User’s Guide

AutoDock

Automated Docking of Flexible Ligands to Receptors

3
Version 3.0.5

Garrett M. Morris
David S. Goodsell
Ruth Huey
William E. Hart
Scott Halliday
Rik Belew
Arthur J. Olson
2

Important

AutoDock is distributed free of charge for academic and non-commercial use. There are some caveats, however.
Firstly, since we do not receive funding to support the academic community of users, we cannot guarantee rapid (or
even slow) response to queries on installation and use. While there is documentation, it may require at least some
basic Unix abilities to install. If you need more support for the AutoDock code, a commercial version (with support)
is available from Oxford Molecular (http://www.oxmol.com). If you can’t afford support, but still need help:

(1) Ask your local system administrator or programming guru for help about compiling, using Unix/Linux, etc..

(2) Consult the AutoDock web site, where you will find a wealth of information and a FAQ (Frequently Asked
Questions) page with answers on AutoDock:

http://www.scripps.edu/pub/olson-web/doc/autodock/

(3) If you can’t find the answer to your problem, send your question to the Computational Chemistry List (CCL).
There are many seasoned users of computational chemistry software and some AutoDock users who may
already know the answer to your question. You can find out more about the CCL on the web, at:

http://ccl.osc.edu/ccl/welcome.html

(4) If you have tried (1), (2) and (3), and you still cannot find an answer, send email to [email protected] for
questions about AutoGrid or AutoDock; or to [email protected], for questions about AutoTors.

Thanks for your understanding!

E-mail addresses

Arthur J. Olson, Ph.D.: [email protected]

Peggy Graber, Senior Administrative Assistant: [email protected]
Garrett M. Morris, M.A., D.Phil. [email protected]
David S. Goodsell, Ph.D. [email protected]
Ruth Huey, Ph.D. [email protected]

Fax: + (858) 784-2860

AutoDock, AutoGrid, AutoTors, Copyright © 1991-2000,

The Scripps Research Institute,

Molecular Graphics Laboratory,
Department of Molecular Biology, Mail Drop MB-5,
10550 North Torrey Pines Road,
La Jolla, CA 92037-1000, U.S.A.

Modification date: November 20, 2001 2:20 pm

November 20, 2001

 3

Contents
Automated Docking 5
Introduction 5
Overview of the Method 6
Applications 7
What’s New 9

Theory 11
Overview of the Free Energy Function 11
Grid Maps 13
Van der Waals Potential Energy 15
Modelling Hydrogen Bonds 20
A note on atom type codes: 21
Electrostatic Potential Grid Maps 22

Methodology 24
Getting Started... 24
Setting Up AutoGrid and AutoDock Jobs 25
Preparing the Ligand 26
Ligand Flexibility and Constraints 26
Using AutoTors to Define Torsions in the Ligand 28
Ligand is in PDBQ-format: 28
Ligand is in Mol2-format: 28
AutoTors Output: 28
AutoTors Flags: 29
Running AutoTors 30
Input Stage 31
Root Specification Stage 31
Torsions Detection and Selection Stage 31
Root Expansion and Output Stage 32
Adding Polar Hydrogens to the Macromolecule 32
Running AutoGrid 33
Flexible Docking with AutoDock 33
Monte Carlo Simulated Annealing 35
Genetic Algorithm and Evolutionary Programming Docking 36
Running AutoDock 37
Using the Command Mode in AutoDock 38
Trajectory Files 41
Evaluating the Results of a Docking 42
Visualizing Grid Maps 43
Visualizing Trajectories 43

Shell Scripts and Awk Programs 45

Parameters from AutoDock Version 1 56

AutoDock File Formats 57

Protein Data Bank with Partial Charges: PDBQ 57
PDBQ with Solvation Parameters: PDBQS 57

November 20, 2001

4

AutoGrid Grid Parameter File: GPF 58

AutoGrid Keywords and Commands 59
Grid Map File 61
Grid Map Field File 61
AutoDock Docking Parameter File: DPF 62
Command to set the seed for the random number generator 63
Parameters defining the grid maps to be used 63
Parameters defining the ligand and its initial state 64
Parameters defining ligand step sizes 65
Parameters defining optional ligand torsion constraints 66
Parameter affecting torsional free energy 67
Parameters for ligand internal energies 68
Parameters for simulated annealing searches 68
Parameter to set the amount of output 70
Parameters for trajectory output during SA dockings 70
Parameter for energies of atoms outside the grid 71
Parameter for initializing the ligand in SA 71
Parameters for cluster analysis of docked conformations 71
Parameters for re-clustering the results of several jobs 72
Parameters for genetic algorithm, Lamarckian GA and evolutionary programming searches 72
Command to set genetic algorithm parameters 73
Parameters for local search 74
Commands to choose and set the local search method 75
Commands to perform automated docking 75
Command to perform clustering of docked conformations 76

Example Parameter Files 77

AutoGrid GPF 77
AutoDock DPF 78
Example 1: Docking using Monte Carlo Simulated Annealing 78
Example 2: Clustering Many Dockings 79
Example 3: Docking Using the Lamarckian Genetic Algorithm (LGA) 80

AutoDock References 82
Primary References 82
AutoDock 3.0 82
AutoDock 2.4 82
AutoDock 1.0 83
Reviews of Applications 83
Selected Applications and Citations of AutoDock 83
Web publications 85

November 20, 2001

 Introduction

Automated Docking

Introduction
1. Introduction

The first version of AutoDock1 was distributed to over 35 sites around the world, and that number
has since grown to over 600 sites with the latest versions of AutoDock2,3. This user guide is the
first version to accompany a significantly enhanced version of AutoDock, version 3.0, which
includes powerful new search methods and a new empirical free energy function3.

The program AutoDock was developed to provide an automated procedure for predicting the
interaction of ligands with biomacromolecular targets. The motivation for this work arises from
problems in the design of bioactive compounds, and in particular the field of computer-aided drug
design. Progress in biomolecular x-ray crystallography continues to provide a number of impor-
tant protein and nucleic acid structures. These structures could be targets for bioactive agents in
the control of animal and plant diseases, or simply key to understanding of a fundamental aspect
of biology. The precise interaction of such agents or candidate molecules is important in the
development process. Indeed, AutoDock can be a valuable tool in the x-ray structure determina-
tion process itself: given the electron density for a ligand, AutoDock can help to narrow the con-
formational possibilities and help identify a good structure. Our goal has been to provide a
computational tool to assist researchers in the determination of biomolecular complexes.

In any docking scheme two conflicting requirements must be balanced: the desire for a robust and
accurate procedure, and the desire to keep the computational demands at a reasonable level. The
ideal procedure would find the global minimum in the interaction energy between the substrate
and the target protein, exploring all available degrees of freedom (DOF) for the system. However,
it must also run on a laboratory workstation within an amount of time comparable to other compu-
tations that a structural researcher may undertake, such as a crystallographic refinement. In order
to meet these demands a number of docking techniques simplify the docking procedure. Still one
of the most common techniques in use today is manually-assisted docking. Here, the internal and
orientational degrees of freedom in the substrate are under interactive control. While the energy
evaluation for such techniques can be sophisticated, the global exploration of configurational
space is limited. At the other end of the spectrum are automated methods such as exhaustive
search and distance geometry. These methods can explore configurational space, but at the cost of
a much simplified model for the energetic evaluation.

1. Goodsell, D.S. & Olson, A.J. (1990) “Automated Docking of Substrates to Proteins by Simulated Anneal-
ing”, Proteins: Str. Func. Genet., 8, 195-202.
2. Morris, G. M., Goodsell, D. S., Huey, R. and Olson, A. J. (1996), "Distributed automated docking of
flexible ligands to proteins: Parallel applications of AutoDock 2.4", J. Computer-Aided Molecular Design,
10: 293-304.
3. Morris, G. M., Goodsell, D. S., Halliday, R.S., Huey, R., Hart, W. E., Belew, R. K. and Olson, A. J.
(1998), "Automated Docking Using a Lamarckian Genetic Algorithm and and Empirical Binding Free
Energy Function", J. Computational Chemistry, 19: 1639-1662.

5
 Introduction

The original procedure developed for AutoDock used a Monte Carlo (MC) simulated annealing
(SA) technique for configurational exploration with a rapid energy evaluation using grid-based
molecular affinity potentials. It thus combined the advantages of exploring a large search space
Introduction

and a robust energy evaluation. This has proven to be a powerful approach to the problem of dock-
ing a flexible substrate into the binding site of a static protein. Input to the procedure is minimal.
The researcher specifies a rectangular volume around the protein, the rotatable bonds for the sub-
strate, and an arbitrary or random starting configuration, and the procedure produces a relatively
unbiased docking.

2. Overview of the Method

Rapid energy evaluation is achieved by precalculating atomic affinity potentials for each atom
type in the substrate molecule in the manner described by Goodford 4. In the AutoGrid procedure
the protein is embedded in a three-dimensional grid and a probe atom is placed at each grid point.
The energy of interaction of this single atom with the protein is assigned to the grid point. An
affinity grid is calculated for each type of atom in the substrate, typically carbon, oxygen, nitrogen
and hydrogen, as well as a grid of electrostatic potential, either using a point charge of +1 as the
probe, or using a Poisson-Boltzmann finite difference method, such as DELPHI 5,6. The energetics
of a particular substrate configuration is then found by tri-linear interpolation of affinity values of
the eight grid points surrounding each of the atoms in the substrate. The electrostatic interaction is
evaluated similarly, by interpolating the values of the electrostatic potential and multiplying by
the charge on the atom (the electrostatic term is evaluated separately to allow finer control of the
substrate atomic charges). The time to perform an energy calculation using the grids is propor-
tional only to the number of atoms in the substrate, and is independent of the number of atoms in
the protein.

The docking simulation is carried out using one of a number of possible search methods. The
original AutoDock supported only one search method, although version 3.0 now has several.

The original search algorithm was the Metropolis method, also known as Monte Carlo simulated
annealing. With the protein static throughout the simulation, the substrate molecule performs a
random walk in the space around the protein. At each step in the simulation, a small random dis-
placement is applied to each of the degrees of freedom of the substrate: translation of its center of
gravity; orientation; and rotation around each of its flexible internal dihedral angles. This dis-
placement results in a new configuration, whose energy is evaluated using the grid interpolation
procedure described above. This new energy is compared to the energy of the preceding step. If
the new energy is lower, the new configuration is immediately accepted. If the new energy is
higher, then the configuration is accepted or rejected based upon a probability expression depen-
dent on a user defined temperature, T. The probability of acceptance is given by:

4. Goodford, P.J. (1985) “A Computational Procedure for Determining Energetically Favorable Binding
Sites on Biologically Important Macromolecules”, J. Med. Chem., 28, 849-857.
5. Sharp, K., Fine, R. & Honig, B. (1987) Science, 236, 1460-1463.
6. Allison, S.A., Bacquet, R.J., & McCammon, J. (1988) Biopolymers, 27, 251-269.

6
 Introduction

∆E 
 – ---------
 kBT
P ( ∆E ) = e

Introduction
where ∆E is the difference in energy from the previous step, and kB is the Boltzmann constant. At
high enough temperatures, almost all steps are accepted. At lower temperatures, fewer high
energy structures are accepted.

The simulation proceeds as a series of cycles, each at a specified temperature. Each cycle contains
a large number of individual steps, accepting or rejecting the steps based upon the current temper-
ature. After a specified number of acceptances or rejections, the next cycle begins with a tempera-
ture lowered by a specified schedule such as:
T i = gT i – 1

where Ti is the temperature at cycle i, and g is a constant between 0 and 1.

Simulated annealing allows an efficient exploration of the complex configurational space with
multiple minima that is typical of a docking problem. The separation of the calculation of the
molecular affinity grids from the docking simulation provides a modularity to the procedure,
allowing the exploration of a range of representations of molecular interactions, from constant
dielectrics to finite difference methods and from standard 12-6 potential functions to distributions
based on observed binding sites.

3. Applications

The original FORTRAN version of AutoDock was initially tested on a number of protein-sub-
strate complexes which had been characterized by x-ray crystallography7. These tests included
phosphocholine binding in a antibody combining site, N-formyltryptophan binding to chymot-
rypsin and N-acetylglucosamine binding to Lysozyme. In almost all cases the results of the
AutoDock simulations functionally reproduced the crystallographic complexes. In further appli-
cations AutoDock was used to predict interactions of substrates with aconitase prior to any crys-
tallographic structures for complexes. In this work we not only predicted the binding mode of
isocitrate, but we demonstrated the utility of AutoDock in generating substrate models during the
early stages of crystallographic proteins structure refinement 8. Citrate docking experiments
showed two binding modes, one of which approximated the experimental electron density deter-
mined for an aconitase-nitrocitrate complex. The docking simulation results provided insight into
the proposed reaction mechanism of the enzyme.

7. Goodsell, D.S. & Olson, A.J. (1990) “Automated Docking of Substrates to Proteins by Simulated Anneal-
ing”, Proteins: Str. Func. Genet., 8, 195-202.
8. Goodsell, D.S., Lauble, H., Stout, C.D & Olson, A.J. (1993) “Automated Docking in Crystallography:
Analysis of the Substrates of Aconitase”, Proteins: Str. Func. Genet., 17, 1-10.

7
 Introduction

One novel and intriguing use of the software was reported from Koshland’s laboratory9. These

?
Introduction

receptor MBP
investigators used the known structures of the maltose-binding protein (MBP) and the ligand
binding domain of the aspartate receptor to predict the structure of the receptor-protein complex
(see diagram below). They used knowledge from mutational studies on MBP to select two

MBP MBP octapeptides

octapeptides on the protein known to be involved in the binding to the aspartate receptor, which

receptor
receptor AutoDock AutoDock
receptor

they docked independently to the model of the receptor using our automated docking code (the
backbones of the peptides were fixed, but the side-chain conformations and overall orientations
were unrestrained).

superimpose
MBP + MBP

9. Stoddard, B.L. & Koshland, D.E. (1992) “Prediction of a receptor protein complex using a binary docking
method.”, Nature, 358 (6389), 774-776.

8
 Introduction

The distance and orientation of the two peptides as docked to the receptor corresponded to that in
the intact MBP, thus enabling a reasonable prediction of the protein-receptor complex. This tech-
nique could be generally useful in situations where there are data on multi-site interactions.

Introduction
superimpose
MBP + receptor

MBP/receptor complex

4. What’s New

In AutoDock version 3.0, we have added a promising new, hybrid search technique that imple-
ments an adaptive global optimizer with local search, based on the work of Rik Belew and Will-
iam Hart10 at the Department of Computational Science, University of California at San Diego.
The global search method is a C++ implementation of a modified genetic algorithm (GA), with
2-point crossover and random mutation. The local search method is based on the optimization
algorithm of Solis and Wets11 (SW), which has the advantage that it does not require gradient
information in order to proceed. The local searcher modifies the phenotype, which is allowed to
update the genotype: clearly this contravenes Mendelian genetics observed in nature, but it does
improve the overall performance of the method. We refer to this hybrid genetic algorithm with
phenotypic local search as a Lamarckian Genetic Algorithm or LGA for short, since it utilizes
(discredited) Lamarckian notion that an adaptations of an individual to its environment can be
inherited by its offspring.

The SW local searcher uses fixed variances which are initially and uniformly 1. These variances
are used for probabilistically determining the change to a particular state variable, like the x-trans-
lation. These variances are either doubled or halved during the search, depending on the number
of consecutive successful or failed moves. Success is a drop in energy. We have modified the clas-
sical SW method to take into account in the variances the relative magnitudes of translations and
rotations (in Angstroms and radians respectively). We call this method pseudo-Solis and Wets
(pSW).

The genome consists of floating point genes each of which encodes one state variable describing
the molecular position, orientation and conformation. This is a departure from the classical GA
approach, which dictates a purely binary implementation. By setting the rate of genetic crossover
to zero, and increasing the rate of genetic mutation, our hybrid GA can mimic an evolutionary

10. William Hart’s doctoral thesis describes this hybrid global-local method, and can be found on the World
Wide Wed at “http://www.cs.sandia.gov/~wehart/abstracts_html/thesis.html”.
11. F.J. Solis and R.J.-B. Wets. (1981) “Minimization by random search techniques”, Mathematical Opera-
tions Research, 6, 19-30.

9
 Introduction

programming (EP) method.

AutoDock can now be used as a standalone energy minimizer by using the command
Introduction

“do_local_only”. Thus the user can now minimize a structure using exactly the same force field
as is used in the dockings. This could be useful, for example, in minimizing a crystal structure to
relieve bad contacts.

The hybrid global-local search routine uses a library of portable routines for random number gen-
eration, based on the method of L’Ecuyer & Cote12. This is code is a transliteration of the original
Pascal carried out by the Department of Biomathematics, University of Texas. This random num-
ber generator (RNG) has the advantage of providing a set of random numbers that are hardware
independent. The simulated annealing (SA) algorithm still uses the built-in “drand48” function on
most platforms, but the drand48 implementation may vary from platform to platform.

Our results show that the Lamarckian GA, (also known as the hybrid GA-SW and GA-pSW meth-
ods) reproduce the crystal complex more reliably using the same number of energy evaluations
than SA does.

