Computers in Biology and Medicine 45 (2014) 20–25

Contents lists available at ScienceDirect

Computers in Biology and Medicine

journal homepage: www.elsevier.com/locate/cbm

Comprehension of drug toxicity: Software and databases

Andrey A. Toropov a,n, Alla P. Toropova a, Ivan Raska Jrb, Danuta Leszczynska c,
Jerzy Leszczynski d
a
IRCCS, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Via La Masa 19, Milano 20156, Italy
b
3rd Department of Medicine, Department of Endocrinology and Metabolism, First Faculty of Medicine, Charles University in Prague and General University
Hospital in Prague, U Nemocnice 1, 12808 Prague 2, Czech Republic
c
Interdisciplinary Nanotoxicity Center, Department of Civil and Environmental Engineering, Jackson State University, 1325 Lynch St, Jackson,
MS 39217-0510, USA
d
Interdisciplinary Nanotoxicity Center, Department of Chemistry and Biochemistry, Jackson State University, 1400 J. R. Lynch Street, PO Box 17910,
Jackson, MS 39217, USA

art ic l e i nf o a b s t r a c t

Article history: Quantitative structure–property/activity relationships (QSPRs/QSARs) are a tool (in silico) to rapidly
Received 30 August 2013 predict various endpoints in general, and drug toxicity in particular. However, this dynamic evolution of
Accepted 18 November 2013 experimental data (expansion of existing experimental data on drugs toxicity) leads to the problem of
critical estimation of the data. The carcinogenicity, mutagenicity, liver effects and cardiac toxicity should
Keywords: be evaluated as the most important aspects of the drug toxicity. The toxicity is a multidimensional
In silico toxicology phenomenon. It is apparent that the main reasons for the increase in applications of in silico prediction of
In silico methods toxicity include the following: (i) the need to reduce animal testing; (ii) computational models provide
QSAR reliable toxicity prediction; (iii) development of legislation that is related to use of new substances; (iv)
Computational toxicology
filling data gaps; (v) reduction of cost and time; (vi) designing of new compounds; (vii) advancement of
Drug toxicity
understanding of biology and chemistry. This mini-review provides analysis of existing databases and
software which are necessary for use of robust computational assessments and robust prediction of
potential drug toxicities by means of in silico methods.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction models. These criteria are briefly described by Organization for

Economic Co-operation and Development (OECD) as Setubal
There are various tools that one could use for prediction of principles: QSARs for regulatory application should: (1) be asso-
properties (activities) of chemical compounds. Among them the ciated with a defined endpoint of regulatory importance, (2) take
quantitative structure–property/activity relationships (QSPRs/ the form of an unambiguous algorithm, (3) have a defined domain
QSARs) methods hold important place. The history of evolution of applicability, (4) be associated with appropriate measures of
of the QSPRs/QSARs techniques contains three basic periods. The goodness of fit, robustness, and predictivity, and (5) have a
first period involved design of molecular descriptors which are mechanistic basis [1].
correlated with important physicochemical parameters and/or The drug discovery establishment has been probably one of the
with various biological endpoints. The statistical quality of QSPR/ original industries to appreciate the QSPR/QSAR technology and
QSAR for all available compounds which were used to build up the still remains its the most important user. In fact, the drug
model was considered as the main result. The second period, the discovery protocol needs to define two groups of endpoints related
statistical quality of model for external “invisible” compounds to new molecular entities (NMEs) [1] which relate to both the
which were not involved in building up model, become the main therapeutic and toxic effects [2].
criterion of quality of QSPR/QSAR. On the other hand, the third Physico-chemical indicators have been increasingly used dur-
period, the definition of chemical space where QSPR/QSAR can be ing the early stages of drug discovery to provide a more compre-
used with satisfactory accuracy, i.e. so-called the domain of hensive understanding of the key properties that affect the
applicability has become the measure of quality for QSPR/QSAR biological functions (i.e. ADME—absorption, distribution, metabo-
lism, and excretion). The most commonly measured physic-
chemical properties are permeability and solubility (due to their
n
Corresponding author. Tel.: þ 39 2 39014595; fax: þ39 2 39014735.
importance in the gastrointestinal absorption of orally adminis-
E-mail addresses: [email protected], tered drugs), and also lipophilicity, integrity, and stability (since
[email protected] (A.A. Toropov). these properties generally define the pharmaceutical potential of

