatives with an
adamantyl group. Bioorg Med Chem Lett 21:435–439.

281. Valeur E, Christmann-Franck S, Lepifre F, Carniato D, Cravo D, Charon C et al

(2011) Structure-based design of 7-azaindole-pyrrolidine amides as inhibitors of 11β-
hydroxysteroid dehydrogenase type I. Bioorg Med Chem Lett 22:5909–5914.

282. Xiong B, Xia G, Liu L, Xue M, Liu H, Yu J et al (2012) Discovery of novel

sulfonamides as potent and selective inhibitors against human and mouse 11β-
hydroxysteroid dehydrogenase type 1. Mol Cell Endocrinol 358:46–52.

283. Maser E, Blum A, Loerz C, Martin HJ, Staab-Weijnitz CA (2012) Momordica

charantia extract, a herbal remedy for type 2 diabetes, contains a specific 11β-
hydroxysteroid dehydrogenase type 1 inhibitor. J Steroid Biochem Mol Biol 128:51–
55.

284. Dhanesha N, Joharapurkar A, Shah G, Kshirsagar S, Dhote V, Sharma A et al

(2012) Inhibition of 11β-hydroxysteroid dehydrogenase 1 by carbenoxolone affects
225
 REFERENCES

glucose homeostasis and obesity in db⁄db mice. Clini Exp Pharmacol Physiol 39:69–
77.

285. Scott JS, deSchoolmeester J, Kilgour E, Mayers RM, Packer MJ, Hargreaves D
et al (2012) Novel acidic 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1)
inhibitor with reduced acyl glucuronide liability: the discovery of 4-[4-(2-
Adamantylcarbamoyl)-5-tert-butyl-pyrazol-1-yl]benzoic Acid (AZD8329). J Med
Chem 55:10136−10147.

286. Goldberg FW, Leach AG, Scott JS, Snelson W, Groombridge SD, Donald CS et
al (2012) Free-Wilson and structural approaches to co-optimizing human and rodent
isoform potency for 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1)
inhibitors. J Med Chem 55:10652−10661.

287. Shen J, Yu J, Liu H, Xia G, Liu L, Xu Z et al (2012) Discovery of 2-Alkyl-1-

arylsulfonylprolinamides as 11β- Hydroxysteroid Dehydrogenase Type 1 Inhibitors.
ACS Med Chem Lett 3:793−798.

288. Wan ZK, Chenail E, Li HQ, Ipek M, Xiang J, Suri V et al (2012) Discovery of
HSD-621 as a potential agent for the treatment of type 2 diabetes. ACS Med Chem
Lett 4:118−123.

289. Kim KY, Parka JS, Rheea SD, Junga WH, Kang NS, Kim HY et al (2012) Anti-
diabetic and anti-adipogenic effects of a novel selective 11β-hydroxysteroid
dehydrogenase type 1 inhibitor in the diet-induced obese mice. Eur J Pharmacol
691:19–27.

290. Xiao X, Ma R, Liu J, Wu l, Sun J, Yang Z et al (2012) Differential expression of

placental 11β-hydroxysteroid dehydrogenases in pregnant women with diet-treated
gestational diabetes mellitus. Steroids 77:798–805.

291. Rew Y, DeGraffenreid M, He X, Jaen JC, McMinn DL, Sun D et al (2012)

Discovery and optimization of benzenesulfonanilide derivatives as a novel class of
11β-HSD1 inhibitors. Bioorg Med Chem Lett 22:3786–3790.

226
 REFERENCES

292. Scott JS, Gill AL, Godfrey L, Groombridge SD, Rees A, Revill J et al (2012)
Optimisation of pharmacokinetic properties in a neutral series of 11β-HSD1
inhibitors. Bioorg Med Chem Lett 22:6756–6761.

293. Christmann-Franck S, Valeur E, Lepifre F, Carniato D, Cravo D, Charon C et al

(2012) Structure-based design of 7-azaindole-pyrrolidine amides as inhibitors of 11β-
hydroxysteroid dehydrogenase type I. Bioorg Med Chem Lett 22:5909–5914.

294. Shen J, Zhang L, Chen J, Ning M, Zou Q, Leng Y (2012) Synthesis and
evaluation of piperidine urea derivatives as efficacious 11β-hydroxysteroid
dehydrogenase type 1 inhibitors in diabetic ob/ob mice. Bioorg Med Chem Lett
22:2748–2752.

295. Ritter K, Bohme T, Engel CK, Farjot G, Gussregen S, Haack T et al (2013) 1,1-
Dioxo-5,6-dihydro-[4,1,2]oxathiazines, a novel class of 11ß-HSD1 inhibitors for the
treatment of diabetes. Bioorg Med Chem Lett 23:4685–4691.

296. Anagnostis P, Katsiki N, Adamidoua F, Athyrosc VG, Karagiannis A, Kita M et

al (2013) 11beta-Hydroxysteroid dehydrogenase type 1 inhibitors: novel agents for
the treatment of metabolic syndrome and obesity-related disorders? Metab Clini Exp
62: 21-33.