There are new keywords that have been added to AutoDock to assist in setting up a docking using
the new methods. Those keywords that pertain to the genetic algorithm are prefixed with the let-
ters “ga_” , those specific to local search have the prefix “ls_” and those specific to Solis and
Wets and pseudo-Solis and Wets have the prefix “sw_”. To use the GA, the “set_ga” directive
must be given Classical Solis and Wets needs “set_sw1”, and pseudo-Solis and Wets requires
“set_psw1”. In order to begin the GA, the keyword “ga_run” must be given along with a number
of runs to be executed.

In order to perform conformational cluster analysis after the dockings, the keyword “analysis”
must be supplied as the final line, otherwise no structural output will be generated.

A small change must now be made to old SA docking parameter files, with the addition of the
keywords “simanneal” and “analysis”. These instruct AutoDock respectively to begin the SA
docking, and when the requested number of runs have been carried out, perform cluster analysis.

12. P. L’Ecuyer and S. Cote. (1991) “Implementing a Random Number Package with Splitting Facilities”,
ACM Transactions on Mathematical Software, 17, 98-111

10
 Theory

Theory
5. Overview of the Free Energy Function

In version 3.0 of AutoGrid and AutoDock, we introduced a new kind of scoring function that is
used during and at the end of the dockings. It is based on the principles of QSAR (quantitative
structure-activity relationships) and was parameterized using a large number of protein-inhibitor
complexes for which both their structure and inhibition constants, or Ki, were known. The user is
encouraged to refer to the description of how this free energy function was derived in the original
literature3.

Theory
∆Gbinding,vacuo
E + I E I

+ Η2Ο ∆Gsolvation(E+I) + Η2Ο ∆Gsolvation(EI)

∆Gbinding,solution
E + I E I +

The above diagram shows the thermodynamic cycle for the binding of an enzyme, E, and an
inhibitor, I, in both the solvated phase and in vacuo. Note the solvent molecules are indicated by
filled circles: they tend to be ordered around the larger molecules, but when E and I bind, several
solvent molecules are liberated and become disordered. This is an entropic effect and is the basis

11
 Theory

of the hydrophobic effect. The solvent ordering around E and I, when both bound and unbound, is
strongly influenced by the hydrogen bonding between these molecules. These hydrogen bonds
between solvent and E, and solvent and I, contribute enthalpic stabilization, and is something we
can estimate in our new free energy function.

According to Hess’s law of heat summation, the change in free energy between two states will be
the same, no matter what the path. So we can calculate the free energy of binding in solvent by the
following equation:

∆Gbinding,solution = ∆Gbinding,vacuo + ∆Gsolvation(EI) - ∆Gsolvation(E+I)

Since we can calculate ∆Gbinding,vacuo from our docking simulation, and can estimate the free energy
change upon solvation for the separate molecules E and I, and for the complex, EI, ∆Gsolvation(EI)
and ∆Gsolvation(E+I) respectively, then it is also possible to calculate the free energy change upon
binding of the inhibitor to the enzyme in solution, ∆Gbinding,solv. Thus, we can estimate the inhibi-
Theory

tion constant, Ki, for the inhibitor, I.

A key point to bear in mind is that most parts of the new scoring function are essentially the same
as the original AutoDock scoring function used in versions prior to 3.0, except that various terms
in the molecular mechanics energy function have been re-scaled by new coefficients, and new
terms have been introduced. These new terms include the desolvation free energy of the ligand,
and an estimate of the loss of conformational degrees of freedom of the ligand upon binding.

The coefficients were derived using linear regression analysis, and we chose the linear regression
model that most closely fit the observed inhibition constant data. Thus the user should not modify
these coefficients lightly.

For the curious, these coefficients are defined in “gpf3gen.awk” and “dpf3gen.awk”, and their
variable names and values are as follows:

#
# Free energy model 140n coefficients:
#
FE_vdW_coeff = 0.1485
FE_estat_coeff = 0.1146
FE_hbond_coeff = 0.0656
FE_tors_coeff = 0.3113
FE_desol_coeff = 0.1711

So for example, in Table 2 on page 19 and Table 3 on page 21, the values used in the AutoDock
3.0 scoring function will be as given except the van der Waals coefficients and well depth ener-
gies, ε, will be scaled by 0.1485, the electrostatic energy will be scaled by 0.1146, and the hydro-
gen bonding terms will be scaled by 0.0656. The new terms for loss of torsional degrees of
freedom upon binding and the ligand desolvation free energy will be scaled by 0.3113 and 0.1711
respectively. The torsional term is actually the number of rotatable bonds in the ligand that rotate
heavy atoms, multiplied by the coefficient, 0.3113. Hydroxyl rotors, for example, are not counted.
This number is actually determined by AutoTors, and is written in the ligand PDBQ file after the

12
 Theory

new keyword, “TDOF”. This can also be set in the DPF, using the keyword “torsdof n 0.3113”,
where n is the number of heavy-atom rotatable bonds.

Note that the new desolvation free energy term is only calculated for aliphatic and aromatic car-
bon atoms in the ligand. We found that the quality of the final empirical free energy model was
not affected by the inclusion of the heteroatoms N and O in this term, so these are ignored for sim-
plicity and speed of calculation. This does mean that the ligand input PDBQ file must distinguish
between carbon atoms that are aliphatic and those that are aromatic. This is done by changing the
atom names of the aromatic carbons so that the initial ‘C’ is replaced by ‘A’. For example, if the
ligand happened to be a peptidomimetic inhibitor which contained a Phe-sidechain, then the atom
names in the PDBQ file would have to be changed from CG, CD1, CD2, CE1, CE2 and CZ, into
AG, AD1, AD2, AE1, AE2 and AZ, respectively. To help the user do this automatically, Auto-
Tors has a new option that can be invoked using the ‘-A’ flag on the command line. This will look
for all planar cyclic carbons and assume these are aromatic: it will then change their atom names
automatically. The user can also override the default angle AutoTors uses to determine planarity.
See the Section below on AutoTors.

Theory
This also means AutoGrid will calculate two different types of carbon grid maps, one for ali-
phatic carbons in the ligand (*.C.map), and one for aromatic carbons in the ligand (*.A.map).
These will be input by AutoDock for the docking calculations, and the internal energy parameters
must also specify the values for the aromatic and aliphatic carbon atoms.

Another point to bear in mind is that AutoGrid 3.0 can calculate smoothed pairwise potentials:
the lowest energy within a user-defined distance is stored at the current position. This has the
effect of ‘widening’ the basin of affinity. This smoothing distance is set using the keyword
‘smooth’ in the grid parameter input file to AutoGrid. It is very important that the value of
‘smooth’ is not changed from 0.5Å, since the free energy function was calibrated using this set-
ting. If smooth is set to 0.0Å, for example, the calculated free energies will probably be too high.

6. Grid Maps

AutoDock requires pre-calculated grid maps, one for each atom type present in the ligand being
docked. This helps to make the docking calculations extremely fast. These maps are calculated by
AutoGrid. A grid map consists of a three dimensional lattice of regularly spaced points, sur-
rounding (either entirely or partly) and centered on some region of interest of the macromolecule
under study. This could be a protein, enzyme, antibody, DNA, RNA or even a polymer or ionic
crystal. Typical grid point spacing varies from 0.2Å to 1.0Å, although the default is 0.375Å
(roughly a quarter of the length of a carbon-carbon single bond). Each point within the grid map
stores the potential energy of a ‘probe’ atom or functional group that is due to all the atoms in the
macromolecule.

The user must specify an even number of grid points in each dimension, nx, ny and nz. This is
because AutoGrid adds a central point, and AutoDock requires an odd number of grid points.
The probe’s energy at each grid point is determined by the set of parameters supplied for that par-

13
 Theory

ticular atom type, and is the summation over all atoms of the macromolecule, within a non-
bonded cutoff radius, of all pairwise interactions.

The following figure illustrates the main features of a grid map:

grid spacing /Å

grid point

ny+1
probe atom
Theory

nz+1

nx+1

The ligand can be seen in the centre of the grid map, buried inside the active site of the protein. In
this case, the grid map encompasses the whole protein. The grid spacing is the same in all three
dimensions.

As mentioned in the description of the new free energy function, the user can smooth the pairwise
potentials, by storing the lowest energy within a given distance of the current pairwise separation.
The value of this specified in the GPF, and should not be changed from ‘smooth 0.5’ in order to
use the function described in the literature3.

In addition, in AutoGrid 3.0, the user must use a new utility program to specify the atomic frag-
mental volume and atomic solvation parameters for each atom in the macromolecule, and for all
the carbon atoms in the ligand. This requires the assignment of these parameters to the macromol-
ecule using the program ‘addsol’ to create a PDBQS file. This resembles a PDB formatted file, but
in addition gives the partial charges and solvation parameters for each atom. The solvation param-
eters for the ligand ‘probe’ atoms are specified in the GPF, by the “sol_par” keyword, and should
not be modified unless the free energy function is not needed.

14
 Theory

One last addition to the GPF, is the introduction of a ‘constant’ keyword. This was introduced to
penalize hydrogen bonds lost upon ligand binding. This defines a constant energy that is added to
all the values in a grid map. The rationale for this is as follows: when a ligand goes from the
unbound to the bound state, and is capable of making hydrogen bonds, it may or may not lose the
enthalpic stabilization of one or more of these H-bonds. We assume that a ligand in the aqueous
phase accepts and donates as many hydrogen bonds as it can. However, we found that a plot of the
total hydrogen bonding energy for the bound ligand in the protein complex, versus the maximum
number of possible hydrogen bonds the ligand could form, indicated that on average only 36% of
the maximum well depth stabilization was achieved for each possible hydrogen bond. Thus a
ligand atom in the complex that has a hydrogen bonding capacity must experience at least this
amount of stabilization before it can be formed. Those that cannot are penalized by this amount.

7. Van der Waals Potential Energy

Theory
The pairwise potential energy, V(r), between two non-bonded atoms can be expressed as a func-
tion of internuclear separation, r, as follows,
– br
Ae C
V ( r ) = ------------ – -----66-
r r

Graphically, if reqm is the equilibrium internuclear separation, and ε is the well depth at reqm, then:

V(r)

repulsive, exchange energy +Ae-br/r

reqm

0 reqm
r

ε
attractive, dispersion energy -C6/r6

The exponential, repulsive, exchange energy is often approximated thus,

15
 Theory

A C 12
---e –br ≈ --------
12
r r

Hence pairwise-atomic interaction energies can be approximated using the following general
equation,
C Cm
V ( r ) ≈ -----n-n – ------
–n –m
m
= Cnr – Cmr
r r

where m and n are integers, and Cn and Cm are constants whose values depend on the depth of the
energy well and the equilibrium separation of the two atoms’ nuclei. Typically the 12-6 Lennard-
Jones parameters (n=12, m=6) are used to model the Van der Waals’ forces13 experienced between
Theory

two instantaneous dipoles. However, the 12-10 form of this expression (n=12, m=10) can be used
to model hydrogen bonds (see “Modeling Hydrogen Bonds” below). Appendix II gives the
parameters which were distributed with the first (FORTRAN-77) version of AutoDock, and
which have been used in numerous published articles.

A revised set of parameters has been calculated, which use the same Van der Waals radius of a
given atom for all pairwise distances, no matter what the other atom. Likewise, the well-depths
are consistently related. Let reqm, XX be the equilibrium separation between the nuclei of two like
atoms, X, and let εXX be their pairwise potential energy or well depth. The combining rules for the
Van der Waals radius, reqm, and the well depth, ε, for two different atoms X and Y, are:

r eqm,XY = 1--2- ( r eqm,XX + r eqm,YY )

ε XY = ε XX ε YY

A derivation for the Lennard-Jones potential sometimes seen in text books invokes the parameter,
σ, thus,
1
---
6
r eqm, XY = 2 σ
Then the Lennard-Jones 12-6 potential becomes:

σ σ 6
12
V 12-6 ( r ) = 4 ε XY --- –  ---
 r  r

Hence, the coefficients C12 and C6 are given by:

13. van der Waals, J. H. (1908) Lehrbuch der Thermodynamik, Mass and Van Suchtelen, Leipzig, Part 1

16
 Theory

C 12 = ε XY r eqm, XY
12

C 6 = 2 ε XY r eqm, XY
6

We can derive a general relationship between the coefficients, equilibrium separation and well
depth as follows. At the equilibrium separation, reqm, the potential energy is a minimumand equal
to the well depth: in other words, V(reqm) = -ε. The derivative of the potential with respect to sepa-
ration will be zero at the minimum potential:

dV nC n mC m
n+1
- + -----------
= – ---------- m+1
- = 0
dr r r

therefore:
nC n mC m

Theory
----------
n+1
- = -----------
m+1
-
r r
so:
m +1
nC n r n (m – n)
C m = ---------------------
- = ---- C n r
n+1 m
mr

Substituting Cm into the original equation for V(r), then at equilibrium we obtain,

(m – n)
C n nC n r eqm
– ε = -------
n
- – ----------------------
m
-
r eqm mr eqm

Rearranging:
m n (m – n)
C n  --------------------------------------------
mr eqm – nr eqm r eqm 
- = –ε
 mr r
n m 
eqm eqm

Therefore, the coefficient Cn can be expressed in terms of n, m, ε and reqm thus:

m
C n = -------------- εr eqm
n
n–m

and, substituting into original equation for V(r),

n
C m = -------------- εr eqm
m
n–m

17
 Theory

In summary, then, we obtain the general equation for any n, m:

m n
-------------- εr -------------- εr eqm
n m
n – m eqm n–m
V ( r ) ≈ -----------------------
n
- – -----------------------
m
-
r r

One final point worth making here is the effect of the ‘smooth 0.500’ command on the pairwise
potentials. This is set in the AutoGrid input file (also known as the ‘GPF’). This is best illustrated
with a diagram; note that this has the effect of widening the region of maximum affinity at ε, and
also reduces the potential energy at r=0 to a finite value:

V(r)
Theory

Standard van der Waals potential

(unsmoothed)

0 reqm
r

‘Smoothed’ van der Waals potential

ε

Example reqm and ε parameters for various AMBER atom types of carbon are shown in Table 1.

Table 1: AMBER parameters for carbon atom types.

reqm ε
AMBER atom type
_ --- / Å / kcal mol-1

C, C*, CA, CB, CC, CD, CE, CF, CG, CH, CI, CJ, CM, CN, CP 1.850 0.12

C2 1.925 0.12

C3 2.000 0.15

CH 1.850 0.09

CT 1.800 0.06

Using the equations describing C12 and C6 above, the following new set of 12-6 parameters were

18
 Theory

calculated shown in Table 2. These parameters may be used with AutoDock version 3.0, or alter-
natively, you may use or derive your own. Remember the linear regression coefficients for the van
der Waals term has not been applied to the parameters in this table.

Table 2: Self-consistent Lennard-Jones 12-6 parameters, before multiplication by the free energy model coefficients.

Atoms i-j reqm,ij εij C12 C6

----------/ Å / kcal mol-1 / kcal mol-1Å12 / kcal mol-1Å6

C-C 4.00 0.150 2516582.400 1228.800000

C-N 3.75 0.155 1198066.249 861.634784

C-O 3.60 0.173 820711.722 754.059521

C-S 4.00 0.173 2905899.052 1418.896022

C-H 3.00 0.055 29108.222 79.857949

N-C 3.75 0.155 1198066.249 861.634784

Theory
N-N 3.50 0.160 540675.281 588.245000

N-O 3.35 0.179 357365.541 505.677729

N-S 3.75 0.179 1383407.742 994.930149

N-H 2.75 0.057 10581.989 48.932922

O-C 3.60 0.173 820711.722 754.059521

O-N 3.35 0.179 357365.541 505.677729

O-O 3.20 0.200 230584.301 429.496730

O-S 3.60 0.200 947676.268 870.712934

O-H 2.60 0.063 6035.457 39.075098

S-C 4.00 0.173 2905899.052 1418.896022

S-N 3.75 0.179 1383407.742 994.930149

S-O 3.60 0.200 947676.268 870.712934

S-S 4.00 0.200 3355443.200 1638.400000

S-H 3.00 0.063 33611.280 92.212017

H-C 3.00 0.055 29108.222 79.857949

H-N 2.75 0.057 10581.989 48.932922

H-O 2.60 0.063 6035.457 39.075098

H-S 3.00 0.063 33611.280 92.212017

H-H 2.00 0.020 81.920 2.560000

The above parameters yield the following graphs, for C, N, O and H atom types; the curves in
order of increasing well-depth are: HH << CH < NH < OH << CC < CN < CO < NN < NO < OO:-

19
 Theory
Theory

Grid maps are required only for those atom types present in the ligand being docked. For example,
if the ligand being docked is a hydrocarbon, then only carbon and hydrogen grid maps would be
required. In practice, however, non-polar hydrogens would not be modeled explicitly, so just the
carbon grid map would be needed, for ‘united atom’ carbons. This saves both disk space and com-
putational time.

8. Modelling Hydrogen Bonds

Hydrogen bonds are frequently important in ligand binding. These interactions can be modeled
explicitly in AutoDock.

In order to save having two types of hydrogen grid maps, and thus conserve disk space, we nor-
mally use ligands with just one type of hydrogen, namely polar hydrogens. Polar hydrogens can
be defined here as those bonded to heteroatoms like nitrogen and oxygen, while non-polar hydro-
gens are bonded to carbon atoms.