0010-4825/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compbiomed.2013.11.013
 A.A. Toropov et al. / Computers in Biology and Medicine 45 (2014) 20–25 21

a compound) [3–8]. Often, the ADME concept is expanded by Table 1

toxicity [5,6]. Apparently, this potential hazardous quality of NME Classification scheme for substances involved in the drug discovery [34].
should be taken into account with no less care than possible
Class Comments
therapeutic effects.
There are several endpoints that relate to potential hazardous 1 Known to be both mutagenic and carcinogenic
effects(such as the liver effects of drug candidates [9]; cardiac toxicity 2 Known to be mutagenic but unknown carcinogenic potential
[10], and blood–brain barrier of compounds that has influence upon 3 Structurally alerting compound, unrelated to the active pharmaceutical
ingredient (API) and of unknown mutagenic potential
both drug efficiency and drug toxicity [11–13]) which should be 4 Structurally alerting compound related to the API
estimated during the early stages of drug discovery. Finally carcino- 5 No structural alerts or sufficient evidence for absence of mutagenicity
genicity [14] and mutagenicity [15] of a drug also should be estimated
during the initial stages of drug discovery.
Due to huge cost and time necessary for research and develop-
ment related to drug design, many in silico methods have been (mg/kg, mg/m3, ppm, etc.). Thus, the search for data on different
developed to provide accurate prediction of pharmacokinetic prop- kinds of toxicity for various substances is complex enough task.
erties, such as Absorption, Distribution, Metabolism, Excretion, and However, there are a plethora of private and public data
Toxicity (ADMET), in very early stage of drug discovery [3–8,14,15]. resources available for developing toxicity models. Recent reviews
Recently [16,17], the physiologically based pharmacokinetic/phar- summarized available public toxicity databases [35,36] (Table 2).
macodynamic (PBPK/PD) approaches (for quantitatively description of On the other hand, the large number and variety of sources of
the metabolism) have been suggested. In fact, PBPK models represent drug toxicity (their number is increasing day by day) lead to two
synthesis of mathematical calculations and experimental measure- questions: (i) where one can get information on toxicity of the
ments [17]. Nevertheless, this approach becomes an necessary tool for potential therapeutic agent; and (ii) how one can estimate
the drug discovery [18] because allows to take into account chemical accuracy and reliability of the data. The problem of adequate and
structures of potential metabolites [19]. fast estimation of drug toxicity lead to creation of international
Various approaches could be used to analyze possible toxicity of organizations such as Food and Drug Administration (FDA), Center
drugs. Apparently, classes of substances which are potential NMEs of Drug Evaluation and Research (CDER), and the Registration,
define the methodological details of the drug toxicity analysis. It Evaluation, Authorization and Restriction of Chemicals (REACH)
appears that organic compounds are the most known source of the legislation of the EU [32,35,36]. In spite of a number of organiza-
NMEs. However, this group of NME contains many sub-classes tions involved, harmonizing the reporting of chemical toxicity data
[2,5,7]. Peptides [20,21] are also source of NMEs, and this case often to facilitate comparison between available data sources and
involves application of “unclassic” QSAR approaches. Relatively new databases remains a critical need [37].
class of potential NMEs consist of various nano materials such as It should be noted that these international organizations
fullerene derivatives [22–25] nanoparticles [26,27], and others [28]. encourage the use of in silico methods in order to answer two