297. Kim KY, Park SB, Jung WH, Kang NS, Park JS, Bae GH et al (2013) Anti-
diabetic and anti-inflammatory effect of a novel selective 11β-HSD1 inhibitor in the
diet-induced obese mice. Eur J Pharmacol 721:70–79.

298. Harno E, Cottrell EC, Keevil BG, DeSchoolmeester JD, Andersen H, Turnbull
AV et al (2013) 11-Dehydrocorticosterone causes metabolic syndrome, which is
prevented when 11β-HSD1 is knocked out in livers of male mice. Endocrinol
154:3599-3609.

299. Nair SK, Matthews JJ, Cripps SJ, Cheng H, Hoffman JE, Smith C et al (2013) N-
(Pyridin-2-yl) arylsulfonamide inhibitors of 11b-hydroxysteroid dehydrogenase type
1: Strategies to eliminate reactive metabolites. Bioorg Med Chem Lett 23:2344–2348.

300. Navarrete-Vazquez G, Alaniz-Palacios A, Hidalgo-Figueroa S, González-

Acevedo C, Avila-Villarreal G, Estrada-Soto S et al (2013) Discovery, synthesis and
227
 REFERENCES

in combo studies of a tetrazole analogue of clofibric acid as a potent hypoglycemic

agent. Bioorg Med Chem Lett 23:3244–3247.

301. Venier O, Pascal C, Braun A, Namane C, Mougenot P, Crespin O et al (2013)

Discovery of SAR184841, a potent and long-lasting inhibitor of 11β-hydroxysteroid
dehydrogenase type 1, active in a physiopathological animal model of T2DM. Bioorg
Med Chem Lett 23:2414–2421

302. Ge RS, Lin H, Hu GX, Guo J, Ge Y, Liang G et al (2013) Mono-carbonyl

curcumin analogues as 11β-hydroxysteroid dehydrogenase 1 inhibitors. Bioorg Med
Chem Lett 23:4362–4366.

303. Odermatt A, Hofer S, Kratschmar DV, Schernthannera B, Vuorinena A,

Schustera D et al (2013) Synthesis and biological analysis of benzazol-2-yl piperazine
sulfonamides as 11β-hydroxysteroid dehydrogenase 1 inhibitors. Bioorg Med Chem
Lett 23:5397–5400.

304. Robl JA, Ye XY, Yoon DY, Chen SY, Nayeem A, Golla R et al (2014) Synthesis
and structure–activity relationship of 2-adamantylmethyl tetrazoles as potent and
selective inhibitors of human 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1).
Bioorg Med Chem Lett 24: 654–660.

305. Fardella CE, Lagos CF, Vecchiola A, Allende F, Fuentes CA, Tichauer JE et al
(2014) Identification of novel 11β-HSD1 inhibitors by combined ligand- and
structure-based virtual screening. Mol Cell Endocrinol 384:71–82.

306. Huang Y, Yuan X, Li H, Bai H, Su Z, Xiang Q et al (2014) Synthesis of novel

curcumin analogues for inhibition of 11β-hydroxysteroid dehydrogenase type 1 with
anti-diabetic properties. Eur J Med Chem 77:223-230.

307. Akiyama N, Akiyama Y, Kato H, Kuroda T, Ono T, Imagawa K et al (2014)

Pharmacological evaluation of adipose dysfunction via 11β-hydroxysteroid
dehydrogenase type 1 in the development of diabetes in diet-induced obese mice with
cortisone pellet implantation. J Pharmacol Exp Ther 349:66-74.

308. Navarrete-Vazqueza G, Morales-Vilchisa MG, Estrada-Sotoa S, Ramírez-

Espinosaa JJ, Hidalgo-Figueroaa S, Nava-Zuazoa C et al (2014) Synthesis of 2-{2-
228
 REFERENCES

[(α/β-naphthalen-1-ylsulfonyl)amino]-1,3-thiazol-4-yl} acetamides with 11β-

hydroxysteroid dehydrogenase inhibition and in combo antidiabetic activities. Eur J
Med Chem 74:179–186.

309. Lee Y, Shin YJ, Ahn SK (2014) 3-Amino-N-adamantyl-3-methylbutanamide

derivatives as 11β-hydroxysteroid dehydrogenase 1 inhibitor. Bioorg Med Chem Lett
24:1421–1425.

310. Rousseau GG, Pierreux CE, Stafford J, Demonte D, Scott DK, Vandenhaute J et
al (1999) Antiglucocorticoid activity of hepatocyte nuclear factor-6. Proc Natl Acad
Sci U S A 96:8961–8966.

311. Gasteiger J, Rudolph C, Sadowski J (1990) Automatic generation of 3D atomic

coordinates for organic molecules. Tetrahedron Comp Method 3:537-547.

312. Muppalaneni NB, Rao AA (2012) Virtual screening of plant derived compounds
for aldose reductase inhibition using molecular docking. Bioinformation 8:980-983.