If you want to model non-polar hydrogens as well, you would need a separate map for such

20
 Theory

hydrogens. You could use the atom type code ‘h’ for non-polar hydrogens, and ‘H’ for polar
hydrogens. Use 12-6 to distinguish non-polar hygrogens, and 12-10 for polar hydrogens.

The user must specify the appropriate 12-10 parameters in the AutoGrid parameter file, and on
the correct lines. Pairwise atomic interaction energy parameters are always given in blocks of 7
lines, in the order: C, N, O, S, H, X, M. X and M are “spare” atom types: If there were phosphorus
atoms in the receptor, X could be used as P. For example, to model donor hydrogens in the ligand,
12-10 parameters would be needed in the hydrogen parameter block, but only for H-bond accep-
tors, N,O and S (second, third and fourth lines in the H-parameters). The other parameters remain
as 12-6 Lennard-Jones values (C,H,X and M). In order to keep the symmetry of pairwise energet-
ics (H-O is the same as O-H), the user must specify 12-10 parameters for H (fifth line) in the N, O
and S-parameter blocks.

Table 3: Self-consistent Hydrogen bonding 12-10 parameters, before multiplication by the free energy model coefficients.

Atoms i-j reqm,ij εij C12 C10

Theory
----------/ Å / kcal mol-1 / kcal mol-1Å12 / kcal mol-1Å6

N-H 1.90 5.00 55332.873 18393.199

O-H 1.90 5.00 55332.873 18393.199

S-H 2.50 1.00 298023.224 57220.459

AutoGrid detects hydrogen bond parameters in the grid parameter file, if either n is not 12 or m is
not 6. If so, the pairwise interaction is modulated by a function of the cosine of the hydrogen bond
angle. This takes into account the directionality of hydrogen-bonds.

AutoGrid incorporates the angular dependence of the hydrogen bond potential. The ideal hydro-
gen bond would have an angle, θ, of 180˚ between the lone-pair of the acceptor atom, the polar
hydrogen and the donor atom, thus:

probe
θ macromolecule
O H N
donor
acceptor

As θ decreases, the strength of the hydrogen bond diminishes. There are no hydrogen bonds when
θ is 90˚ or less.

8.1 A note on atom type codes:

Note: If you use hydrogen bonding for nitrogen, you may need to distinguish between nitrogens
that can be acceptors and those that can be donors. The above settings for N and H would allow

21
 Theory

>NH to accept a hydrogen bond. To avoid this, such nitrogens should be treated as 12-6 non-
hydrogen bonders: used ‘n’ as the atom type code instead of ‘N’. This would mean, of course, an
extra grid map.

If you use polar and non-polar hydrogens, for example, with atom type codes of ‘H’ and ‘h’, you
must edit the atom names in the PDBQ files by hand. This would apply to different flavours of
nitrogen, ‘N’ for polar and ‘n’ for non-polar; or carbon, ‘C’ for aliphatic carbons and ‘A’ for aro-
matic carbons.

9. Electrostatic Potential Grid Maps

In addition to the atomic affinity grid maps, AutoDock requires an electrostatic potential grid
map. Polar hydrogens must be added, if hydrogen-bonds are being modeled explicitly. Partial
atomic charges must be assigned to the macromolecule. The electrostatic grid can be generated by
Theory

AutoGrid, or by other programs such as MEAD14 or DELPHI15, which solve the linearized Pois-
son-Boltzmann equation. AutoGrid calculates Coulombic interactions between the macromole-
cule and a probe of charge e, +1.60219x10-19 C; there is no distance cutoff used for electrostatic
interactions. A sigmoidal distance-dependent dielectric function is used to model solvent screen-
ing, based on the work of Mehler and Solmajer16,
B
ε(r) = A + -------------------------
– λBr
-
1 + ke

where: B = ε0 - A; ε0 = the dielectric constant of bulk water at 25˚C = 78.4; A = -8.5525, λ =

0.003627 and k = 7.7839 are parameters.

Charges must be stored in PDBQ format in order for AutoGrid to read them. PDBQ is an aug-
mented form of the standard PDB format, in which an extra column is used to store the partial
atomic charges (hence the “Q” in “PDBQ”). Columns 71-76 of the PDB file hold the partial
atomic charge (the older form of PDBQ contains charges in columns 55-61).

Charges can be assigned using a molecular modeling program. Unix shell scripts are provided to
convert from Insight9517 “.car” files (“cartopdbq”) and SYBYL18 “.mol2” files

14. Bashford, D. and Gerwert, K. (1992) “Electrostatic calculations of the pKa values of ionizable groups in
bacteriorhodopsin”, J. Mol. Biol., 224, 473-486; Bashford, D. and Karplus, M. (1990) “pKas of ionizable
groups in proteins - atomic detail from a continuum electrostatic model.”, Biochemistry, 29, 10219-10225;
MEAD is available from Donald E. Bashford, Dept. Molecular Biology, Mail Drop MB1, The Scripps
Research Institute, 10666 North Torrey Pines Road, La Jolla, CA 92037.
15. Gilson, M.K. and Honig, B. (1987) Nature, 330, 84-86; DELPHI is available from Biosym Technologies,
9685 Scranton Road, San Diego, CA 92121-2777, USA.
16. Mehler, E.L. and Solmajer, T. (1991) “Electrostatic effects in proteins: comparison of dielectric and
charge models” Protein Engineering, 4, 903-910.
17. Biosym/MSI, 9685 Scranton Road, San Diego, California 92121-3752, USA.
18. Tripos Associates, Inc., 1699 South Hanley Road, Suite 303, St. Louis, Missouri 63144-2913, USA.

22
 Theory

(“mol2topdbq”). See also “q.amber” and “q.kollua”, in the appendices.

Theory

23
 Methodology

Methodology
10. Getting Started...

This section describes very quickly the method for setting up a docking using the AutoDock pro-
grams. You should find all these utilities under the “share” and “bin” directories. Before you
start, add these two lines to your .cshrc: “setenv AUTODOCK_UTI /path/to/the/directory/
share” and “set path=($path $AUTODOCK_UTI)”. Make sure you “source .cshrc” also.

(1) The macromolecule first needs polar hydrogens to be added and then partial atomic
charges to be assigned. This can be done efficiently in SYBYL, e.g., using the
“Biopolymer” menu, adding “Essential_Only” hydrogens and assigning “KOLLUA”
partial charges to the protein. Create the PDBQS file19 for the macromolecule. Save
the protein in “mol2” format, and then convert into PDBQS format using
“mol2topdbqs”. This also assigns atomic solvation parameters and creates
“macro.pdbqs”:

% mol2topdbqs macro.mol2

(2) If you already have a PDBQ-formatted version of your macromolecule, say

“macro.pdbq”, you must assign the atomic solvation parameters to it. The “addsol”
program will input “macro.pdbq” and output a PDBQS file, “macro.pdbqs”:
Methodology

% addsol macro.pdbq macro.pdbqs

(3) Create the ligand PDBQ file20 using “deftors”21, to define any torsions that you want to
be explored during the docking. (Label the ligand with “Atom ID” or atom serial
numbers in a molecular viewer. This will help in assigning the atoms):

% deftors lig.mol2

(4) Create the GPF (grid parameter file) and the DPF (docking parameter file).

% mkgpf3 lig.pdbq macro.pdbqs

% mkdpf3 lig.pdbq macro.pdbqs

These create files with names derived from the ligand and macromolecule files,

19. This contains the PDB records in addition to the partial atomic charges and atomic solvation parameters.
20. This contains the root atoms and the branches and torsions defining the rotatable bonds in the ligand, as
well as the partial atomic charges.
21. The script deftors uses the program AutoTors to assign root atoms and torsions.

24
 Methodology

namely “macro.gpf” and “lig.macro.dpf”22.

(5) Edit the GPF and then use AutoGrid to calculate the grid maps.

% autogrid3 -p macro.gpf -l macro.glg &

(6) Edit the DPF and then perform the dockings using AutoDock.

% autodock3 -p lig.macro.dpf -l lig.macro.dlg &

(7) To view docking results in a molecular modelling program, use “get-docked”, to cre-
ate a PDB formatted file. It will be called “lig.macro.dlg.pdb” and will contain all
the docked conformations output by AutoDock in the “lig.macro.dlg” file.

% get-docked lig.macro.dlg

Or to view in AVS, use “mkdlgfld”,”mkatmtypfld” and “mkbndfld”.

(8) You don’t need to do this step. But if you are interested, you can calculate the energy
of a given ligand conformation in the crystal structure you used to calculate the maps:

% autodock3 -p lig.macro.dpf -l lig.macro.epdb.log -c <

lig.macro.epdb.com

where the AutoDock command file “lig.macro.epdb.com” contains the two com-
mands, “epdb lig.pdbq” and “stop” on separate lines.

Methodology
There are several Unix shell scripts and “awk” programs to help set up default parameter files for
AutoGrid and AutoDock. They are described in more detail in the Appendix. The user must
check their input “gpf” and “dpf” files, to ensure the defaults look reasonable. The user can adjust
the default parameters using a text editor like “vi” or “emacs”. These parameters are described in
the sections “AutoGrid Parameter File Format” and “AutoDock Parameter File Format”, in the
appendices.

11. Setting Up AutoGrid and AutoDock Jobs

Let us suppose that the user wishes to test AutoDock by trying to reproduce an x-ray crystallo-
graphic structure of a ligand-enzyme complex taken from the Brookhaven Protein Data Bank. The
first step is to split the desired PDB file into two separate PDB files, one containing all the heavy
atoms of the enzyme, the other containing those of the ligand. Both files should retain the exten-

22. The stems differ because the grid parameter file is specific to the macromolecule only, but the docking
parameter file is specific to both the ligand and the macromolecule. Therefore, try and keep the ligand and
macromol filename stems short.

25
 Methodology

sion ‘.pdb’.

Note: Care should be taken when the PDB file contains disordered residues, where alternate loca-
tion indicators (column 17) have been assigned. For each such atom, the user must select only one
of the possible alternate locations (preferably that with the highest occupancy value).

We will discuss in the next sections, the steps needed to prepare the parameter files for AutoGrid
and AutoDock. If desired, the user may specify rotatable bonds in the ligand (receptor flexibility
is not allowed). To help this definition, there is a program called AutoTors. This utility interac-
tively queries the user about the rigid portion of the molecule (the “root”) and rotatable torsions
(the “branches” and “torsions”). Then it outputs the ligand in PDBQ format for AutoDock. It can
even process partial charges on the hydrogens to create a polar-hydrogen only version of the
ligand. This will be discussed in greater detail below.

12. Preparing the Ligand

Initially you must add hydrogens to all atoms in the ligand, ensuring their valences are completed.
This can be done using a molecular modeling package. Make sure that the atom types are correct
before adding hydrogens. You may want to specify the pH, depending on whether charged or neu-
tral carboxylates and amides are desired.

Next, assign partial atomic charges to the molecule. AMPAC or MOPAC can be used to generate
partial atomic charges for the ligand. These charges must be written out in PDBQ format, which
has the same columns as a Brookhaven PDB format, but with an added column of partial atomic
Methodology

charges (see ‘cartopdbq’ and ‘mol2topdbq’ in Appendix I).

13. Ligand Flexibility and Constraints

To allow flexibility in the ligand, it is necessary to assign the rotatable bonds. It is a good idea to
have handy a plot of the ligand, labelled by atom name, and a second labelled by atom serial num-
ber (atom ID). AutoDock can handle up to MAX_TORS rotatable bonds: this parameter is defined
in “autodock.h”, and is ordinarily set to 32. If this value is changed, AutoDock must be recom-
piled.

Torsions are defined in the PDBQ file using the following tokens or keywords:

ROOT / ENDROOT
BRANCH / ENDBRANCH
TORSION / ENDTORSION

These keywords use the metaphor of a tree. See the diagram below for an example. The “root” is

26
 Methodology

defined as the fixed portion of the ligand, from which rotatable ‘branches’ sprout. Branches within
branches are possible, and torsions are a special case of branches, where the two atoms at either
end of the rotatable bond have only two nearest neighbors (unlike branches which can have three
or more). Nested rotatable bonds are rotated in order from the “leaves” to the “root”.

O2
NCH H2

BRA TORS
C10
C8 O3
C6 C9
ROOT C5

C7
C4 H1
O1
H21
C2 C3
N2
C1
H22 +
N1 H11

H12

The PDBQ keywords must be carefully placed, and the order of the ATOM or HETATM records
may need to be changed in order to fit into the correct branches. The PDBQ keywords can be
abbreviated to no less than the first 4 letters. To assist the user in placing these keywords correctly,

Methodology
and in re-ordering the ATOM or HETATM records in the ligand PDBQ file, it is best to use the
interactive program AutoTors (see below).

Note: AutoTors, AutoGrid and AutoDock do not recognize PDB “CONECT” records, neither do
they output them.

“CONSTRAIN” defines a single, optional distance constraint, between two flexible parts of the
ligand. It is not normally used in docking. This retains only those conformations where this dis-
tance is within a certain range of values. In docking, a conformation which violates this constraint
is instantly rejected; it does not increment the rejections-counter in simulated annealing, its
energy is not evaluated, nor is the steps-counter incremented. This PDBQ keyword has the follow-
ing syntax:

CONSTRAIN atom1 atom2 lower upper

The first two parameters are the atom serial numbers of the two atoms to be constrained, and the
last two are the lower and upper bounds for this distance, in Angstroms. This can be particularly
useful when docking say two proteins: a loop from one protein can be cut out and the ends con-
strained to have roughly the same value as in the original protein.

27
 Methodology

The next sections describe the input files needed for AutoTors, and how to run it.

14. Using AutoTors to Define Torsions in the Ligand

This section describes input and output files used and generated by AutoTors. Input consists of
one or two files, depending on whether the ligand is in our “AutoDock-standard” PDBQ-format,
or in Sybyl’s mol2-format. PDBQ-format is the default; mol2-format is allowed with the “-m”
flag (see below).

14.1 Ligand is in PDBQ-format:

When the ligand is in PDBQ format, AutoTors also needs a “bnd” or bond file, which describes
the connectivity of the atoms in the ligand. In this example, the bond file is “oligo.bnd”, and
“oligo.pdbq” is the input PDBQ file; “oligo.out.pdbq” is created and contains all the ROOT,
BRANCH and TORS keywords needed to define the torsions selected by the user.

% autotors oligo.bnd oligo.pdbq oligo.out.pdbq

The “.bnd” file, contains information about the covalent bonds in the ligand. The bonds are
described by the serial numbers of the atoms in the input PDBQ file, with one line per bond. For
example, if C10 is the atom appearing on the first “ATOM” line in the PDBQ file, and it is bonded
to N18 which appears on the 17th line in the PDBQ file, this information appears as a discrete line
in the “.bnd” file as: “1 17”. The output of “pdbtoatm” is an “atm” file, which can be converted to
Methodology

a “bnd” file, using “atmtobnd”. For example, to generate a “bnd” file, use something like this
command:

% pdbtoatm vitc.pdbq | atmtobnd > vitc.bnd

14.2 Ligand is in Mol2-format:

When the ligand is in SYBYL-mol2 format, no “bnd” bond file is required, in addition to -m flag.
This is because the mol2 file contains both atom coordinates and bonding information. So, for
example the following command would read in the “lead.mol2” file and after interactively
requesting which torsions to rotate, AutoTors would write out “lead.out.pdbq”:

% autotors -m lead.mol2 lead.out.pdbq

14.3 AutoTors Output:

The output filename is defined by the last AutoTors command-line argument. Output consists of
PDBQ-formatted lines, rearranged as required by AutoDock, according to the user’s specification
of the fixed ROOT portion of the molecule, and the allowed rotatable bonds in the rest of the mol-

28
 Methodology

ecule. AutoTors inserts the ROOT, ENDROOT, BRANCH, ENDBRANCH, TORS, and END-
TORS lines in the necessary places.

14.4 AutoTors Flags:

-m <input_ligand_mol2_file>

This flag is used when the input file is in Tripos ’mol2’ format (produced by SYBYL). When
it is entered on the command line, the program uses only 1 file for both kinds of input (the bond
data input file and the pdbq data input file) and uses the second file specified for output. [If the
user runs the program with the -m flag AND three file parameters on the command line, the first
file will be opened for reading the input needed by the program, the second opened for writing
and the third ignored. This means any contents in the second file will get over-written and lost.]

-h

This flag causes the program to detect non-polar hydrogens, that is hydrogen atoms bonded to
carbon atoms, to merge the charge of each with the charge of the carbon to which it is bonded and
to delete the line of output data pertaining to that hydrogen. At the end of the program, a count of
the number of non-polar hydrogens which have been merged in this fashion is written to the
screen.

-o

This flag is used only in conjunction with the -h flag when the pdbq data is in the older pdbq
format. It causes the program to obtain charge data from column 55 instead of column 70. (Its use

Methodology
along with the -m flag is an error but this is disregarded.)

-a

This flag instructs AutoTors to disallow torsion rotations in amide and peptide bonds, (C=O)-
-(NH).

-b

This flag is useful for peptides. It disallows rotations in backbone torsions, including phi, psi
and omega (peptide) torsions.

-c

This will add atom connectivity to the ATOM records in the output pdbq file.

