Biosimulation

Seminar: Molecular Docking

with AutoDock4

by Jennifer Loschwitz
WS 2017/2018
1
 Task: Re-docking with AutoDock
• Evaluation of the ability of AutoDock
– To reproduce the native binding mode of a ligand that is extracted from a protein
crystal structure and then re-docked into this crystal structure
– How good are the results?

• Learning how to use the docking program AutoDock with ADT

– Prediction of crystal binding mode of cyclooxygenase (COX) inhibitor flurbiprofen
(FLP) in isozymes COX-1 (PDB code: 1CQE)
– Pharmaceutical inhibition of COX can provide relief from the symptoms of
inflammation and pain
– COX produces prostanoids, including thromboxane and prostaglandins such as
prostacyclin

2
 Docking program: AutoDock
• http://autodock.scripps.edu/

• For the usage of AutoDock:

– AutoDock4: Performing the docking of the ligand to a set of
grids describing the target protein
– AutoGrid4: Pre-calculation of grids for the protein-ligand
interaction
– AutoDockTools (ADT): Graphical interface for AutoDock and
AutoGrid

• How to install AutoDock and ADT:

– http://autodock.scripps.edu/faqs-help/faq/how-do-i-install-
autodock-on-linux-and-mac-os-x
– http://autodock.scripps.edu/resources/adt

• Docking tutorials with AutoDock:

– http://autodock.scripps.edu/faqs-help/tutorial
3
AutoDockTools (ADT)

4
 AutoDockTools (ADT)
• ADT program is based on the Python Molecular Viewer (PMV) and provides
a graphical user interface (GUI) to perform docking with AutoDock

• Command for starting of ADT:

– > /Path/to/adt &

• PMV menu
– Some general methods available in PMV.
– This includes tools
• to import and export molecules
• to select and modify molecules
• to create images from molecules

• ADT menu
– Provides some AutoDock-specific tools

• Viewer
– Showing all molecules imported to ADT

• Dashboard
– Quick selection, visualization, and coloring of molecules currently displayed in the 5
viewer
 General steps for docking
with AutoDock
1. Ligand preparation

2. Calculation of protein-ligand interactions

using AutoGrid

3. Docking of the ligand in the crystal binding

pose using AutoDock and search algorithms

4. Analysis of docking results

6
 Steps for the ligand preparation
• Loading of the ligand structure in ADT menu
– ADT menu → Ligand → Input → Open → ligand.pdb
– An information window about the ligand:
• Adding of missing hydrogen atoms
• Number of aromatic carbon atoms
• Number of rotation bonds

• Creation of the torsion tree for ligand flexibility

– ADT menu → Ligand → Torsion Tree → Detected Root
– Position of an optimal ligand center as green sphere to start
the conformational changes of the ligand

• Saving the ligand with the torsion tree as pdbqrt

file
– ADT menu → Ligand → Output → Save as PDBQT →
ligand.pdbqt
– AutoDock uses the pdbqt file format for ligand and receptor
– pdpqt file contains atomic coordinates, partial charges,
AutoDock 4 atom-types, torsion tree to represent the rigid
and rotatable pieces of the ligand and number of torsional
degrees of freedom in the ligand

More Information about pdbqt:

http://autodock.scripps.edu/faqs-help/faq/what-is-the-format-of-a-pdbqt-file 7
 General steps for docking
with AutoDock
1. Ligand preparation

2. Calculation of protein-ligand interactions

using AutoGrid

3. Docking of the ligand in the crystal binding

pose using AutoDock and search algorithms

4. Analysis of docking results

8
 Steps for calculation of protein-
ligand interactions using AutoGrid
• Using a 3D grid over the protein
1. Choosing types of the probe atoms depending of ligand
 NA, OA, C, A (aromatic carbon)

2. Setting probe atoms of one type on every grid point of the

box
Repeat for
every probe 3. Calculating interactions with the environment using
atom type AutoDock force field

4. Saving the potential values in a grid file (Thickness of the

points means the strength of the interaction)

9
 AutoDock force field
• Only intermolecular interactions are considered in AutoDock
compared to MD force fields:

1. VdW-forces via Lennard-Jones potential:

2. Hydrogen bonds to water and protein:

3. Electrostatics via Coulomb law:

4. Desolvation for hydrophobic effect (AutoDock4):

More Information about AutoDock force field: 10

http://autodock.scripps.edu/resources/science/equations/
 Steps for running AutoGrid
• Loading and saving the protein structure and choosing the probe atom types depending of the ligand
– ADT menu → Grid → Macromolecule → Open → save as protein.pdbqt
– ADT menu → Grid → Set Map Types → Choose Ligand