Finally, one also needs to consider an important challenge of the questions mentioned above, as well as for the harmonizing of
drug discovery—so-called drug–drug interactions [29–31]. available data on the toxicity in general, and drug toxicity in
The average success rate for NMEs in all therapeutic areas particular [37].
combined, starting from first-in-human studies to registration One can obtain practical skills of estimation of the data related
during 1991–2000 period was approximately 11%. In 2003, the to toxicity after visiting web sites listed in Table 2.
US Food and Drug Administration (FDA) approved 21 NMEs,
however, during next years, this number has decreased (only 15
NMEs approved in 2010) [32]. Although lack of efficacy is a major 3. In silico toxicology tools
contributor to a disappointment, toxicity can also be a cause of
failure in drug development [1]. In silico toxicology is generally, but not exclusively, a predictive
Thus, the systematization of the information that is necessary science. The approaches used for helping to define safety and
for drug discovery is possible only with involving of multimillion discovery efforts in therapeutics represent a large number of chemi-
databases and reliable software [33]. In this minireview, we cal–biological informatics-based programs. Toxicology-oriented com-
discuss possible ways of the systematization of the in silico putational approaches as a rule are based on building toxicity
methods that can be used for fast, preliminary estimation of databases. This gives possibility to carry out the QSAR analysis
toxicity of compounds with the possible therapeutically effects. (modeling) [35].
Fig. 1 shows the general scheme of the QSAR analysis and
illustrates the roles of the databases and QSAR models.
The main reasons for the in silico prediction of toxicity in
2. Sources of drug toxicity data general are [36]: (i) Pressure to reduce animal testing; (ii)
Computational models provide suitable toxicity prediction; (iii)
Apparently, as concluded by the results of a recent study, colla- Legislation (Governmental policies in both the European Union
boration between industry, computational software developers and (EU) and North America) has encouraged and, in some cases,
regulatory researchers led to the development of a toxicity database mandated the use of computational techniques to predict toxicity.
and classification rules for NMEs [34]. Table 1 contains the basic For example, the US EPA has utilized QSAR to assist in the pre-
principles of classification for substances available via database [34]. manufactory notification of new chemicals, especially where no
There are various kinds of toxic endpoints. An endpoint can be toxicity data are physically in hand. This requirement for models
related to different organisms, e.g. rats, mice, fishes, birds, and has inspired considerable advancement in the prediction of acute
human. From point of view of praxis most important toxicity is toxicity for environmental endpoints); (iv) Filling data gaps;
one that is related to human, but these data are very limited. (v) Cost and time reduction; (vi) Identification of new toxicological
Therefore, there are attempts to select organism with some similar- problems; (vii) Designing of new compounds; (viii) Higher
ity to organism of human (rat, dog, monkey, etc.). An endpoint can throughput and in silico approaches have higher reproducibility
be related to various routes of adsorption (inhalation, oral, skin, etc.). if the same model is used. Again, this in silico approaches have low
Consequently, a toxic endpoint can be measured by different units compound synthesis requirement; (ix) In silico models have the
22 A.A. Toropov et al. / Computers in Biology and Medicine 45 (2014) 20–25