-e

This instructs AutoTors to use the atom types given in the mol2file. This can only be used
with the ’-m’ mol2-format flag.

29
 Methodology

-r

This sets the ROOT to be the non-hydrogen atom closest to the center of the molecule.

-M

This instructs AutoTors to use the ROTATABLE_BOND and ANCHOR information in a Tri-
pos SYBYL mol2 formatted file, to define the ROOT and active torsions.

-A
-A +<angle>

This flag causes the program to check rings for aromaticity. If all the ring atoms are ’co-pla-
nar’ enough, the program replaces ’C’ by ’A’ for all carbons in the ring. This distinction is neces-
sary in the AutoDock 3.0 Force Field computations. By default, the test for planarity is whether
the angle between two adjacent atoms’ normal vectors is less than or equal to 7.5 degrees.

The user can specify what cut-off to use (and thus override the default of 7.5 degrees) by typ-
ing in a different angle after the -A flag. The angle should be given in degrees. Note: this number
must be preceeded by a plus sign, ’+’. For example:

% autotors -A +6.2 myfile.bnd myfile.pdbq myfile.out

would cause autotors to use 6.2 degrees for this aromaticity cut-off angle between adjacent atoms.
This depends on how ’warped’ the ring is: some crystal structures can have aromatic rings that are
quite distorted from planarity.
Methodology

The -m, -h and -a flag may appear in any order. The -o flag must be given after the -h flag.
Placement of these flags should follow these two examples. Square brackets denote optional flags:

% autotors [-h][-o][-a][-c] pept.bnd pept.pdbq pept.out.pdbq

For SYBYL-mol2 input, e.g.:

% autotors -m [-h][-o][-a][-c] drug.mol2 drug.out.pdbq

15. Running AutoTors

MAX_TORS: AutoDock is set up to allow a maximum number of torsions. If AutoTors detects

more torsions than are permitted, a warning to that effect is given and it is up to the user to reduce
the number of torsions, either by deleting or selecting the appropriate number of torsions.
MAX_TORS is defined in the file “autodock.h”; if this definition is changed, the autodock-
executable must be re-made, using the appropriate Makefile.

30
 Methodology

There are four stages in running AutoTors:

15.1 Input Stage

(a) Data about bonds in the molecule are used to construct a tree -like structure. Each line of
bond data consists of two integers corresponding to the line numbers (in the pdbq file) of the
atoms involved. These integers are used as the ’id’s of the atoms in the molecule. Once the new
ids are read in, the pre-existing TREE is searched for ATOM_NODEs with either of these ids. If
only one such node is found, an ATOM_NODE is created for the other id and linked to the
already-entered node in the appropriate way. If the id of the pre-existing node was the first of the
two integers on the line of bond data, one of the node’s next links is set to the new node and one of
the new node’s prev links is set to the pre-existing node. In the other case, the opposite linking
pattern is set. If neither id can be found in the member ids of ATOM_NODEs in the TREE, two
new ATOM_NODEs are created, linked together in the appropriate way and held in a temporary
data structure for later linking into the ’TREE’ in phase b. If both of the ids are already in the
TREE, the two nodes with these ids are linked as appropriate. Moreover, this case signals the
detection of a cycle and the existing TREE is processed to detect the members of this cycle and to
store resulting information about the new cycle. The members of new cycle i are stored in two
dimensional global array cycle[i]. The number of members of cycle i is stored in cycle_size[i].

(b) After all the bond data is entered, the program attempts to attach any pairs which had not
been linked to pre-existing atoms. To do this, it searchs through the TREE AT MOST once for
each id in input data, attempting to add each unattached pair to the TREE. If any unattached pair
remains after this process, the input data is flawed and the program exits early with an error mes-
sage to that effect.

Methodology
15.2 Root Specification Stage

After all the nodes are created and connected, in whatever order and direction the bond data spec-
ifies, the user interacts with the program to select the portion of the molecule to be considered the
’ROOT.’ This is the section of the molecule which will remain rigid and NOT undergo any tor-
sions. This phase has two parts:

(a) Cycles detected are listed on the screen. The user either selects one of these cycles to be the
root (by entering the appropriate number) OR selects none of them as the root (by entering ’0’).

(b) The user can modify the list of root atoms at this point by adding atoms to the rootlist (by
entering the atom ’id.’). The user leaves this phase by entering ’q’ (to quit).

If no root atoms are specified, a message to this effect is written to the screen and the program
will exit at this point.

15.3 Torsions Detection and Selection Stage

Once at least one root atom has been designated, the program next processes the TREE, changing

31
 Methodology

the direction of the links in the TREE as appropriate (so that all links previous to the root are pre-
vious and all nexts are nexts), and accumulating a list of possible torsions which the user edits
interactively.

(a) The TREE is traversed in a depth-first order traversal at this point detecting possible tor-
sions. Torsions cannot occur between root atoms NOR between atoms in a cycle. Moreover, tor-
sions are not permitted between atoms and their ’leaves’ (attached atoms which have no other
connection). In the case of the user-specified ’-a’ flag, amide bond torsions are not permitted. Dur-
ing this traversal, a linked list of possible torsions is built.

(b) The user is given the list of the torsions detected in the molecule and has an opportunity to
modify this list. Torsions can be deleted OR selected at this point.(Depending on whether only a
few are to be deleted or if only a few are to be selected). The user leaves this phase by entering ’q’
(for quit).

15.4 Root Expansion and Output Stage

The rest of program follows at this point with no further input from the user. The TREE is tra-
versed again and the rootlist is expanded to include any atoms which are between the existing
root section and ’BRANCH’es as defined by active torsions. Next, the TREE is traversed and
new_id numbers are assigned sequentially to the atoms in the TREE, starting with the root.
Finally, a last traversal through the TREE is made and output is written to the file specified by the
user on the command line. ’REMARK’ lines are written first describing each possible torsion and
its status at the end of the program. Next, the expanded list of root atoms is output preceeded by
a’ROOT’ line and followed by an ’ENDROOT’ line. Last, the rest of atoms in the molecule are
output with appropriate ’BRANCH’, ’TORS’, ’ENDBRANCH’ and ’ENDTORS’ lines inserted
Methodology

as dictated by active torsions.

16. Adding Polar Hydrogens to the Macromolecule

When modeling hydrogen bonds explicitly, it is necessary to add polar hydrogens to the macro-
molecule. Then the appropriate partial atomic charges must be assigned. This can be achieved by
the user’s preferred method, e.g. using InsightII, Quanta, Sybyl, AMBER or CHARMm. Alter-
natively, one of the shell scripts described in the Appendix can be used. The charged macromole-
cule must be converted to PDBQS format so that AutoGrid can read it.

Note that most modeling systems add polar hydrogens in a default orientation, typically assuming
each new torsion angle is 0˚ or 180˚. Without some form of refinement, this can lead to spurious
locations for hydrogen-bonds. One option is to relax the hydrogens and perform a molecular
mechanics minimization on the structure. Another is to use a program like “pol_h” which takes as
input the default-added polar hydrogen structure, samples favorable locations for each movable
proton, and selects the best position for each. This “intelligent” placement of movable polar
hydrogens can be particularly important for tyrosines, serines and threonines.

32
 Methodology

17. Running AutoGrid

AutoGrid requires an input grid parameter file, which usually has the extension “.gpf”. The com-
mand is issued as follows:

% autogrid3 -p macro.gpf -l macro.glg &

where ‘-p macro.gpf’ specifies the grid parameter file, and ‘-l macro.glg’ the log file output during
the grid calculation. The ‘&’ ensures that the this job will be run in the background. This whole
line can be prefixed with the ‘nice’ command to ensure other processes are not unduly affected.
The log file will inform the user of the maximum and minimum energies found during the grid
calculations.

AutoGrid writes out the grid maps in ASCII form, for readability and portability; AutoDock
expects ASCII format grid maps. For a description of the format of the grid map files, see the
appendices.

Check the minimum and maximum energies in each grid map: these are reported at the end of the
AutoGrid log file (here, it is “macro.glg”). Minimum van der Waals’ energies and hydrogen
bonding energies are typically -10 to -1 kcal/mol, while maximum van der Waals’ energies are
around +105 kcal/mol. Electrostatic potentials tend to range from around -103 to +103 kcal/mol: if
these are both 0, this is a fairly clear indication that there are no partial charges on the macromol-
ecule.

Methodology
As well as the grid maps, AutoGrid creates two AVS-readable files, with the extensions ‘.fld’, and
‘.xyz’. The former is a field file summarizing the grid maps, and the latter describes the spatial
extent of the grids in Cartesian space. (To read the grid maps into AVS, use a “read field” mod-
ule.)

The ‘-o’ flag can be used on the AutoGrid command line to signify that the ‘.pdbq’ file specified
in the grid parameter file is in ‘old’ PDBQ format (charges are stored in columns 55-61).

18. Flexible Docking with AutoDock

As already described in the Introduction, AutoDock can use Monte Carlo simulated annealing
(SA), a genetic algorithm (GA), a hybrid genetic algorithm-local search (LGA), an evolutionary
programming (EP) or a pure local search (SW or pSW) engine in order to explore the conforma-
tional states of a flexible ligand.

Quaternion rotations23 have been implemented in handling the rigid body orientation of the

23. Shoemake, K. (1985) “Animating Rotation with Quaternion Curves” SIGGRAPH ‘85, 19, 245-254.

33
 Methodology

ligand. It was found that this gave finer control over the movement of the ligand, and gave better
docked solutions than with the alternative Eulerian rotations. Quaternions also avoid the gimbal
lock problem that Eulerian angles suffer from.

A docking “job” is a single AutoDock process, which carries out a number of independent dock-
ing “runs”, each of which begins with the same initial conditions. A simnulated annealing (SA)
run is a sequence of constant temperature annealing cycles. A genetic algorithm (GA, LGA or EP)
run consists of a series of generations. Each job can be seeded with a user-defined or a time-
dependent random-number generator seed. If time-dependent seeds are requested, this value is
updated each time a run starts, so 10 runs in one job get 10 different seed values.

The various parameters for the docking are usually stored in a docking parameter file, or “DPF”.
This is passed to AutoDock using a command line flag (-p). These flags will be discussed in
greater detail later on. It is advisable to do a short run to check the DPF, before committing to
spending billions of computer cycles. If there is any problem, a short run should find it.

Whatever search engine is chosen, the DPF must define the following: the random number gener-
ator seed or seeds using “seed”; the atom “types” in the ligand, that match the grid maps pro-
duced by AutoGrid; the “fld” field file that describes the spatial extents of the grids; and the
names of the “map” files themselves. AutoDock must be told what filename contains the ligand to
“move”, and “about” which x,y,z coordinate the rotations and translations will be centered. The
x,y,z values used in the “about” command must be in the same coordinate frame as the coordi-
nates in the ligand PDBQ file specified in the “move” command.

Currently in AutoDock 3.0, the intial state of the ligand can only be set using SA. All evolution-
ary search methods, GA, LGA and EP, automatically start with a random population. It is not pos-
Methodology

sible to seed a population with user-defined individuals in version 3.0

The initial translation and quaternion of this ligand may be set in SA dockings only, using the
“tran0” and “quat0” keywords.

The step sizes for making changes to the state variables affect SA and the evolutionary methods,
GA, LGA and EP. They are defined using the “tstep”, “qstep” and “dstep” keywords. The
default values are: translation, 0.2 Å, rigid-body orientation and dihedral angles, 5˚.

If the ligand is conformationally flexible, the user may specify, for SA only, the number and initial
values of the initial dihedral angles using “ndihe” and “dihe0”. If the keyword “random” is given
instead of explicit values, the ligand starts the SA with a random conformation.

The internal non-bonded potential parameters are defiend using the “intnbp_coeffs” or
“intnbp_r_eps” keywords. The former accepts coefficients while the latter accepts equilibrium
separations in angstroms and well depths in kcal/mol. The latter input method is more intuitive.

The user should specify the level of output during dockings, using “outlev”. Essentially, the
higher this integer, the more output is generated. A value of 1 is normally used.

34
 Methodology

If the user gives the “analysis” command, then after all the docking runs are completed in a
given job, cluster analysis or ‘structure binning’ will be performed. This is based on positional
root mean square deviation of corresponding atoms, ranking the resulting families of docked con-
formations in order of increasing energy. AutoDock writes out a histogram showing the number
of conformations in each cluster, and represents it ‘graphically’ using a bar chart of ‘#’ symbols.
Search the AutoDock log file for the phrase ‘HISTOGRAM’ all in upper-case, and you will see the
cluster analysis results.

The default method for structure binning allows for symmetry rotations. For example, a tertiary
butyl can be rotated by +/-120˚ and it will be chemically equivalent to the original conformation.
In other cases it may be desirable to bypass this similar atom type checking and calculate the rms
on a one-for-one basis: this can be done using the “rmsnosym” keyword. When clustering the con-
formations, the root mean square deviation tolerance “rmstol” and reference structure “rmsref”
filename should be specified. Typical values for rmstol range from 0.5 to 1.5 Å.

AutoDock’s analysis tool compares all the docked conformations with one-another, and if two
conformations have an rmsd that is less than the rmstol value, they are both stored in the same
cluster. This is repeated for all conformations, and the clusters are output ranked in order of
increasing energy from most negative to most positive. To perform the cluster analysis, the key-
word “analysis” must be given after the dockings have finished, on the last line of the DPF. It uses
the ‘rmstol’, ‘rmsref ’ and ‘rmsnosym’ commands set earlier in the DPF.

The next sections describe the parameters specific to the different search engines.

19. Monte Carlo Simulated Annealing

Methodology
During each constant temperature cycle of Monte Carlo simulated annealing, random changes are
made to the ligand’s current position, orientation, and conformation, if flexibile. The new state is
then compared to its predecesso. If its new energy is lower than the previous, this new state is
immediately accepted. However, if the new state’s energy is higher than the last, it is accepted
probabilistically. This probability depends upon the energy and cycle temperature (see the first
equation in Section 2). Generally speaking, at high temperatures, many states will be accepted,
while at low temperatures, the majority of these probabilistic moves will be rejected.

The user can choose whether to select the minimum energy state found during a cycle to be used
as the initial state for the next cycle, or the last state. The best docking results tend to be achieved
by selecting the minimum energy state from the previous cycle.

The initial annealing temperature “rt0” should be of the order of the average ∆E found during the
first cycle. This ensures that the ratio of accepted to rejected steps is high at the start. A typical
automated docking job may have an initial annealing temperature “rt0” of 500 (depending on the
system’s average ∆E) and a temperature reduction factor “rtrf” of 0.85-0.95 /cycle. Gradual
cooling is recommended, to avoid “simulated quenching”, which tends to trap systems in local
minima.

35
 Methodology

Depending on the degree of complexity of the problem, a relatively good search is given by 50
Monte Carlo “cycles”, and a maximum of 30,000 steps rejected “rejs” or 30,000 steps accepted
“accs”. 10 “runs” may or may not give a range of possible binding modes,. Multiple runs also
give relative energies. A schedule of 100 runs, 50 cycles, 3,000 steps accepted, 3,000 steps
rejected will provide more highly populated clusters, hinting at the ‘density of states’ for a given
conformation. A short test job would be: 1 run, 50 cycles, 100 accepted, 100 rejected steps.

The user must specify the maximum step a state variable can make in one step. Furthermore, these
can be adjusted during Monte Carlo simulated annealing, if a reduction factor (a fraction from 0
to less than 1) for translations and rotations is given. At the start of each cycle, the range from the
previous cycle is multiplied by this constant to give the new range, for translational and angular
displacements.

If desired, the states can be sampled during a docking and output to a trajectory file. This file con-
tains all the state variables required to define each sampled conformation, position and orientation
of the ligand. The user can specify the range of cycles to be sampled. This allows the selection of
the last few cycles when the docking will be nearing the final docked conformation, or the selec-
tion of the whole run.

20. Genetic Algorithm and Evolutionary Programming Docking

Since the new search methods were implemented in an object-oriented fashion, there is a new way
of specifying the parameters, that the user should be aware of. All the relevant parameters should
be specified first. Then, in order to use the genetic algorithm, the user must set up a global opti-
Methodology

mizer object using the “set_ga” command. Otherwise, if this set_ga command is given before
the “ga_*” parameters are specified, AutoDock will ignore these GA parameters and use the
default GA object, which has default parameters built-in.

Both the GA and LGA begin with a population of random ligand conformations in random orien-
tations and at random translations. The user must decide the number of individuals in the popula-
tion, using “ga_pop_size”: we have typically found 50 to be a good value. AutoDock counts the
number of energy evaluations and the number of generations as the docking run proceeds: the run
terminates if either limit is reached (“ga_num_evals” and “ga_num_generations” respectively).
The user can set the number of the best individuals in the current population that automatically
survive into the next generation, using “ga_elitism”: typically this is 1. The user can specify the
rate of gene mutation using “ga_mutation_rate” and the rate of gene crossover
“ga_crossover_rate”; typically these are 0.02 and 0.80 respectively, although setting
“ga_crossover_rate” to 0.00 reduces the genetic algortihm (GA) to an evolutionary program-
ming (EP) method. If the EP approach is used, you should also use an increased mutation rate to
ensure a good exploration of the search space. The number of generations for picking the worst
individual is set by “ga_window_size” and is usually 10.