313. Serra JR, Jurs PC, Kaiser KL (2011) Linear regression and computational neural
network prediction of tetrahymena acute toxicity for aromatic compounds from
molecular structure. Chem Res Toxicol 14:1535-1545.

314. Paliwal S, Seth D, Yadav D, Yadav R, Paliwal, S (2011) Development of a

robust QSAR model to predict the affinity of pyrrolidine analogs for dipeptidyl
peptidase IV (DPP- IV). J Enz Inhib Med Chem 26:129-140.

315. Riahi S, Pourbasheer E, Ganjali MR, Ali PN, Moghaddam Z (2008) QSPR study
of the distribution coefficient property for hydantoin and 5- arylidene derivatives. A
genetic algorithm application for the variable selection in the MLR and PLS methods.
J Chinese Chem Soc 55:1086-1093.

316. Gupta L, Patel A, Karthikeyan C, Trivedi P (2010) QSAR studies on

dihydroalkoxy- benzyl-oxopyrimidines (DABOs) derivatives, a new series of potent,
broad-spectrum non-nucleoside reverse transcriptase inhibitors. J Curr Pharmaceutic
Res 1:19-25.

229
 REFERENCES

317. Paliwal S, Singh S, Pal M (2012) In silico ligand based design of

indolylpiperidinyl derivatives as novel histamine H1 receptor antagonists. Drug
Discov Therap 6:69-77

318. Kruzlicova D, Danihelova M, Veverka M (2012) Quantitative structure

antioxidant activity relationship of quercetin and its new synthesized derivatives.
Nova Biotech Chimic 11:37-44.

319. Myint KZ, Xie XA (2010) Recent advances in fragment -based QSAR and multi-
dimensional QSAR methods. Int J Mol Sci 11:3846-3866.

320. Wu B, Rega MF, Wei J, Yuan H, Dahl R, Zhang Z et al (2009) Discovery and
binding studies on a series of novel pin1 ligands. Chem Biol Drug Design 73:369–
379.

321. Varmuza K, Filzmoser P, Dehmer M (2013) Multivariate linear QSPR/QSAR

models: rigorous evaluation of variable selection for PLS. Comput Struc Biotech J
5:1-10.

322. Karelson M, Lobanov VS (1996) Quantum-chemical descriptors in QSAR/QSPR

studies. Chem Rev 96:1027-1044.

323. Linusson A, Elofsson M, Andersson IE, Dahlgren MK (2010) Statistical

molecular design of balanced compound libraries for QSAR modeling. Curr Med
Chem 17:2001-2016.

324. Mitra I, Saha A, Roy K (2013) Predictive modeling of antioxidant coumarin

derivatives using multiple approaches: descriptor-based QSAR, 3D pharmacophore
mapping, and HQSAR. Sci Pharm 81:57–80.

325. Yu H, Jin H, Sun L, Zhang L, Sun G, Wang Z, Yu Y (2013) Toll-like receptor 7

agonists: chemical feature based pharmacophore identification and molecular docking
studies. PLoS ONE 8:e56514.

326. Rastelli G, Pacchioni S, Sirawaraporn W, Sirawaraporn R, Parenti MD, Ferrari

AM (2003) Docking and database screening reveal new classes of plasmodium
falciparum dihydrofolate reductase inhibitors. J Med Chem 46:2834-2845.

230
 REFERENCES

327. Gupta RK, Kumar D, Chaudhary AK, Maithani M, Singh R (2012) Antidiabetic
activity of Passiflora incarnata Linn. in streptozotocin-induced diabetes in mice. J
Ethnopharmacol 139: 801-806.

328. Eleazu CO, Eleazu KC, Chukwuma S, Essien UN (2013) Review of the
mechanism of cell death resulting from streptozotocin challenge in experimental
animals, its practical use and potential risk to humans. J Diab Metabol Disord 12:1-7.

231
 ANEXURE-I

Research Paper Communicated/Published

S. No. Title Name of Journal Status

Pharmacophore modeling,
virtual screening, molecular
Tetrahedron
1. docking and biological Under review
Letters
evaluation to identify novel
anti-diabetic lead compounds

Simultaneous Modeling of 4-
(aminomethyl)-N-
methylbenzamide based
Glucagon Receptor Indian Journal of
2. Under review
Antagonist (GRA): Chemistry-B
Application of Equation and
Neural Network Based
QSAR

Quantitative Structure
Activity Relationships
Medicinal
(QSAR) of N6 substituted
3. Chemistry Published
adenosine receptor agonists
Research
as potential antihypertensive
agents

Pharmacophore Based In-

Silico High Throughput Medicinal
4. Screening to Identify Novel Chemistry Published
Topoisomerase-I Inhibitors Research
against Renal Cancer

Quantitative Structure
Activity Relationship Studies
Journal of
5. of Topoisomerase I Published
Chemistry
Inhibitors as Potent Anti-
Breast Cancer Agents