Table 2
Databases which are sources of available data on toxicity with the following classification: A ¼ numerical data on experimental toxic endpoints; B ¼numerical data on
calculated (or extracted from several references) toxic endpoints; C ¼documentation related to toxicity (references and/or descriptions).

Database Accessibility Description Class

FAERS http://www.fda.gov/Drugs/ Adverse Effects Reporting system (FAERS) of post-market safety surveillance for all C
GuidanceComplianceRegulatoryInformation/Surveillance/ approved drug and therapeutic biologic products
AdverseDrugEffects/default.htm
ACToR http://actor.epa.gov/actor/faces/ACToRHome.jsp ACToR ( Aggregated Computational Toxicology Resource ) is EPA's online warehouse of A
publicly available chemical toxicity data.ACToR provides the numerical data on over
500,000 environmental chemicals searchable by chemical name and/or by chemical
structure
Cal/EPA http://www.oehha.ca.gov/risk/ChemicalDB/index.asp State of California EPA Toxicity. User can obtain information about CAS number, use, A
list of synonyms, and a group of criteria for risk assessment of a wanted substance
CCRIS http://toxnet.nlm.nih.gov/cgi-bin/sis/htmlgen?CCRIS Chemical carcinogenesis research information system (CCRIS). The numerical data on C
various carcinogenic endpoints (mice, rats, ames salmonella typhimurium, and
human) for over 8000 compounds
CPDB http://potency.berkeley.edu/ University of California, Berkeley, carcinogenic potency database contains long-term A
animal cancer tests on 1547 chemicals
Drugs@FDA http://www.fda.gov/Drugs/InformationOnDrugs/ Information about brand name and generic prescription and over-the-counter human A
ucm135821.htm drugs and biological therapeutic products
DSSTox http://www.epa.gov/ncct/dsstox/index.html User can obtain data on toxicity represented by PDF, SDF or XLS files A
ECOTOX http://cfpub.epa.gov/ecotox/ User can use quick and/or advanced database query. There is an user guide. There are C
links to other databases on toxicity
EXTOXNET http://extoxnet.orst.edu/ghindex.html University-based database of issues related to pesticide toxicology A
FDA http://www.accessdata.fda.gov/scripts/plantox/index.cfm US FDA/CFSAN database with references to scientific literature describing studies of C
Poisonous the toxic properties of plants
Plant
Database
Gene-Tox http://toxnet.nlm.nih.gov/cgi-bin/sis/htmlgen?GENETOX US NLM Peer-reviewed genetic toxicology test data for over 3000 chemicals A
HERA http://www.heraproject.com/RiskAssessment.cfm Human and Environmental Risk Assessment on Ingredients and Household Cleaning C
Products; toxicity and risk data on ingredients supplied and formulated by European
manufacturers
Household http://hpd.nlm.nih.gov/ This database contains over 12,000 consumer brands with numerical criteria their C
Products health effects
IRIS http://cfpub.epa.gov/ncea/iris/index.cfm Integrated Risk Information System; a compilation of electronic reports on C
environmental substances and their potential to cause human health effects. User can
obtain PDF file with description of toxicological review of a substance in detail
ITER http://www.tera.org/iter/ Database of human health risk values and cancer classifications for over 600 C
environmental chemicals
JECDB http://dra4.nihs.go.jp/mhlw_data/jsp/SearchPageENG.jsp Japanese Ministry of Health, Labour and Welfare Chemical Toxicity Database; toxicity C
data for 369 chemicals
LAZAR http://www.in-silico.de/ Lazy structure–activity relationships database; provides QSAR predictions for liver B
toxicity, mutagenicity, and carcinogenicity
MR http://www.atsdr.cdc.gov/mrls/index.html User can obtain data on minimal risk (MR) levels for hazardous substances A
represented in this database
N-Class http://apps.kemi.se/nclass/ Using special interface, user can define a wanted compound. The system provide C
database, information on large list of different endpoints related to this compound
KemI
NPIC http://npic.orst.edu National Pesticide Information Center through Oregon State University and US EPA C
provides science-based information about pesticides including toxicity. In fact the
database is online encyclopedia for pesticides
NTP http://ntp.niehs.nih.gov/ US NIH/NIEHS National Toxicology Program testing status and information of agents C
registered in the US of public health interest. User can obtain documents related to
different substances and protocols of definition of different toxic endpoints as well as
information on other aspects of toxicology in general
PAN Pesticide http://www.pesticideinfo.org/ Pesticide Action Network North America; data on 6500 pesticides, insecticides and C
herbicides including toxicity, water pollution, ecological toxicity, uses, and regulatory
status. In fact the database is a digest of pesticides
Riskline, KemI http://apps.kemi.se/riskline/ Contains information on both environment and health Useful for classification and B
labelling. Provides links to references, associated with a chemical. User can obtain a
set of abstracts related to indicated substance
STITCH http://stitch.embl.de/ Search Tool for Interactions of Chemicals (STITCH). Knowledge database to explore C
known and predicted interactions between proteins and small-molecule chemicals for
understanding of molecular and cellular functions. Over 68,000 chemicals are
represented
TEXTRATOX http://www.vet.utk.edu/TETRATOX/index.php The University of Tennessee Institute of Agriculture. A collection of aquatic toxic B
potency data for more than 2400 industrial organic compounds
TOXNET http://toxnet.nlm.nih.gov/ Databases on toxicology, hazardous chemicals, environmental health, and toxic A
releases. User can obtain data on toxic endpoint related to different animals and
human as well as data on physicochemical endpoints, such as boiling points, water
solubility, logP (octanol–water) etc.
ToxRefDB http://www.epa.gov/ncct/toxrefdb/ US EPA relational database of standard toxicity test results for pesticides and other C
environmental chemicals including acute, subchronic, chronic, reproductive, and
developmental toxicity in support of the ToxCast program. User can obtain data
represented by XLS files
 A.A. Toropov et al. / Computers in Biology and Medicine 45 (2014) 20–25 23

Fig. 1. Graphical representations of information (i) from databases; and (ii) from QSAR analysis.

Table 3
Software available for QSAR analysis of toxic endpoints.