If the user wants to perform a local search (LS), and for the Larmarckian GA (LGA), the user
must specify the local search parameters first (ls_*), and then set them (set_sw1 or set_psw1).

36
 Methodology

The maximum number of local search iterations is set by “sw_max_its”: this is typically about
300. The maximum number of consecutive successes or failures are both typically 4, and should
be set by “sw_max_succ” and “sw_max_fail” respectively. The size of the local search space to
sample is set by “sw_rho” and is usually 1.0. The lower bound on rho, “sw_lb_rho”, sets the
smallest step size that a move can make before terminating the local search, and is usually 0.01.
The probability that an individual in the population will experience local search is set by
“ls_search_freq”, and is typically about 0.07.

After specifying the local search parameters using the “ls_*” keywords, the user must set up a
local optimizer object using “set_sw1” or “set_psw1”, for Solis and Wets or pseudo Solis and
Wets. The former is the standard implementation of the local search, while the latter allows the
variances which control a step’s size to differ from gene to gene. This latter method, pseudo-SW,
is preferable in docking, since a 1 Å-step in translational space is small in comparison to a 1
radian-step in rotation space. The pseudo-SW local search takes its cue about the relative sizes of
the translational, orientational and torsional step sizes from the tstep, qstep and dstep values
set earlier in the AutoDock input parameter file.

Having set all these parameters and made all these choices, the user must tell AutoDock what to
do.

To perform a number of local searches (or “energy minimizations”) the user should put this line
into the DPF: “do_local_only 50”, where the number after the command is the number of local
search dockings to perform.

To carry out simulated annealing dockings, the command “simanneal” should be given; this will
perform the number of runs set by the “runs 10” command, which here would be 10 runs.

Methodology
To do a number of standard genetic algorithm (GA) dockings, give the “ga_run 10“ command,
but do not use the “set_sw1” or “set_psw1” commands in the same DPF. In this example,
AutoDock would do 10 GA dockings.

To use the Lamarckian genetic algorithm (LGA) in dockings, you must still use the “ga_run 10”
command, but you must have specified either the “set_sw1” or “set_psw1” command in one of
the preceding lines of the DPF.

Finally, after the docking command, you will almost certainly want to perform cluster analysis on
your search results. Give the ’analysis’ command, and after the last docking run is completed,
AutoDock will perform conformational clustering and then output a histogram ranked by increas-
ing energy.

21. Running AutoDock

Once the grid maps have been prepared by AutoGrid and the docking parameter file, or DPF, is
ready, the user is ready to run an AutoDock job. A docking is started from the command line

37
 Methodology

using the following command:

% autodock3 [-o][-k][-i][-u][-t] -p lig.macro.dpf [-l

lig.macro.dlg] &

Input parameters are specified by “-p lig.macro.dpf”, and the log file containing the output and
results from the docking is defined by “-l lig.macro.dlg”. This is the normal usage of
AutoDock, and performs a standard docking calculation.

-o

This can be added to the command line, to signify that the input file specified in the docking
parameter file is in old PDBQ format, with charges in columns 55-61.

-k

keep the original residue number of the input ligand PDBQ file. Normally AutoDock re-numbers
the starting position to residue-number 0, and any cluster-representatives are numbered incremen-
tally from 1, according to their rank (rank 1 is the lowest energy cluster).

-i

This is used to ignore any grid map header errors that may arise due to conflicting filenames. This
overrides the header checking that is normally performed to ensure compatible grid maps are
being used.
Methodology

-u

This returns a message describing the command line usage of AutoDock.

-t

This instructs AutoDock to parse the PDBQ file to check the torsion definitions, and then stop.

The Unix script “job” can be used to submit an AutoDock job, and then perform additional post-
processing, such as profiling, extracting job-information and creating a field file for AVS display
of the docked results. See the Appendix for more details.

22. Using the Command Mode in AutoDock

AutoDock can be run in “command mode”, using the “-c” flag thus:

% autodock3 -p lig.macro.dpf -l lig.macro.clg -c

38
 Methodology

When AutoDock has read in the grid maps specified in “lig.macro.dpf”, the program gives the
message “COMMAND MODE” and waits for the user to issue a command from the standard input.
These commands are described in more detail below.

An alternative way of using the command mode is to edit a file containing the commands you
wish AutoDock to execute (say “command.file”) and channel the output to a file (say “com-
mand.output”), thus:

% autodock3 -p lig.macro.dpf -l lig.macro.clg -c < command.file

> command.output &

AutoDock can also be used in a UNIX pipe command. This is valuable when an alternative search
procedure is desired. Here, the alternative search procedure issues commands to the standard out-
put, and reads the results from the standard input. In this case, AutoDock is behaving as an energy
server for the alternative search-procedure program.

There are eight recognized commands: AutoDock’s command interpreter is not case sensitive.

“eval” Evaluate this state’s total energy.

“epdb” Evaluate the energy of the named PDBQ file.
“outc” Output the last state’s PDB-formatted Cartesian coordinates.
“oute” Output (non-bond and electrostatic) energy breakdown, by atom.
“traj’’ Convert an SA trajectory file into PDB-formatted Cartesian coordinates.
“stop”, “exit”, “quit”Stop the AutoDock command mode interpreter.

eval

Methodology
Evaluates the total energy of a state defined by the subsequent state variables. This command uti-
lizes the trilinear interpolation routine in AutoDock along with the supplied grid maps defined in
the parameter file specified after the ‘-p’ flag to return this energy. The internal energy of the
ligand is also taken into account, as dictated by the values of the torsion angles supplied in the
ntor lines following this eval command line; ntor is the number of torsion angles defined in the
ligand PDBQ file, as described in the section “Defining Torsions in AutoDock”. The usage of this
command is:

eval <float> <float> <float> <float> <float> <float> <float>} Tx, Ty, Tz, Qx, Qy, Qz,
Qw (in ˚)
<float> } ith torsion angle, in ˚.
:
: ntor lines.

where: Tx, Ty, Tz are the coordinates of the center of rotation of the ligand; Qx, Qy, Qz is the unit
vector describing the direction of rigid body rotation, about which a rotation of angle Qw degrees
will be applied. The following ntor lines hold the torsion angles in degrees, given in the same
order as described in the AutoDock log file.

39
 Methodology

epdb
Calculates the energy of the molecule provided in the PDBQ file, thus:

epdb lig.pdbq

where: “lig.pdbq” is the PDBQ formatted coordinates of a molecule for which the interaction
energy with the macromolecule will be returned. The ‘-o’ flag supplied at the AutoDock execu-
tion line specifies the old format of PDBQ, with charges in columns 55-61; otherwise it is assumed
that the charges are in columns 71-76.

This command is useful when the state variables for a given molecule are not known, e.g. the x-
ray crystallographic conformation of the ligand.

outc
Returns the coordinates of the ligand at its current transformed position (in the form of a PDB
REMARK). The x,y,z coordinates will be determined by the state variables supplied to the eval
command.

oute
Returns the total internal energy of the ligand and the total energy of the complex, at the current
state variables. These two REMARK lines are written in PDB format, to the command output channel
and the log file.

traj
Methodology

Convert a simulated annealing “.trj” file into PDBQ format. Usage:

traj lig.trj

where “lig.trj” is a simulated annealing (SA) trajectory file written out by an earlier run of
AutoDock. This trajectory file contains the state variables for the states sampled during a simu-
lated annealing docking simulation. The torsions are assumed to be in exactly the same order as
the input ligand PDBQ file. The torsion angles in the trajectory file are relative to the input ligand’s
conformation.

See also the Appendix, script “runtrj”; and the next section, “Trajectory Files”.

stop, exit, quit

Halts the execution of AutoDock. A value of 0 is returned by the program, and the message
“autodock: Successful Completion” is written to the log file and standard error. Timing informa-
tion is also written. Note: “stop”, “exit” and “quit” are synonymous.

40
 Methodology

23. Trajectory Files

Note: Trajectories can only be generated during a simulated annealing run: they are not available
for the population-based genetic algorithm methods.

A trajectory (of state variables) can be written out during a normal simulated annealing docking ,
if the trajectory-frequency (set by the keyword “trjfrq”) in the docking parameter file is greater
than zero. This value defines the output frequency, in steps, for states sampled during the run. The
default trajectory filename extension is “.trj” . These state variables are all that is needed to
regenerate the coordinates of the ligand. The trajectory control parameter (either “A” or “E”)
allows the user to record only accepted moves (A); or, moves which are either accepted or
rejected (E). Just for information, a sample “.trj” trajectory file is shown below; you will not
need to create such files (unless you feel like creating an animation!):

______________________________________________________________________________

ntorsions 2
run 1
cycle 1
temp 300.000000
state 1 A -3.745762 -1.432243 -9.518171 23.713793 23.076145 0.713534 -0.023818 0.700216
30.606248
-4.894825
2.661499
:
:

Methodology
state 6 R -12.679995 -1.452641 -9.259430 21.634645 23.135242 0.653369 -0.440832
0.615448 39.127316
-31.636299
10.261519
state 7 a -8.746072 -1.458231 -9.080998 21.356874 23.325665 0.648312 -0.448577 0.615200
41.075955
-37.935175
11.918847
:

______________________________________________________________________________

There are several keywords: “run” and “cycle” are self-explanatory; “ntorsions” is the total
number of changing torsions in the ligand; “temp” is the annealing temperature for all subsequent
entries, unless otherwise stated. Each “state” record has the format:

state nstep acc_rej_code e_total e_internal x y z qx qy qz qw

where:
nstep = the number of the step, within this cycle;
acc_rej_code = ‘A’ = an accepted move whose energy was lower than its previous state;
= ‘a’ = an accepted move whose energy was higher than its previous
state, which nevertheless passed the Monte Carlo probability test at

41
 Methodology

this temperature;
= ‘R’ = a rejected move.
= ‘e’ = an edge-hit, also a treated as a rejected move.
e_total = total energy of the system, ligand + macromolecule;
e_internal = internal energy of ligand only;
x,y,z = translation of ligand center;
qx,qy,qz,qw = quaternion, which describes the ligand’s orientation;

In order to get a coordinate-based trajectory file, for visualization, the command mode of
AutoDock must be used to regenerate the coordinates from the state variables. Use the “traj”
command with the name of the pre-calculated trajectory file. For example, suppose there is a com-
mand file called “trj.conv.com” that contains:

_____________________________________________________________________________

traj lig.trj
stop

_____________________________________________________________________________

AutoDock would be executed a second time using the following command,

% autodock3 -p lig.macro.dpf -l lig.macro.trj.conv.log -c <

trj.conv.com > trj.conv.out

24. Evaluating the Results of a Docking

Methodology

At the end of an AutoDock job in which more than one run was performed, the program outputs a
histogram of clusters and their energies. Look in the lig.macro.dlg file for the word ‘HISTO-
GRAM’ all in upper-case. The clustering or structure binning of docked conformations is deter-
mined by the rms tolerance specified in Å by the “rmstol” keyword. The best conformation from
each cluster (i.e. that with the lowest energy) is written out in PDBQ format at the end of the log
file.

Use the UNIX grep command to extract information from the docking log file. For example,

% grep “^DPF>” lig.macro.dlg | sed ‘s/^DPF> //’

would extract all the lines that begin with “DPF>”, and pipe them into the stream editor, “sed”, to
strip out the “DPF>” prompts. Since each line read in from the input DPF is echoed in the log file
on such lines, this UNIX command would recover the original DPF that was used to generate
“lig.macro.dlg”.

To extract the conformations from the docking log file just use the “dockedtopdb” command:

42
 Methodology

% dockedtopdb lig.macro.dlg > lig.macro.dlg.pdb

This writes out a PDB formatted file, and uses the ‘MODEL’ and ‘ENDMDL’ records to denote the dif-
ferent dockings. Check that your molecular modelling package or viewer can parse these PDB
records.

Or if you need a PDBQ formatted file, use “dockedtopdbq” thus:

% dockedtopdbq lig.macro.dlg > lig.macro.dlg.pdbq

These docked structures can be read into any appropriate molecular modeling program, and the
results compared, where possible, with the experimental data.

The table of ranked clusters , under the heading ‘CLUSTERING HISTOGRAM’ in the log file, shows
the final docked energy for each conformation, and the rms difference in Å between the lowest
energy member of the cluster and every other member. The rms for the lowest member of the
group is by definition zero. You can extract this clustering histogram very easily using this com-
mand, which will print the results to the terminal:

% gethis lig.macro.dlg

After this table in the ‘lig.macro.dlg’ log file, the docked structures are output in PDBQ format.
Each conformation has a set of REMARK records, one of which describes the rms difference
between itself and the coordinates specified in the original input PDBQ file.This can be useful for
comparing how close each docked conformation is to the experimentally determined position.

Methodology
25. Visualizing Grid Maps

If you have access to AVS, you can visualize the grid maps by using a read field module. The user
must specify the ‘.fld’ file that was created by AutoGrid, in order to read in the grid maps. An
extract scalar module selects the grid map of interest, e.g. carbon affinity or electrostatics. The
resulting grid map data can be analyzed using arbitrary slicer and isosurface modules, in order to
examine cross sections and iso-energy contours respectively. Negative energy contours are most
informative for the atomic affinity grid maps, since they reveal favorable regions of binding.

26. Visualizing Trajectories

Trajectories can also be read into AVS using the read field module. The trajectory file is essen-
tially a set of “stacked” or concatenated PDB frames, and must be read in as a two dimensional
field (being the number of atoms in the ligand, and the number of frames in the trajectory file). By
paging through this field, using the orthogonal slicer, continuous replay of the trajectory can be

43
 Methodology

achieved using an animate integer module to control which PDBQ frame is selected by the
orthogonal slicer. This animates the sequence of sampled states and allows the user to view in real
time the progress of the docking simulation.

Unlike AVS, gOpenMol is free, and can be dowloaded via the web from:

http://laaksonen.csc.fi/gopenmol/gopenmol.html

It has various AutoDock tools, including the ability to read in and view AutDock trajectories: see:

http://laaksonen.csc.fi/gopenmol/help/main_autodock_widget.html

_______
Methodology

44
 Appendices

Appendix I: Shell Scripts and Awk Programs

cartopdbq
Usage: cartopdbq lig.car > lig.pdbq

Converts from Biosym InsightII “.car” format to PDBQ format

check-qs
Usage: check-qs lig
Needs: lig.pdbq
Creates: lig.err

Checks partial atomic charges in PDBQ file; any non-integral charges are reported.

checkqs
Usage: checkqs lig
Needs: lig.pdbq
Creates: lig.err

Sorts the input PDBQ file by residue number before running the result through check-qs.

clamp
Usage: clamp grid.map > grid.map.NEW

Clamps any AutoGrid map values that exceed ECLAMP (normally set to 1000.0)

cnvmol2topdbq Appendices
Usage: cnvmol2topdbq lig.mol2 > lig.pdbq
Needs: lig.mol2
Creates: lig.pdbq

This converts from (fixed format) Tripos SYBYL mol2 fiormat into PDBQ format, but stores all the
residues’ chain-IDs specified by the SUBSTRUCTURE records in the mol2 file. These chain-IDs are
then output when the PDBQ lines are written.

45
 Appendices

deftors
Usage: deftors lig.mol2
Creates: lig.pdbq, lig.err

Sets up rotatable bonds for AutoDock. This script launches AutoTors, with the -A +15.0, -a, -h
and -m flags; it also checks the charges in the output PDBQ file, with check-qs.

dpf3gen
Usage: dpf3gen lig.pdbq > lig.dpf

This is normally used by mkdpf3, so you should not use this script by hand.

It generates a pre-cursor to a default AutoDock docking parameter file. You must edit the file
before using it. This reads in the small molecule PDBQ file, detects all atom types present in the
lig.pdbq; and creates a docking parameter file for AutoDock. Note the user must replace the tags
<lig> and <macromol> by appropriate filename stems.

This uses equilibrium separations and well depths to define pairwise energy potentials, rather than
coefficients.

get-coords
Usage: get-coords lig.vol > lig.txt

This is used as part of prepare, prepare-gpf+dpf, prepare II and prepare III. It

takes the .vol file created by pdb-volume and creates a line that can be used in the grid
parameter file to specify the center of the maps.

get-docked
Usage: get-docked lig.macro.dlg
Creates: lig.macro.dlg.pdb
Appendices

This extracts the docked records from a docking log file. This is very useful when wanting to view
the results of a docking in a molecular modelling program or molecular viewer. It is essentially
the same as the ‘dockedtopdb’ awk program.

gethis
Usage: gethis lig.macro.dlg

46
 Appendices

This outputs the histogram from the confomational analysis section of the AutoDock log file,
lig.macro.dlg, and writes it to the screen.

getready
Usage: getready lig.pdb
Needs: pdbinfo, pdbsplitchains, pdbwaters, pdbdewater
Creates: lig.info, lig_.atm.pdb, lig_.het.pdb, lig_.wat.pdb

This is a very useful script to get started. It will split a PDB file into separate files, each containing
a different chain, and will split each of these chains into ATOM, non-water HETATM and water
containing PDB files. Also the .info file is a useful summary description of the PDB file.

gpf3gen
Usage: gpf3gen lig.pdbq[s] > lig.gpf

This is used by mkgpf3, and should not be used by hand; otherwise the user must edit certain tags
by hand before this can be used by AutoGrid.