• Parameters of the 3D grid

1. ADT menu → Grid → Grid Box
2. Opening of a window with the grid settings (size and center)
3. The center of the box should be located on the ligand: Center → Center on ligand
4. Saving of the current settings: File → Close saving current
5. Creation of a gpf file containing the grid parameters: ADT menu → Grid → Output → Save GPF

11
 Steps for running AutoGrid
• Starting of AutoGrid
1. ADT menu → Run → Run AutoGrid
2. Opening of a window → choose the gpf file → Launch
– Appearing of a small window, showing that the calculation is
running and disappearing automatically when the calculation
is finished

12
 General steps for docking
with AutoDock
1. Ligand preparation

2. Calculation of protein-ligand interactions

using AutoGrid

3. Docking of the ligand in the crystal binding

pose using AutoDock and search algorithms

4. Analysis of docking results

13
 Searching of the global minimum
1. Generation of ligand
conformation and
orientation

2. Selection of some
solutions via scoring
function

3. Generation of new
ligand conformation
and orientation based
on selected one
 which way to go?

4. Go back to step 2 until

the solution is good
enough or the number
of maximum iterations
is reached

14
 Genetic algorithms
for the global optimization
• Genetic algorithms (GA) based on the evolution
– Encoding of the molecule: Mapping
– Population of the individuals: Parents and Offspring
– Inheritance: Mutation and Crossover
– Selection: Scoring function
– Elitism: How many top individuals will be automatically survived
into the next generation?

15
 Genetic algorithms
for the global optimization
• Scoring function (fitness → binding energy of the ligand ΔG)

– n0 = Number of offspring of the corresponding individual i

– fkworst = Worst individual in the last generation N (highest ΔG)
– fi = Fitness of the individual i (ΔG)
– < f > = Mean fitness of the population (< ΔG >)

• Convergence and termination criteria

– fkworst = < f > (Convergence)
– Number of GA runs
– Number of energy evaluations

• Genetic algorithms
– Lamarckian genetic algorithm (LGA)
– Darwinian genetic algorithm
More information about LGA:
G.M. Morris et al., Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function, 16
J. Comput. Chem. ,1998.
Steps of the genetic algorithm
for the global optimization

17
 Steps for setting of the LGA
and running AutoDock
• Loading of the ligand and protein structure
– ADT menu → Docking → Macromolecule → Set Rigid
Filename → protein.pdbqt
– ADT menu → Docking → Ligand → Choose →
ligand.pdbqt
– Appearing of an information window about atom types,
number of torsion angles and center of the ligand

• Settings of the LGA

– ADT menu → Docking → Search parameters → Genetic
Algorithm
– Number of GA Runs: Number of independent
optimization runs
– Maximum Number of energy evals: Number of energy
evaluations that are performed during one optimization
run → high number gives better results but also needs
more calculation time
– ADT menu → Docking → Output → Lamarckian GA →
save as dpf

• Running of docking with AutoDock

– ADT menu → Run → Run Auto-Dock → choose of dpf
file → launch

18
 General steps for docking
with AutoDock
1. Ligand preparation

2. Calculation of protein-ligand interactions

using AutoGrid

3. Docking of the ligand in the crystal binding

pose using AutoDock and search algorithms

4. Analysis of docking results

19
 Steps for analyzing
the docking results
• Loading the results
– ADT menu → Analyze → Dockings → Open → choose dgl file

• Showing the clustering to evaluate the convergence

– ADT menu → Analyze → Clusterings → Show
– Open of a window with a clustering histogram between the different docking solutions
– In clustering, similar docking solutions form one cluster (RMSD 0.5 Å or 2 Å)
– A docking experiment is successful with respect to convergence if the majority of all
available docking poses falls into the same cluster

20
 Steps for analyzing
the docking results
• Showing the conformation ranked by the binding energy
– ADT menu → Analyze → Conformations → Play, ranked by Energy
– Button &: More information about the generated conformations
– Button & → Show Info: Detailed information of values for the current docking pose:
1. binding_energy: Binding energy between protein and ligand
2. refRMS: RMSD value to the reference native binding pose
3. Inhibition_constant: Inhibition constant of the ligand to protein

21
 Investigation of the
predicted ligand poses
• A RMSD < 2 Å between the result and the crystal ligand pose
indicates a good docking solution.
– How good are your docking results?

• Are there any differences among the resulting docking

solutions, or did all docking trials converged to a similar
binding pose?

• Which value has the binding energy of the top-ranked ligand?

• Do you expect that you will obtain the same results if you
would repeat the docking?
– Compare your results with those of your neighbors by analyzing the
generated binding poses and the corresponding binding energies

22