Software Accessibility Description

ChemProp http://www.ufz.de/index.php? ChemProp contains several programs which provide batch processing in terms of QSAR runs for
en=10684 compound lists
CODESSA http://www.codessa-pro.com/ Generator of large set of various molecular descriptors for multiple regression analysis, artificial
neuron networks, partial least squares method, etc. (commercial)
CORAL http://www.insilico.eu/coral Free available software and data on some endpoints including toxicity. There are detailed descriptions
together with references on published articles
DRAGON http://www.ccl.net/qsar/archives/ Generator of large set of various molecular descriptors for multiple regression analysis, artificial
0403/0458.html neuron networks, partial least squares method, etc. (commercial)
GOLPE http://www.miasrl.com/golpe.htm GOLPE (Generating Optimal Linear PLS Estimations) is a system of programs for 3D QSAR (commercial)
OECD ToolBox http://www.oecd.org/env/ehs/ Contains standardized software and databases related to QSPR/QSAR problematics
PRECLAV http://www.softpedia.com/get/ PRECLAV (PRoperty Evaluation by CLAss Variables) is a packet of programs for QSPR/QSAR analysis
Science-CAD/PRECLAV.shtml (commercial)
RmSquare http://aptsoftware.co.in/rmsquare/ Estimation of prtedictability of QSPR/QSAR models with special metrics (Rm2). There are detailed
description together with references on published articles
T.E.S.T http://www.epa.gov/nrmrl/std/ Toxicity Estimation Software Tool (TEST) will enable users to easily estimate acute toxicity using the
qsar/qsar.html QSAR methodologies: (i) Hierarchical method (ii) FDA method (iii) Single-model method (iv)
Consensus method, etc. TEST can be used for the analysis of LC50(48 h), LC50 (96 h), BCF, and
mutagenicity
Virtual Computational http://www.vcclab.org/ This software provides on-line tools, which can be helpful for QSPR/QSAR analyses of various
Chemistry Laboratory endpoints (solubility, octanol/water partition coefficient, etc.)

ability to predict ADME related properties on virtual structures health, the manufacture, storage, distribution and release of these
enabling exploration of chemical space without the need to carry substances to the environment after their use, are controlled and
out chemical laboratory synthesis and experimental testing until regulated at local, national and international levels by different
a likely hit or candidate molecule is selected; (x) Development of governments and regulatory agencies worldwide.
understanding of basic biology and chemistry (In the modelling of An important aspect of the in silico methods involved in the
acute toxicological endpoints, much has been gained regarding estimation of the drug toxicity is so-called organ toxicity [37]. High
mechanisms of action). For many modelling approaches, it may be quality QSARs are built up for the following kinds of organ toxicity
assumed that compounds fitting the same QSAR are acting by the [37]: hepatotoxicity, liver necrosis, liver relative weight gain, liver
same mechanism of action. This has allowed researchers to define lipid accumulation, nephrotoxicity, kidney necrosis, kidney rela-
the chemical domain of certain mechanisms [36]. tive weight gain, nephron injury.
Currently, the toxic potential of large quantities of industrial If there are a group of QSAR models available for a serious of
chemicals including pharmaceuticals, cosmetics, pesticides and compounds, the search for the consensus of these models seems to
other synthetic or semi-synthetic chemicals is often required to provide robust and accurate approach [4,36].
be assessed by using standard animal models. This comprises the It should be noted that the Monte Carlo technique [38–41] can
basic test protocol for risk assessments necessary for their approval be involved in solution of the QSAR problems related to drug
as a registered product to be launch into the market. With toxicity. Detailed description of the Monte Carlo technique is
increasing concern about the environmental pollution and human available in the literature [42].
24 A.A. Toropov et al. / Computers in Biology and Medicine 45 (2014) 20–25