This generates a precursor to a grid parameter file. It takes lig.pdbq as its input file, detects all
atom types present, and creates the properly formatted parameter file for AutoGrid. It uses equi-
librium separations and well depths to define pairwise energy potential. It also assigns atomic sol-
vation parameters, based on Stouten, P.F.W., FrÖmmel, C., Nakamura, H., and Sander, C. (1993),
"An effective solvation term based on atomic occupancies for use in protein simulations", Molec-
ular Simulation, 10, 97-120.

histable
Usage: histable lig.macro.dlg
Creates: lig.macro.dlg.tbl

This extracts the histogram from the docking log file, and counts all the ‘#’ symbols, writing the
result in a table file. This is suitable for input to a variety of graph drawing programs and spread-
sheets.
Appendices

job3
Usage: job3 lig.macro > lig.macro.joblog &

Launches a single AutoDock 3.0 job. It assumes that “lig.macro.dpf” exists, and executes
AutoDock using the arguments:

47
 Appendices

autodock3 -p lig.macro.dpf -l lig.macro.dlg

You must edit this script the first time you use it, so that the environment variables $root, $bin and
$sh are correctly set equal to, respectively: the path to the root of AutoDock tree, the architecture-
dependent binary subdirectory and the Unix scripts subdirectory. The file lig.macro.joblog
contains the output from the job script.

makebox
Usage: makebox macro.gpf >! macro.gpf.box.pdb
Creates: macro.gpf.box.pdb

This creates a PDB file from the grid parameter file ‘macro.gpf’, that shows how big and where
the grid box will be when AutoGrid calculates the grid maps. You can use this ‘box molecule’ to
help refine the center and number of grid points in the grid maps, before you run AutoGrid.

If you colour the ‘box molecule’ by atom type, i.e. red for oxygen, green for carbon, and blue for
nitrogen, then the edges of this box will be coloured-coded to indicate the Cartesian axes. R,G,B
will correspond to x,y,z, respectively. Your molecule viewer must obey the CONECT records in
the ‘macro.gpf.box.pdb’ file, even though the corresponding bonds may appear too long, other-
wise the edges of the grid box will not be displayed.

mkbox
Usage: mkbox macro.gpf >! macro.gpf.box.pdb
Creates: macro.gpf.box.pdb

This is very similar to ‘makebox’, except that this puts a phosphorus atom at the minimum x, min-
imum y and minimum z coordinates of the box. This helps to convey which directions are +x, +y
and +z. Once again, if oxygen is red, carbon is green and nitrogen is blue, then R,G,B will corre-
spond to x,y,z, respectively.

mkdlgfld
Usage: mkdlgfld lig.macro.dlg
Appendices

Needs: lig.macro.dlg
Creates: lig.macro.dlg.fld

Only needed for AVS users.

This extracts the “AVSFLD” records from an AutoDock log file, and puts them in
lig.macro.dlg.fld. These “AVSFLD” descriptors must be removed before the file can be used in
AVS.

48
 Appendices

mkdpf3
Usage: mkdpf3 lig.pdbq macromol.pdbqs
Needs: dpf3gen, dpf3gen.awk (AWK program)
Creates: lig.macro.dpf

This creates a default docking parameter file for AutoDock 3.0; it needs the ligand in PDBQ for-
mat and the macromolecule in PDBQS format. It uses the script dpf3gen, which in turn calls the
awk program ‘dpf3gen.awk’. The lig.macro.dpf docking parameter file is based on the atom
types detected in the input lig.pdbq file. See dpf3gen above.

mkgpf3
Usage: mkgpf3 lig.pdbq macromol.pdbqs
Needs: gpf3gen, gpf3gen.awk (AWK program), pdbcen (AWK program)
Creates: macro.gpf

This creates a default grid parameter file for AutoGrid 3.0; it needs gpf3gen.awk and pdbcen,
both awk programs. See gpf3gen above.

mol2fftopdbq
Usage: mol2fftopdbq lig.mol2 > lig.pdbq
Needs: lig.mol2
Creates: lig.pdbq

Converts from free formatted SYBYL mol2 into AutoDock PDBQ format. Chain-IDs specified in
the mol2 file by the SUBSTRUCTURE records are incorporated into the PDBQ file.

mol2topdbq
Usage: mol2topdbq lig.mol2
Needs: lig.mol2
Creates: lig.pdbq Appendices
Converts from fixed-format SYBYL mol2 into AutoDock PDBQ format, and automatically
names the output based on the stem of the input mol2 file. Do not use “mol2topdbq lig.mol2 >
lig.pdbq”, because “lig.pdbq” is automatically created.

mol2topdbqs
Usage: mol2topdbqs lig.mol2

49
 Appendices

Needs: lig.mol2
Creates: lig.pdbqs

Converts from SYBYL mol2 format into AutoGrid 3.0 PDBQS format, by calling mol2topdbq
then running addsol on the intermediate PDBQ file. Like mol2topdbq, it also removes any lone
pairs (using “rem-lp”), and automatically names the output based on the stem of the input mol2
file. There is no need to use “mol2topdbq lig.mol2 > lig.pdbq”, because “lig.pdbqs” is au-
tomatically created.

pdbcen
Usage: pdbcen lig.pdb
Creates: a “gridcenter” line in AutoGrid GPF format, holding the x,y,z coordinates of the
molecule.

This calculates the center of a molecule supplied in PDB format, and outputs a line holding the
x,y,z coordinates of the molecule for inclusion in an AutoGrid 3.0 grid parameter file (GPF).

pdb-center
Usage: pdb-center [ lig.pdb | lig.pdbq ] > lig2.pdb

Calculates the center of mass of each residue; writes these coordinates out using REMARK
records.

pdb-center-all
Usage: pdb-center-all [ lig.pdb | lig.pdbq ] > lig2.pdb

Calculates the center of mass of each residue; writes these coordinates out using REMARK
records.Also calculates the center of all the residues.

pdb-distance
Appendices

Usage: pdb-distance macro.pdb

The first line of the macro.pdb file defines the center of the distance profile. It is just a copy of the
line containing the atom of interest, which will be the origin for the distance calculations. How-
ever, it must have the ATOM or HETATM record replaced with a non-PDB tag, ‘FROM’. The
x,y,z coordinates in this FROM line will then be used to calculate the distance to the center of
each residue in the protein. Finally, this awk program outputs a bar chart using ‘#’ symbols, show-
ing the distance from this point to each residue. This can be useful to identify all the residues

50
 Appendices

nearest a particular ligand atom, or near an active site.

pdbdewater
Usage: pdbdewater macro.pdb >! macro.dry.pdb

This removes any water records from a PDB file.

pdbinfo
Usage: pdbinfo macro.pdb

Builds a summary of the contents of a PDB file.

pdb-volume
Usage: pdb-volume [ lig.pdb | lig.pdbq ] > lig2.pdb

Calculates the center of mass of each residue. Writes out REMARKs showing these coordinates.
Draws ASCII diagram showing volume extents of each residue.

pdbqtobnd
Usage: pdbqtobnd lig
Needs: lig.pdbq
Creates: lig.bnd

Creates “lig.bnd” from the existing “lig.pdbq” ligand PDBQ file. Note this script needs just the
stem of the file name. This script executes “pdbqtoatm” and “atmtobnd”: the latter is an execut-
able, not a script, so it must be compiled for each architecture and operating system used.

pdbqtopdb Appendices
Usage: pdbqtopdb lig.pdbq > lig.pdb

Converts from AutoDock PDBQ to PDB format.

pdbsplitchains
Usage: pdbsplitchains macro.pdb

51
 Appendices

Creates separate PDB files that contain each of the chains in macro.pdb. The chain IDs are used
to name the new PDB files. If there is no chain ID, the underscore character, ‘_’ is used.

pdbtoatm
Usage: pdbtoatm lig.pdbq > lig.atm

This creates a Connolly ATM formatted file “lig.atm” from the ligand PDBQ file, “lig.pdbq”.
This is used to create input for the utility program atmtobnd to generate a bond connectivity
file.

pdbdewaters
Usage: pdbwaters macro.pdb > macro.wet.pdb

Extracts the waters from macro.pdb into macro.wet.pdb.

prepare
Usage: prepare m s

where: m.pdb and s.pdbq contain the receptor and ligand respectively. Prepare performs
the following eight steps. The macromolecule ‘.pdb’ filename stem is represented by “m”, and
Appendices

52
 Appendices

the ligand ‘.pdbq’ filename stem by “s”:

prepare m s
m.pdb s.pdb

grep ^[AT][TE][OR][M ] AMPAC/MOPAC, Insight95, Sybyl

pdb-volume
m.enz s.pdbq s.vol
renumber-residues get-coords
gpf-gen
dpf-gen

m.rnm.log m.rnm s.gpf

protonate
s.dpf

m.polH
Key:
q.kollua
m = macromolecule;
check-qs s = ligand.
m.err m1.pdbq

1. Extracts all ATOM and TER records from m.pdb into m.enz;
2. Renumbers residues to avoid problems in protonate-step;
3. Adds polar hydrogens to m.enz, creating m.polH;
4. Somewhat crudely assigns partial atomic charges to m.polH, creating m.pdbq;
5. Checks charges in m.pdbq, all errors held in m.err;
6. Creates s.gpf, a parameter file for AutoGrid, based on ligand file s.pdbq;
7. Creates s.vol, a volume dimensions file; and finally,
8. Creates s.dpf, a parameter file for AutoDock, based on ligand file s.pdbq;

Its arguments are the stem of the filename of the macromolecule ‘.pdb’ file and that of the ligand
PDBQ file. See the flowchart below for more details. It shows what files are created by ‘pre-
pare’, and which scripts or programs are used. Steps 1.-4. are better carried out with a reliable
molecular modeling system: these steps can produce some odd results unless carefully checked.
Appendices

The user must check the m.err error file to ensure there are no non-integral charges, either on
any residue in the macromolecule, or on the macromolecule as a whole. If there are, then the user
must repair the m.pdbq file. This problem can arise if there are atoms for which no coordinates
were assigned by the crystallographer, e.g. due to ambiguous electron density. Assuming there
were no problems, s.gpf and s.dpf should be successfully produced.

53
 Appendices

prepare-gpf+dpf
Usage: prepare-gpf+dpf macro lig

Executes only steps 6. through 8.

rem-lp
Usage: rem-lp lig.pdbq
or: rem-lp lig.mol2
Creates: lig.pdbq
or: lig.mol2

This removes the lone-pairs (atom name = LP) added by some molecular modelling programs,
such as SYBYL, and adds their partial charges on to that of the atom to which they were attached
(SG in cysteines and SD in methionines). Otherwise, AutoDock treats lone-pairs as carbon atoms.
(Note: if you need lone-pairs, you can force AutoGrid to calculate a grid map for “LP” atoms,
using the atom code “L” in the “types” commands of AutoGrid and AutoDock).

renumberatoms
Usage: lig.pdb > lig2.pdb

Used to renumber the atom IDs in the first column of the ATOM and HETATM records of a PDB
file. Also updates the CONECT records appropriately.

renumber-residues
Usage: lig.pdb > lig.rnm

Used by prepare to renumber residues in the macromolecule contiguously. This step is needed
prior to using protonate, which may fail if there are gaps in the residue numbers.

resrange
Appendices

Usage: resrange lig.pdb

This is handy to summarise the range(s) of residues in a given protein PDB file.

54
 Appendices

runtrj
Usage: runtrj lig
Needs: lig.dpf, lig.trj
Creates: lig.tcom, lig.tlg and lig.tout

This creates an AutoDock command file, lig.tcom, which is then used to convert the trajectory
written in state variables (lig.trj), into a trajectory written in cartesian coordinates. lig.trj is
created by an earlier run of AutoDock, in which trjfrq was set to a non-zero value.

stats
Usage: stats columns.dat

This is a very useful, general awk program. Use it to calculate the minimum, maximum, mean and
standard deviation for each column of numbers in an input file, here ‘columns.dat’. Any alpha-
numeric columns will be ignored.

Appendices

55
 Appendices

Appendix II: Parameters from AutoDock Version 1

Table II.1: Lennard-Jones C6 parameters (AutoDock version 1.0)

C N O S H

C 1127.684 783.3452 633.7542 1476.364 226.9102

N 783.3452 546.7653 445.9175 1036.932 155.9833

O 633.7542 445.9175 368.6774 854.6872 124.0492

S 1476.364 1036.932 854.6872 1982.756 290.0756

H 226.9102 155.9833 124.0492 290.0756 46.73839

Table II.2: Lennard-Jones C12 parameters (AutoDock version 1.0)

C N O S H

C 1272653. 610155.1 588883.8 1569268. 88604.24

N 610155.1 266862.2 249961.4 721128.6 39093.66

O 588883.8 249961.4 230584.4 675844.1 38919.64

S 1569268. 721128.6 675844.1 1813147. 126821.3

H 88604.24 39093.66 38919.64 126821.3 1908.578

Table II.3: Hydrogen bonding 12-10 parameters (AutoDock version 1.0)

Atoms i-j C12 C10

Appendices

O-H 75570. 23850.

N-H 75570. 23850.

S-H 2657200. 354290.

56
 Appendices

Appendix III: AutoDock File Formats

The formats will sometimes be given with notation such as ’%d’ to indicate a decimal integer;
’%6.3f’ for a floating point number with up to 6 characters and 3 digits after the decimal place; or
’%-7s’ for a left-justified string 7 characters wide. This notation is compatible with C, C++, awk/
nawk/gawk, and with a slight modification, Python.

The “◊” symbol is used to indicate one space.

The type of an argument is described using the following style:

<string> = an alphanumeric string. In most cases, this is a valid filename;

<character> = a single letter;
<integer> = a decimal integer;
<positive_integer> = a decimal integer greater than zero;
<long_integer> = a decimal integer in the “long” range (depends on computer);
<float> = a floating point or real number.

III 1. Protein Data Bank with Partial Charges: PDBQ

Extension: .pdbq

The name ’PDBQ’ derives from ’PDB’, the Protein DataBank, and ’Q’, a common symbol for
partial charge. As the name suggests, the PDBQ format is very similar to the PDB format for
ATOM records, with a modification in columns 71-76 (counting the first column as 1, not 0) to
carry the partial charge, as %6.3f. Thus, the format of the whole line is as follows:

“ATOM◊◊%5d◊%-4s%1s%-3s◊%1s%4d%1s◊◊◊%8.3f%8.3f%8.3f%6.2f%6.2f%4s%6.3f\n",
atom_serial_num, atom_name, alt_loc, res_name, chain_id, res_num, ins_code, x, y, z,
occupancy, temp_factor, footnote, partial_charge

In addition to this, there are various records (ROOT, TORSION and BRANCH, for example) in
the ligand PDBQ file that specify which portion of the molecule is rigid and which is flexible. Appendices

III 2. PDBQ with Solvation Parameters: PDBQS

Extension: .pdbqs

This format is derived from the PDBQ format, and is used to specify the atomic solvation param-
eters for the macromolecule, hence the “S”. The format of the lines is:

57
 Appendices

"ATOM◊◊%5d◊%-4s%1s%-3s◊%1s%4d%1s◊◊◊%8.3f%8.3f%8.3f%6.2f%6.2f%4s%6.3f%8.2f%8.2f\n",
atom_serial_num, atom_name, alt_loc, res_name, chain_id, res_num, ins_code, x, y, z,
occupancy, temp_factor, footnote, partial_charge,
atomic_fragmental_volume, atomic_solvation_parameter

The atomic fragmental volume and solvation parameters are derived from the method of Stouten
et al.24

III 3. AutoGrid Grid Parameter File: GPF

Extension: .gpf

The input file is often referred to as a “grid parameter file” or “GPF” for short. The scripts
described in the appendices give these files the extension “.gpf”. In the grid parameter file, the
user must specify the following spatial attributes of the grid maps:

1. the center of the grid map;

2. the number of grid points in each of the x-, y- and z-directions; and

3. the separation or spacing of each grid point.

In addition, the pairwise-atomic interaction energy parameters must be specified. The following
lines are required for each ligand atom type, Y:

4. the grid map filename for atom type Y;

5. seven lines containing the non-bonded parameters for each pairwise-atomic interaction, in
the following order: Y-C, Y-N, Y-O, Y-S, Y-H, Y-X, (X is any other atom type) and Y-M (M is a
metal, say).

Using coefficients Cn, Cm, n and m, the pairwise interaction energy, V(r) is given by:

C Cm
V ( r ) ≈ -----n-n – ------
m
r r

Alternatively, the user can specify reqm, ε, n and m:

Appendices

m n
-------------- εr -------------- εr eqm
n m
n – m eqm n–m
V ( r ) ≈ -----------------------
n
- – -----------------------
m
-
r r

24. Stouten, P. F. W., Frömmel, C., Nakamura, H. and Sander, C. (1993). “An effective solvation term based
on atomic occupancies for use in protein simulations”, Molecular Simulations, 10, 97-120.

58
 Appendices

This latter method of specification is more intuitive for the user, while AutoGrid handles the cal-
culation of the coefficients. By default, the Y-X and Y-M lines are copies of the Y-H line. But in
some systems, such as receptors which consist of DNA/protein complexes, both sulphur and
phosphorus can be present. In this scenario, the Y-X line can be used for modeling interactions
with receptor-phosphorus atoms. A very rough approximation for phosphorus parameters is to
borrow those of carbon.