The comparison of different software includes (i) convenience References

of using; (ii) reliability of results; and (iii) practical scientific and
heuristic potential of output data. The DRAGON and CODESSA are [1] L.G. Valerio Jr, Predictive computational toxicology to support drug safety
most cited programs, but they are commercial (as well GOLPE and assessment, Methods Mol. Biol. 930 (2013) 341.
[2] G.F. Makhaeva, E.V. Radchenko, V.A. Palyulin, E.V. Rudakova, A.Y. Aksinenko, V.
PRECLAV whereas all other programs from Table 3 are free). TEST B. Sokolov, N.S. Zefirov, R.J. Richardson, Organophosphorus compound ester-
and Virtual Computational Chemistry Laboratory are available ase profiles as predictors of therapeutic and toxic effects, Chem. Biol. Interact.
online and have convenient interface. OECD toolbox is official 203 (2013) 231.
program, consequently, the user of this software can get “very law- [3] A. Speck-Planche, V.V. Kleandrova, M.N.D.S. Cordeiro, New insights toward the
discovery of antibacterial agents: multi-tasking QSBER model for the simulta-
abiding” data from point of view of EC administrations. ChemProp neous prediction of anti-tuberculosis activity and toxicological profiles of
have convenient input (molecular structure can be represented by drugs, Eur. J. Pharm. Sci. 48 (2013) 812.
graphical editor, via SMILES and InChI. The output can be repre- [4] A. Tropsha, Best practices for QSAR model development, validation, and
exploitation, Mol. Inf. 29 (2010) 476.
sented by both HTML tables available on the Internet and simple [5] D.K. Yadav, F. Khan, A.S. Negi, Pharmacophore modeling, molecular docking,
text file. PRECLAV is a standardized software with limitation for QSAR, and in silico ADMET studies of gallic acid derivatives for immunomo-
the number of compounds under consideration (maximum is 50). dulatory activity, J. Mol. Model. 18 (2012) 2513.
GOLPE is a tool for fast analysis of large data. RmSquare is a tool to [6] D.K. Yadav, F. Khan, QSAR, docking and ADMET studies of camptothecin
derivatives as inhibitors of DNA topoisomerase-I, J. Chemom. 27 (2013) 21.
estimate predictability of a model using data on experimental and [7] N. Perisic-Janjic, R. Kaliszan, N. Milosevic, G. Uscumlic, N. Banjac N, Chromato-
predicted values of endpoint for test set. The CORAL software is graphic retention parameters in correlation analysis with in silico biological
a tool for QSPR/QSAR analyses of data where the molecular descriptors of a novel series of N-phenyl-3-methyl succinimide derivatives,
J. Pharm. Biomed. Anal. 72 (2013) 65.
structure is represented by SMILES. [8] S. Segan, J. Trifkovic, T. Verbic, D. Opsenic, M. Zlatovic, J. Burnett, B. Solaja,
D. Milojkovic-Opsenic, Correlation between structure, retention, property, and
activity of biologically relevant 1,7-bis(aminoalkyl)diazachrysene derivatives,
J. Pharm. Biomed. Anal. 72 (2013) 231.
[9] D. Fourches, J.C. Barnes, N.C. Day, P. Bradley, J.Z. Reed, A. Tropsha, Cheminfor-
4. Conclusions matics analysis of assertions mined from literature that describe drug-induced
liver injury in different species, Chem. Res. Toxicol. 23 (2010) 171.
In silico methods, able to reliable estimate toxicity of potential [10] A.A. Frid, E.J. Matthews, Regul. Toxicol. Pharm., 56, Prediction of drug-related
cardiac adverse effects in humans-B: use of QSAR programs for early detection
therapeutical agents are complex systems. They are improved and of drug-induced cardiac toxicities (2010) 276.
evolve on the yearly bases. Their improvement is accompanied by [11] K. Roy, I. Mitra, Electrotopological state atom (E-State) index in drug design,
expansion of the number of large databases (Table 2) and increase QSAR, property prediction and toxicity assessment, Curr. Comput. Aided Drug
of the number of various computer programs (Table 3). In silico Des. 8 (2012) 135.
[12] S. Kortagere, S. Ekins, Troubleshooting computational methods in drug
toxicology methods are practical, evidence-based and provide high discovery, J. Pharmacol. Toxicol. Methods 61 (2010) 67.
throughput, with varying accuracy. In silico approaches are of keen [13] M. Fernandez, J. Caballero, L. Fernandez, A. Sarai, Genetic algorithm optimiza-
interest, not only to scientists in the private sector and to academic tion in drug design QSAR: Bayesian-regularized genetic neural networks
(BRGNN) and genetic algorithm-optimized support vectors machines (GA-
researchers worldwide, but also to the public [32,35,36]. They are SVM), Mol. Diversity 15 (2011) 269.
being increasingly evaluated and applied by regulators. Although [14] S. Kar, K. Roy, Development and validation of a robust model for prediction of
there are foreseeable beneficial aspects – including maximizing carcinogenicity of drugs, Indian J. Biochem. Biophys. 48 (2011) 111.
[15] L.G. Valerio Jr, K.P. Cross, Characterization and validation of an in silico
use of prior test data and the potential for minimizing animal use
toxicology model to predict the mutagenic potential of drug impurities,
for future toxicity testing – the primary use of in silico toxicology Toxicol. Appl. Pharm. 260 (2012) 209.
methods in the pharmaceutical sciences involves decision support [16] C.D. Ruark, C.E. Hack, P.J. Robinson, E. Paul, P.E. Anderson, J.M. Gearhart,
information [43]. Quantitative structure–activity relationships for organophosphates binding to
acetylcholinesterase, Arch. Toxicol. 87 (2013) 281.
Obviously, management of drug design is rife with complicated, [17] M. Bartels, D. Ricka, E. Lowe, G. Loizou, P. Price, M. Spendiff, S. Arnold, J. Cocker,
contentious, and risky decisions. These decisions are quite difficult N. Ball, Development of PK- and PBPK-based modeling tools for derivation of
and complex. They involve significant uncertainty, multiple stake- biomonitoring guidance values, Comput. Meth. Prog. Biol. 108 (2012) 773.
[18] A. Chen, M.L. Yarmush, T. Maguire, Physiologically based pharmacokinetic
holder groups, competing objectives, and dynamic, non-linear
models: integration of in silico approaches with micro cell culture analogues,
interdependencies. All these aspects test the limits of unaided Curr. Drug Metab. 13 (2012) 863.
human judgment. In the past, formal methods of risk analysis have [19] J. Kirchmair, M.J. Williamson, J.D. Tyzack, L. Tan, P.J. Bond, A. Bender, R.C. Glen,
been used to evaluate new technologies, but these methods ignore Computational prediction of metabolism: sites, products, SAR, P450 enzyme
dynamics, and mechanisms, J. Chem. Inf. Model. 52 (2012) 617.
decision-relevant qualitative information and rely on a volume of [20] A.A. Toropov, A.P. Toropova, I. Raska Jr, E. Benfenati, G. Gini, QSAR modeling of
quantitative engineering and scientific data that simply does not endpoints for peptides which is based on representation of the molecular
exist for many potential therapeutical agents (e.g. peptides, bio- structure by a sequence of amino acids, Struct. Chem. 23 (2012) 1891.
[21] A.P. Toropova, A.A. Toropov, B.F. Rasulev, E. Benfenati, G. Gini, D. Leszczynska,
polymers, nanomaterials) [44].
J. Leszczynski, QSAR models for ACE-inhibitor activity of tri-peptides based on
The main problem of the field (in silico estimation of drug representation of the molecular structure by graph of atomic orbitals and
toxicity) is absence of the standardization of available information SMILES, Struct. Chem. 23 (2012) 1873.
for the both the databases and the software. Unfortunately, this [22] A.A. Toropov, A.P. Toropova, I. Raska Jr, E. Benfenati, G. Gini, Development of
QSAR models for predicting anti-HIV-1 activity using the Monte Carlo method,
problem is characterized by not only the “methodological disper- Cent. Eur. J. Chem. 11 (2013) 371.
sion”, but also the geographical one [36], i.e. very eclectic criteria [23] A.P. Toropova, A.A. Toropov, E. Benfenati, D. Leszczynska, J. Leszczynski, QSAR
for estimation of data which are using in various countries. modeling of measured binding affinity for fullerene-based HIV-1 PR inhibitors
by CORAL, J. Math. Chem. 48 (2010) 959.
Thus, harmonization, systematization, and standardization of
[24] A.A. Toropov, A.P. Toropova, E. Benfenati, D. Leszczynska, J. Leszczynski, InChI-
criteria and methods involved in the in silico drug discovery should based optimal descriptors: QSAR analysis of fullerene[C60]-based HIV-1 PR
be the critical issue at the international level for the nearest future inhibitors by correlation balance, Eur. J. Med. Chem. 45 (2010) 1387.
[35,36,43]. [25] A.A. Toropov, A.P. Toropova, E. Benfenati, D. Leszczynska, J. Leszczynski, SMILES-
based optimal descriptors: QSAR analysis of fullerene-based HIV-1 PR inhibitors by
means of balance of correlations, J. Comput. Chem. 31 (2010) 381.
[26] A.A. Toropov, A.P. Toropova, E. Benfenati, G. Gini, T. Puzyn, D. Leszczynska,
J. Leszczynski, Novel application of the CORAL software to model cytotoxicity
of metal oxide nanoparticles to bacteria Escherichia coli, Chemosphere 89
Conflict of interest Statement
(2012) 1098.
[27] T. Puzyn, B. Rasulev, A. Gajewicz, X. Hu, T.P. Dasari, A. Michalkova,
None declared H.-M. Hwang, A. Toropov, D. Leszczynska, J. Leszczynski, Using nano-QSAR
 A.A. Toropov et al. / Computers in Biology and Medicine 45 (2014) 20–25 25