The “elecmap” line in the grid parameter file is the filename of the electrostatic potential grid
map. The following parameter, “dielectric”, if negative, indicates that the distance-dependent
dielectric function of Mehler and Solmajer3 will be used. If positive, however, the value of that
number will be used as a constant dielectric. For example, if the value were 40.0, then a constant
dielectric of 40 would be used.

The AutoGrid parameter file format is described below.

AutoGrid Keywords and Commands

receptor <string>
Macromolecule filename, in PDBQ format.

gridfld <string>
The grid field filename, which will be written in a format readable by AutoDock and AVS25. The
filename extension must be ‘.fld’.

npts <integer> <integer> <integer>

Number of x-, y- and z-grid points. Each must be an even integer number. When added to the cen-
tral grid point, there will be an odd number of points in each dimension. The number of x-, y- and
z-grid points need not be equal.

spacing <float>
The grid point spacing, in Å (see the diagram on page 8). Grid points must be uniformly spaced in
AutoDock: this value is used in each dimension.

gridcenter <float> <float> <float>

gridcenter auto
The user can explicitly define the center of the grid maps, respectively the x, y and z coordinates of
the center of the grid maps (units: Å, Å, Å.) Or the keyword “auto” can be given, in which case
Appendices

AutoGrid will center the grid maps on the center of mass of the macromolecule.

types <string>
1-letter names of the atom types present in the ligand; e.g. if there are carbons, nitrogens, oxygens
and hydrogens, then this line will be “CNOH”; there are no delimiters.

25. “AVS” stands for “Application Visualization System”; AVS is a trademark of Advanced Visual Systems
Inc., 300 Fifth Avenue, Waltham, MA 02154.

59
 Appendices

smooth <float>
This is always 0.5Å, when using the AutoDock 3.0 free energy function. It is used to smooth the
pairwise atomic affinity potentials (both van der Waals and hydrogen bonds). See the Theory sec-
tion for more details.

map <string>
Filename of the grid map, for ligand atom type Y; the extension is usually “.map”.

nbp_coeffs <float> <float> <integer> <integer>

Either “nbp_coeffs” or “nbp_r_eps” keywords can be used to define Lennard-Jones or hydrogen
bond interaction energy parameters. The keyword “nbp_coeffs” specifies coefficients and expo-
nents, in the order “Cn Cm n m”, delimited by spaces; n and m are integer exponents. The units of
Cn and Cm must be kcal mol-1 Ån and kcal mol-1 Åm respectively; n and m have no units.

nbp_r_eps <float> <float> <integer> <integer>

Alternatively, the user can employ “nbp_r_eps” to specify the equilibrium distance and well
depth, epsilon, for the atom pair. The equilibrium separation has units of Å and the well depth,
epsilon, units of kcal mol-1. The integer exponents n and m must be specified too.

In either case, the order of the parameters must be: Y-C, Y-N, Y-O, Y-S, Y-H, Y-X, and Y-M. Repeat
1 “map” line and the 7 “nbp_coeffs”or “nbp_r_eps” lines, for each atom type, Y, present in the
ligand being docked.

sol_par <float> <float>

This is used to define the atomic fragmental volume and solvation parameters, and should not be
changed from the Stouten values used to calibrate the AutoDock 3.0 free energy function.

constant <float>
This is added to all the values in a grid map, and is only set to a non-zero, positive number for
hydrogen bonding maps. This value is essentially the penalty for un-formed hydrogen bonds in
the complex.

elecmap <string>
Filename for the electrostatic potential energy grid map to be created; filename extension ‘.map’.

dielectric <float>
Appendices

Dielectric function flag: if negative, AutoGrid will use distance-dependent dielectric of Mehler
and Solmajer3; if the float is positive, AutoGrid will use this value as the dielectric constant.

fmap <string>
(Optional.) Filename for the so-called “floating” grid map26; filename extension ‘.map’. In such

26. This grid map is not used in AutoDock 3.0; its utility is under investigation, and may be included in a
later version.

60
 Appendices

floating grids, the scalar at each grid point is the distance to the nearest atom in the receptor. These
values could be used to guide the docking ligand towards the receptor’s surface, thus avoiding
non-interesting, empty regions.

III 4. Grid Map File

Extension: .map

The first six lines of each grid map hold header information which describe the spatial features of
the maps and the files used or created. These headers are checked by AutoDock to ensure that
they are appropriate for the requested docking. The remainder of the file contains grid point ener-
gies, written as floating point numbers, one per line. They are ordered according to the nested
loops z( y( x ) ). A sample header from a grid map is shown below:

______________________________________________________________________________

GRID_PARAMETER_FILE vac1.nbc.gpf
GRID_DATA_FILE 4phv.nbc_maps.fld
MACROMOLECULE 4phv.new.pdbq
SPACING 0.375
NELEMENTS 50 50 80
CENTER -0.026 4.353 -0.038
125.095596
123.634560
116.724602
108.233879
:

______________________________________________________________________________

III 5. Grid Map Field File

Extension: .maps.fld

This is essentially two files in one. It is both an AVS field file, and and AutoDock input file with
AutoDock-specific information ‘hidden’ from AVS in the comments at the head of the file.
AutoDock uses this file to check that all the maps it reads in are compatible with one-another and
Appendices
itself. For example, in this file, the grid spacing is 0.375 Angstroms, there are 60 intervals in each
dimension, the grid is centered near (46,44,14), it was calculated around the macromolecule
‘2cpp.pdbqs’, and the AutoGrid parameter file used to create this and the maps was ‘2cpp.gpf ’.
This file also points to a second file, ‘2cpp.maps.xyz’, which contains the minimum and maxi-
mum extents of the grid box in each dimension, x, y, and z. Finally, it lists the grid map files that
were calculated by AutoGrid, here ‘2cpp.C.map’, ‘2cpp.O.map’ and ‘2cpp.e.map’.

______________________________________________________________________________

61
 Appendices

# AVS field file

#
# AutoDock Atomic Affinity and Electrostatic Grids
#
# Created by autogrid3.
#
#SPACING 0.375
#NELEMENTS 60 60 60
#CENTER 46.508 44.528 14.647
#MACROMOLECULE 2cpp.pdbqs
#GRID_PARAMETER_FILE 2cpp.gpf
#
ndim=3 # number of dimensions in the field
dim1=61 # number of x-elements
dim2=61 # number of y-elements
dim3=61 # number of z-elements
nspace=3 # number of physical coordinates per point
veclen=3 # number of affinity values at each point
data=float # data type (byte, integer, float, double)
field=uniform # field type (uniform, rectilinear, irregular)
coord 1 file=2cpp.maps.xyz filetype=ascii offset=0
coord 2 file=2cpp.maps.xyz filetype=ascii offset=2
coord 3 file=2cpp.maps.xyz filetype=ascii offset=4
label=C-affinity # component label for variable 1
label=O-affinity # component label for variable 2
label=Electrostatics # component label for variable 2
#
# location of affinity grid files and how to read them
#
variable 1 file=2cpp.C.map filetype=ascii skip=6
variable 2 file=2cpp.O.map filetype=ascii skip=6
variable 3 file=2cpp.e.map filetype=ascii skip=6

______________________________________________________________________________

III 6. AutoDock Docking Parameter File: DPF

Extension: .dpf

AutoDock 3.0 has an interface based on keywords. This is intended to make it easier for the user
to set up and control a docking job, and for the programmer to add new commands and function-
ality. The input file is often referred to as a “docking parameter file” or “DPF” for short. The
scripts described in the appendices give these files the extension “.dpf”.
Appendices

All delimiters where needed are white spaces. Default values, where applicable, are given in
square brackets [thus]. A comment must be prefixed by the “#” symbol, and can be placed at the
end of a parameter line, or on a line of its own.

Although ideally it should be possible to give these keywords in any order, not every possible
combination has been tested, so it would be wise to stick to the following order.

62
 Appendices

Command to set the seed for the random number generator

seed <long_integer>
seed time
seed pid
seed <long_integer> <long_integer>
seed time <long_integer>
seed <long_integer> time
seed time pid
seed pid <long_integer>
seed <long_integer> pid
seed pid time
There are two possible random number generator libraries. One is the system’s own implementa-
tions, and the second is the platform-independent library from the University of Texas Biomedical
School. If the user gives just one argument to “seed”, then AutoDock will use the system’s imple-
mentation of the random number generator and corresponding system seed call. On most plat-
forms, these are “drand48” and “srand48”. The platform-independent library, however, requires
two seed values. Giving two arguments to “seed” tells AutoDock to use the platform-independent
library for random number generation.

The random-number generator (RNG) for each docking job can be ‘seeded’ with either a user-
defined, a time-dependent, or process-ID-dependent seed. These two seeds can be any combina-
tion of explicit long integers, the keyword “time” or the keyword “pid”. When two arguments to
seed are given, the portable RNG is used; when one is given, the built-in RNG (usually the
“drand48” C-function) is used. The portable RNG is required for the genetic algorithm and the
Solis and Wets routines. The portable RNG cannot be used with the simulated annealing routine:
this needs just one seed parameter. The keyword, “time” gives the number of seconds since the
epoch. The epoch is referenced to 00:00:00 CUT (Coordinated Universal Time) 1 Jan 1970. The
“pid” gives the UNIX process ID of the currently executing AutoDock process, which is reading
this parameter file.

Parameters defining the grid maps to be used

types <string>
Atom names for all atom types present in ligand. Each must be a single character, and only one of:
C, N, O, S, H, X, or M. The maximum number of characters allowed in this line is
ATOM_MAPS, which is defined in the “autodock.h” include file. Do not use any spaces to delimit
the types: they are not needed.
Appendices
fld <string>
Grid data field file created by AutoGrid and readable by AVS (must have the extension “.fld”).

map <string>
Filename for the first AutoGrid affinity grid map of the 1st atom type. This keyword plus filename
must be repeated for all atom types in the order specifed by the “types” command. In all map
files a 6-line header is required, and energies must be ordered according to the nested loops z( y( x

63
 Appendices

) ).

map <string>
Filename for the electrostatics grid map. 6-line header required, and energies must be ordered
according to the nested loops z( y( x ) ).

Parameters defining the ligand and its initial state

move <string>
Filename for the ligand to be docked. This contains most importantly, atom names, xyz-coordi-
nates, anb partial atomic charges in PDBQ format. (Filename extension should be “.pdbq”).

about <float> <float> <float>

Use this keyword to specify the center of the ligand, about which rotations will be made. (The
coordinate frame of reference is that of the ligand PDBQ file.) Usually the rotation center of the
ligand is the mean x,y,z-coordinates of the molecule. Inside AutoDock, the “about” xyz-coordi-
nates are subtracted from each atom’s coordinates in the input PDBQ file. So internally, the
ligand’s coordinates become centered at the origin. Units: Å, Å, Å.

tran0 <float> <float> <float>

tran0 random
Initial coordinates for the center of the ligand, in the same frame of reference as the receptor’s
grid maps. The ligand, which has been internally centered using the “about” coordinates, has the
xyz-coordinates of the initial translation “tran0 x y z” added on. Every run starts the ligand
from this location.

Alternatively, the user can just give the keyword “random” and AutoDock will pick random initial
coordinates instead.

If there are multiple runs defined in this file, using the keyword “runs”, then each new run will
begin at this same location.

The user must specify the absolute starting coordinates for the ligand, used to start each run. The
user should ensure that the ligand, when translated to these coordinates, still fits within the vol-
ume of the grid maps. If there are some atoms which lie outside the grid volume, then AutoDock
will automatically correct this, until the ligand is pulled completely within the volume of the
grids. (This is necessary in order to obtain complete information about the energy of the initial
Appendices

state of the system.) The user will be notified of any such changes to the initial translation by
AutoDock. (Units: Å, Å, Å.)

quat0 <float> <float> <float> <float>

quat0 random
[1, 0, 0, 0˚]
Respectively: Qx, Qy, Qz, Qw. Initial quaternion (applied to ligand) - Qx, Qy, Qz define the unit vector
of the direction of rigid body rotation, and Qw defines the angle of rotation about this unit vector,

64
 Appendices

in ˚ . (Units: none,none,none, ˚.)

Qy
A quaternion. Qw

Qx
Qz

Alternatively, the user can just give the keyword “random” and AutoDock will pick a random unit
vector and a random rotation (between 0˚ and 360˚) about this unit vector. Each run will begin at
this same random rigid body rotation.

ndihe <integer>
Number of dihedrals or rotatable bonds in the ligand. This may be specifed only if rotatable bonds
have been defined using ROOT, BRANCH, TORS etc. keywords in the PDBQ file named on the
“move” line. The number supplied to this command must agree with the number of torsions
defined in this ligand PDBQ file. If this keyword is used, then the next keyword, dihe0, must
also be specified. Note that if ndihe and dihe0 are not specified and there are defined torsions
in the ligand PDBQ file, AutoDock assumes that the chi1, chi2, chi3, etc. are all zero, and does not
change the initial ligand torsion angles. (See also “torsdof” below).

dihe0 <float> ...

Initial relative dihedral angles; there must be ndihe floating point numbers specified on this
line. Each value specified here will be added to the corresponding torsion angle in the input
PDBQ file, at the start of each run. Torsion angles are only specified by two atoms, so the defini-
tion of rotations is relative. Units: ˚.

Parameters defining ligand step sizes

tstep <float>
tstep <float> <float>
[2.0 Å]
The first form, with one argument, defines the maximum translation jump for the first cycle that
the ligand may make in one simulated annealing step. When “trnrf” is less than 1, the reduction
factor is multiplied with the tstep at the end of each cycle, to give the new value for the next cycle.
Appendices
The second form allows the user to specify the value for the first cycle and the last cycle:
AutoDock then calculates the reduction factor that satisfies these constraints. Units: Å.

qstep <float>
[50.0˚]
Maximum orientation step size for the angular component, w, of quaternion. Units: ˚.

dstep <float>

65
 Appendices

[50.0˚]
Maximum dihedral (torsion) step size. Units: ˚.

Parameters defining optional ligand torsion constraints

barrier <float>
[10000.0]
(Optional) This defines the energy-barrier height applied to constrained torsions. When the tor-
sion is at a preferred angle, there is no torsion penalty: this torsion’s energy is zero. If the torsion
angle falls within a disallowed zone, however, it can contribute up to the full barrier energy. Since
the torsion-energy profiles are stored internally as arrays of type ‘unsigned short’, only positive
integers between 0 and 65535 are allowed.

gausstorcon <integer> <float> <float>

(Optional) Adds a constraint to a torsion. The torsion number is identified by an integer. This
identifier comes from the list at the top of the AutoTors-generated input ligand PDBQ file (on the
REMARK lines). An energy profile will be calculated for this torsion. An inverted Gaussian bell
curve is added for each new constraint. To completely specify each Gaussian, two floating point
numbers are needed: the preferred angle and the half-width respectively (both in degrees). Note
that the preferred angle should be specified in the range -180˚ to +180˚; numbers outside this
range will be wrapped back into this range. This angle, χ, is relative to the original torsion angle
in the input structure. The half-width is the difference between the two angles at which the energy
is half the barrier (B/2 in the diagram above). The smaller the half-width, the tighter the con-
straint.

B
Energy / kcal mol-1

torsion 0˚
constraint χ
half-width
B, barrier
B/2 +ve -ve
/ kcal mol-1

0
-180˚ 0˚ 180˚
χ, preferred angle
Appendices

Torsion angle / ˚

If you wish to constrain to absolute-valued torsion angles, it will be necessary to zero the initial
torsion angles in the ligand, before input to AutoTors. The problem arises from the ambiguous 2-
atom definition of the rotatable bond B-C. To identify a torsion angle unambiguously, 4 atoms

66
 Appendices

must be specified: A-B-C-D:

A
A
+ve
C B
B
D
D
χ
looking down the B-C bond

The sign convention for torsion angles which we use is anti-clockwise (counter-clockwise) are
positive angles, clockwise negative. In the above diagram, looking down the bond B-C, the dihe-
dral angle A-B-C-D would be positive.

There is no limit to the number of constraints that can be added to a given torsion. Each new tor-
sion-constraint energy profile is combined with the pre-existing one by selecting the minimum
energy of either the new or the existing profiles.

showtorpen
(Optional) (Use only with “gausstorcon”) This switches on the storage and subsequent output of
torsion energies. During each energy evaluation, the penalty energy for each constrained torsion,
as specified by the “gausstorcon” command, will be stored in an array. At the end of each run, the
final docked conformation’s state variables are output, but with this command, the penalty energy
for each torsion will be printed alongside its torsion angle.

hardtorcon <integer> <float> <float>

(Optional) This command also adds a torsion constraint to the <integer>-th torsion, as numbered
in the AutoTors-generated REMARKs. The first float defines the preferred relative angle, and the
second specifies the full width of the allowed range of torsion angles (both in degrees). This type
of torsion constraint is “hard” because the torsion is never allowed to take values beyond the range
defined. For example, “hardtorcon 3 60. 10.” would constrain the third torsion to values between
55˚ and 65˚.