to predict the cytotoxicity of metal oxide nanoparticles, Nat. Nanotechnol. 6 [36] S. Kar, K. Roy, Predictive toxicology using QSAR: a perspective, J. Indian Chem.
(2011) 175. Soc. 87 (2010) 1455.
[28] A. Gajewicz, B. Rasulev, T.C. Dinadayalane, P. Urbaszek, T. Puzyn, [37] E. Myshkin, R. Brennan, T. Khasanova, T. Sitnik, T. Serebriyskaya, E. Litvinova,
D. Leszczynska, J. Leszczynski, Advancing risk assessment of engineered A. Guryanov, Y. Nikolsky, T. Nikolskaya, S. Bureeva, Prediction of organ toxicity
nanomaterials: application of computational approaches, Adv. Drug Delivery endpoints by QSAR modeling based on precise chemical-histopathology
Rev. 64 (2012) 1663. annotations, Chem. Biol. Drug Des. 80 (2012) 406.
[29] V. Hamon, D. Horvath, C. Gaudin, J. Desrivot, C. Junges, A. Arrault, M. Bertrand,
[38] E. Ibezim, P.R. Duchowicz, E.V. Ortiz, E.A. Castro, QSAR on aryl-piperazine
P. Vayer, QSAR modelling of CYP3A4 inhibition as a screening tool in the
derivatives with activity on malaria, Chemom. Intell. Lab. Syst. 110 (2012) 81.
context of drug–drug interaction studies, Mol. Inf. 31 (2012) 669.
[30] T. Ishikawa, H. Hirano, H. Saito, K. Sano, Y. Ikegami, N. Yamaotsu, S. Hirono, [39] J.C. Garro Martinez, P.R. Duchowicz, M.R. Estrada, G.N. Zamarbide, E.A. Castro,
Quantitative structure–activity relationship (QSAR) analysis to predict drug–drug QSAR study and molecular design of open-chain enaminones as anticonvul-
interactions of ABC transporter ABCG2, Mini-Rev. Med. Chem. 12 (2012) 505. sant agents, Int. J. Mol. Sci. 12 (2011) 9354.
[31] J.A. Rhoades, Y.K. Peterson, H.J. Zhu, D.I. Appel, C.A. Peloquin, J.S. Markowitz, [40] J. García, P.R. Duchowicz, M.F. Rozas, J.A. Caram, M.V. Mirífico, F.M. Fernández,
Prediction and in vitro evaluation of selected protease inhibitor antiviral drugs E.A. Castro, A comparative QSAR on 1,2,5-thiadiazolidin-3-one 1,1-dioxide
as inhibitors of carboxylesterase 1: a potential source of drug–drug interac- compounds as selective inhibitors of human serine proteinases, J. Mol.
tions, Pharm. Res. 29 (2012) 972. Graphics Modell. 31 (2011) 10.
[32] A. Mullard, FDA drug approvals, Nat. Rev. Drug Discovery 10 (2011) (2010) 82. [41] L.M.A. Mullen, P.R. Duchowicz, E.A. Castro, QSAR treatment on a new class of
[33] A.M. Helguera, R.D. Combes, M. Pérez González, M.N.D.S. Cordeiro, Applica- triphenylmethyl-containing compounds as potent anticancer agents,
tions of 2D descriptors in drug design: a DRAGON tale, Curr. Top. Med. Chem. 8 Chemom.. Intell. Lab. Syst. 107 (2011) 269.
(2008) 1628.
[42] A.P. Toropova, A.A. Toropov, E. Benfenati, G. Gini, D. Leszczynska, J. Leszczynski,
[34] K.L. Dobo, N. Greene, C. Fred, S. Glowienke, J.S. Harvey, C. Hasselgren, R. Jolly,
CORAL: quantitative structure–activity relationship models for estimating
M.O. Kenyon, J.B. Munzner, W. Muster, R. Neft, M.V. Reddy, A.T. White,
S. Weiner, In silico methods combined with expert knowledge rule out toxicity of organic compounds in rats, J. Comput. Chem. 32 (2011) 2727.
mutagenic potential of pharmaceutical impurities: an industry survey, Regul. [43] L.G. Valerio Jr, In silico toxicology models and databases as FDA critical path
Toxicol. Pharm. 62 (2012) 449. initiative toolkits, Hum. Genomics 5 (2011) 200.
[35] L.G. Valerio Jr, In silico toxicology for the pharmaceutical sciences, Toxicol. [44] I. Linkov, M.E. Bates, B.D. Trump, T.P. Seager, M.A. Chappell, J.M. Keisler, For
Appl. Pharmacol. 241 (2009) 356. nanotechnology decisions, use decision analysis, Nano Today 8 (2013) 5.