Parameter affecting torsional free energy

Appendices
torsdof <integer> <float>
[0, 0.3113]
This specifies respectively the number and the coefficient of the torsional degrees of freedom
(DOF) for the estimation of the change in free energy upon binding, ∆Gbinding. For the purposes of
AutoDock 3.0, the number of torsional DOF is the number of rotatable bonds in the ligand,
excluding any torsions that rotate one or more hydrogen atoms, e.g. hydroxyls, amines, methyls.
By default, the coefficient is 0.3113 kcal mol-1, although the user can override this as necessary.
(Units: none; kcal mol-1).

67
 Appendices

Parameters for ligand internal energies

intnbp_coeffs <float> <float> <integer> <integer>
Respectively: Cn; Cm; n; m. This command specifies the internal pairwise non-bonded energy
parameters for flexible ligands, where:
C Cm
V ( r ) ≈ -----n-n – ------
m
r r

These parameters are needed even if no rotatable bonds were defined in the ligand-PDBQ file.
They are only used in the internal energy calculations for the ligand and must be consistent with
those used in calculating the grid maps. (Units: kcal mol-1 Ån; kcal mol-1 Åm; none; none, respec-
tively).

intnbp_r_eps <float> <float> <integer> <integer>

Respectively: reqm; ε; n; m, This command is an alternative way of specifying the internal pairwise
non-bonded energy parameters for flexible ligands, where AutoDock calculates the pairwise
atomic potential using:
m n
-------------- εr -------------- εr
n m
n – m eqm n – m eqm
V ( r ) ≈ -----------------------
n
- – -----------------------
m
-
r r

The first two arguments specify the equilibrium distance and well depth, epsilon, for the atom
pair. The equilibrium separation has units of Å and the well depth, epsilon, units of kcal mol-1. The
integer exponents n and m must be specified too. Obviously, n ≠ m. (Units: Å; kcal mol-1; none;
none, respectively).

intelec
(Optional) Internal ligand electrostatic energies will be calculated; the products of the partial
charges in each non-bonded atom pair are pre-calculated, and output. Note that this is only rele-
vant for flexible ligands.

Parameters for simulated annealing searches

rt0 <float>
Appendices

[500. cal mol-1].

Initial “annealing temperature”; this is actually the absolute temperature multiplied by the gas
constant R. R = 8.314 J mol-1 K-1 = 1.987 cal mol-1 K-1. (Units: cal mol-1.)

rtrf <float>
Annealing temperature reduction factor, g [0.95 cycle-1]. See the equation at the bottom of page 5.
At the end of each cycle, the annealing temperature is multiplied by this factor, to give that of the

68
 Appendices

next cycle. This must be positive but < 1 in order to cool the system. Gradual cooling is recom-
mended, so as to avoid “simulated quenching”, which tends to trap systems into local minima.

linear_schedule
schedule_linear
linsched
schedlin
These keywords are all synonymous, and instruct AutoDock to use a linear or arithmetic temper-
ature reduction schedule during Monte Carlo simulated annealing. Unless this keyword is given, a
geometric reduction schedule is used, according to the rtrf parameter just described. If the lin-
ear schedule is requested, then any rtrf parameters will be ignored. The first simulated anneal-
ing cycle is carried out at the annealing temperature rt0. At the end of each cycle, the
temperature is reduced by (rt0/cycles). The advantage of the linear schedule is that the sys-
tem samples evenly across the temperature axis, which is vital in entropic calculations. Geometric
temperature reduction schedules on the other hand, under-sample high temperatures and over-
sample low temperatures.

runs <integer>
[10]
Number of automated docking runs.

cycles <integer>
[50]
Number of temperature reduction cycles.

accs <integer>
[100]
Maximum number of accepted steps per cycle.

rejs <integer>
[100]
Maximum number of rejected steps per cycle.

select <character>
[m]
State selection flag. This character can be either m for the minimum state, or l for the last state
found during each cycle, to begin the following cycle.
Appendices

trnrf <float>
[1.0]
Per-cycle reduction factor for translations.

quarf <float>
[1.0]
Per-cycle reduction factor for quaternions.

69
 Appendices

dihrf <float>
Per-cycle reduction factor for dihedrals [1.].

Parameter to set the amount of output

outlev <integer>
[1]
Diagnostic output level. For SA (simulated annealing): 0 = no output, 1 = minimal output, 2 = full
state output at end of each cycle; 3 = detailed output for each step. For GA and GA-LS (genetic
algorithm-local search): 0 = minimal output, 1 = write minimum, mean, and maximum of each
state variable at the end of every generation. Use “outlev 1” for SA, and “outlev 0” for GA and
GA-LS. If you use “outlev 1” with GA-LS, you will generate very large log files.

Parameters for trajectory output during SA dockings

trjfrq <integer>
[0]
Output frequency, n, for trajectory of ligand, in steps. If n = 0, then no trajectory states will be out-
put; otherwise, every nth state will be output. The state consists of 7 floats describing the x,y,z
translation, the x,y,z components of the quaternion unit vector, the angle of rotation about the
quaternion axis; and any remaining floats describing the torsions, in the same order as described
in the input ligand PDBQ file).

trjbeg <integer>
[1]
Begin sampling states for trajectory output at the cycle with this value.

trjend <integer>
[50]
End trajectory output at this cycle.

trjout <string>
[lig.trj]
Trajectory filename. AutoDock will write out state variables to this file every “trjfrq” steps. Use
the “traj” command in AutoDock’s command mode to convert this trajectory of state-variables
into a series of PDB frames. The “traj” command is described in § “Using the Command Mode in
Appendices

AutoDock”; see also § “Trajectory Files”.

trjsel <string>
[E]
Trajectory output flag, can be either ‘A’ or ‘E’; the former outputs only accepted steps, while the
latter outputs either accepted or rejected steps.

watch <string>

70
 Appendices

(Optional) Creates a “watch” file for real-time monitoring of an in-progress simulated annealing
job. This works only if the “trjfrq” parameter is greater than zero. The watch file will be in PDB
format, so give a “.pdb” extension. This file has an exclusive lock placed on it, while AutoDock is
writing to it. Once the file is closed, the file is unlocked. This can signal to a watching visualiza-
tion program that the file is complete and can now be read in, for updating the displayed coordi-
nates. This file is written at exactly the same time as the trajectory file is updated

Parameter for energies of atoms outside the grid

extnrg <float>
[1000.]
External grid energy assigned to any atoms that stray outside the volume of the grid during a
docking. Units: kcal mol-1.

Parameter for initializing the ligand in SA

e0max <float> <positive_integer>
[0., 10000]
This is only used by the simulated annealing method. This keyword stipulates that the ligand’s ini-
tial state cannot have an energy greater than the first value, nor can there be more than the second
value’s number of retries. Typical energy values range from 0 to 1000 kcal/mol. If the initial
energy exceeds this value, a new random state is generated and tested. This process is iterated
until the condition is satisfied. This can be particularly useful in preventing runs starting in excep-
tionally high energy regions. In such cases, the ligand can get trapped because it is unable to take
a long enough translational jump. In those grids were the ligand is small enough to fit into the low
energy regions with ease, there will not be many iterations before a favorable location is found.
But in highly constrained grids, with large ligands, this initialization loop may run almost indefi-
nitely.

Parameters for cluster analysis of docked conformations

rmsref <string>
The root mean square deviation (rmsd) of the docked conformations will be calculated with
respect to the coordinates in the PDB or PDBQ file specified here. This is useful when the experi-
mentally determined complex conformation of the ligand is known. The order of the atoms in this
file must match that in the input PDBQ file given by the move command. These values of rmsd
Appendices
will be output in the last column of the final PDBQ records, after the clustering has been per-
formed.

rmstol <float>
[0.5Å]
rms deviation tolerance for cluster analysis or ‘structure binning’ , carried out after multiple dock-
ing runs. If two conformations have an rms less than this tolerance, they will be placed in the same
cluster. The structures are ranked by energy, as are the clusters. The lowest energy representative

71
 Appendices

from each cluster is output in PDBQ format to the log file. To keep the ligand’s residue number in
the input PDBQ file, use the ‘-k’ flag; otherwise the clustered conformations are numbered incre-
mentally from 1. (Units: Å).

rmsnosym
When more than one run is carried out in a given job, cluster analysis or ‘structure binning’ will
be performed, based on structural rms difference, ranking the resulting families of docked confor-
mations in order of increasing energy. The default method for structure binning allows for atom
similarity, as in a tertiary-butyl which can be rotated by +/-120˚, but in other cases it may be desir-
able to bypass this similar atom type checking and calculate the rms on a one-for-one basis. The
symmetry checking algorithm scans all atoms in the reference structure, and selects the nearest
atom of identical atom type to be added to the sum of squares of distances. This works well when
the two conformations are very similar, but this assumption breaks down when the two conforma-
tions are translated significantly. Symmetry checking can be turned off using the rmsnosym
command; omit this command if you still want symmetry checking.

Parameters for re-clustering the results of several jobs

cluster <string>
(Clustering multi-job output only.) AutoDock will go into ‘cluster mode’. Use this command
only to perform cluster analysis on the combined output, <PDBQfilename>, of several jobs. This
command can be very useful when many jobs have been distributed to several machines and run
in ‘parallel’. The docking parameter file will need the following keywords: rmstol and types;
and optionally write_all_cluster_members and/or rmsnosym.

It is necessary to grep the USER lines along with the ATOM records, since AutoDock parses the
these lines to determine what the energy of that particular conformation was. For more informa-
tion, see the example DPF files given later.

write_all_cluster_members
(Clustering multi-job output only.) This command is used only with the cluster command, to
write out all members of each cluster instead of just the lowest energy from each cluster. This
affects the cluster analysis PDBQ output at the end of each job.

Parameters for genetic algorithm, Lamarckian GA and evolutionary programming searches

ga_pop_size <positive_integer>
Appendices

[50]
This is the number of individuals in the population. Each individual is a coupling of a genotype
and its associated phenotype. Usually, this number is fixed throughout the run. Typical values
range from 50 to 200.

ga_num_evals <positive_integer>
[250000]

72
 Appendices

This is the maximum number of energy evaluations that a GA run should make.

ga_num_generations <positive_integer>
[27000]
This is the maximum number of generations that a GA or LGA run should last.

ga_elitism <integer>
[1]
This is used in the selection mechanism of the GA. This is the number of top individuals that are
guaranteed to survive into the next generation.

ga_mutation_rate <float>
[0.02]
This is a floating point number from 0 to 1, representing the probability that a particular gene is
mutated. This parameter is typically small.

ga_crossover_rate <float>
[0.80]
This is a floating point number from 0 to 1 denoting the crossover rate. Crossover rate is the
expected number of pairs in the population that will exchange genetic material. Setting this value
to 0 turns the GA into the evolutionary programming (EP) method, but EP would probably require
a concomitant increase in the ga_mutation_rate in order to be effective.

ga_window_size <positive_integer>
[10]
This is the number of preceding generations to take into consideration when deciding the thresh-
old for the worst individual in the current population.

ga_cauchy_alpha <float>
[0]

ga_cauchy_beta <float>
[1]
These are floating point parameters used in the mutation of real number genes. They correspond
to the alpha and beta parameters in a Cauchy distribution. Alpha roughly corresponds to the
mean, and beta to something like the variance of the distribution. It should be noted, though, that
the Cauchy distribution doesn’t have finite variance. For the mutation of a real valued gene, a
Appendices

Cauchy deviate is generated and then added to the original value.

Command to set genetic algorithm parameters

set_ga
This command sets the global optimizer to be a genetic algorithm [GA]. This is required to per-
form a GA search. This passes any ’ga_’ parameters specified before this line to the global opti-

73
 Appendices

mizer object. If this command is omitted, or it is given before the ’ga_’ parameters, your choices
will not take effect, and the default values for the optimizer will be used.

To use the traditional genetic algorithm, do not specify the local search parameters, and do not use
the “set_sw1” or “set_psw1” commands.

To use the Lamarckian genetic algorithm, you must also specify the parameters for local
search, and then issue either the ’set_sw1’ or ’set_psw1’ command. The former command uses
the strict Solis and Wets local search algorithm, while the latter uses the pseudo-Solis and Wets
algorithm: see earlier for details about how they differ.

Parameters for local search

sw_max_its <positive_integer>
[50]
This is the maximum number of iterations that the local search procedure apply to the phenotype
of any given individual. This is an unsigned integer. In Bill’s experiments, he used a combination
of iterations and function evaluations. It seems to me, that a value around 30 should be fine.

sw_max_succ <positive_integer>
[4]
This is the number of successes in a row before a change is made to the rho parameter in Solis &
Wets algorithms. This is an unsigned integer and is typically around four.

sw_max_fail <positive_integer>
[4]
This is the number of failures in a row before Solis & Wets algorithms adjust rho. This is an
unsigned integer and is usually around four.

sw_rho <float>
[1.0]
This is a parameter of the Solis & Wets algorithms. It defines the initial variance, and specifies the
size of the local space to sample.

sw_lb_rho <float>
[0.01]
This is the lower bound on rho, the variance for making changes to genes (i.e. translations, orien-
Appendices

tation and torsions). rho can never be modified to a value smaller than “sw_lb_rho”.

ls_search_freq <float>
[0.06]

This is the probability of any particular phenotype being subjected to local search.

74
 Appendices

Commands to choose and set the local search method

Both of these commands, ’set_sw1’ and ’set_psw1’, pass any ’sw_’ parameters set before this
line to the local searcher. If you forget to use this command, or give it before the ’sw_’ keywords,
your choices will not take effect, and the default values for the optimizer will be used.

set_sw1
Instructs AutoDock to use the classical Solis and Wets local searcher, using the method of uni-
form variances for changes in translations, orientations and torsions.

set_psw1
Instructs AutoDock to use the pseudo-Solis and Wets local searcher. This method maintains the
relative proportions of variances for the translations in Å and the rotations in radians. These are
typically 0.2 Å and 0.087 radians to start with, so the variance for translations will always be
about 2.3 times larger than that for the rotations (i.e. orientation and torsions).

Commands to perform automated docking

simanneal
This command instructs AutoDock to do the specifed number of docking runs using the simulated
annealing (SA) search engine. This uses the value set by the “runs” keyword as the number of SA
docking runs to carry out. All relevant parameters for the simulated annealing job must be set first.
These are indicated above by [SA] in each keyword description.

do_local_only <integer>
This keyword instructs AutoDock to carry out only the local search of a global-local search; the
genetic algorithm parameters are ignored, with the exception of the population size. This is an
ideal way of carrying out a minimization using the same force field as is used during the dockings.
The “ga_run” keyword should not be given. The number after the keyword determines how many
dockings will be performed.

do_global_only <integer>
This keyword instructs AutoDock to carry out dockings using only a global search, i.e. the tradi-
tional genetic algorithm. The local search parameters are ignored. The “ga_run” keyword should
not be given. The number after the keyword determines how many dockings will be performed.

ga_run <integer>
Appendices

[10]
This command invokes the new hybrid, Lamarckian genetic algorithm search engine, and per-
forms the requested number of dockings. All appropriate parameters must be set first: these are
listed above by “ga_”.

75
 Appendices

Command to perform clustering of docked conformations

analysis
This performs a cluster analysis on results of a docking, and outputs the results to the log file. The
docked conformations are sorted in order of increasing energy, then compared by root mean
square deviation. If the conformer is within the “rmstol” threshold, it is placed into the same clus-
ter. A histogram is printed showing the number in each cluster, and if more than one member, the
cluster’s mean energy. Furthermore, a table is printed to the docking log file of cluster rmsd and
reference rmsd values.
Appendices

76
 Appendices

Appendix IV: Example Parameter Files

IV 1. AutoGrid GPF

An example AutoGrid parameter file is given below:

______________________________________________________________________________

receptor 3ptb.pdbqs #macromolecule

gridfld 3ptb.maps.fld #grid_data_file
npts 60 60 60 #num.grid points in xyz
spacing .375 #spacing (Angstroms)
gridcenter -1.853 14.311 16.658#xyz-coordinates or ’auto"
types CANH #atom type names
smooth 0.500 #store minimum energy within radius (Angstroms)
map 3ptb.C.map #filename of grid map
nbp_r_eps 4.00 0.0222750 12 6#C-C lj
nbp_r_eps 3.75 0.0230026 12 6#C-N lj
nbp_r_eps 3.60 0.0257202 12 6#C-O lj
nbp_r_eps 4.00 0.0257202 12 6#C-S lj
nbp_r_eps 3.00 0.0081378 12 6#C-H lj
nbp_r_eps 3.00 0.0081378 12 6#C-H lj
nbp_r_eps 3.00 0.0081378 12 6#C-H lj
sol_par 12.77 0.6844 #C atomic fragmental volume, solvation param.
constant 0.000 #C grid map constant energy
map 3ptb.A.map #filename of grid map
nbp_r_eps 4.00 0.0222750 12 6#A-C lj
nbp_r_eps 3.75 0.0230026 12 6#A-N lj
nbp_r_eps 3.60 0.0257202 12 6#A-O lj
nbp_r_eps 4.00 0.0257202 12 6#A-S lj
nbp_r_eps 3.00 0.0081378 12 6#A-H lj
nbp_r_eps 3.00 0.0081378 12 6#A-H lj
nbp_r_eps 3.00 0.0081378 12 6#A-H lj
sol_par 10.80 0.1027 #A atomic fragmental volume, solvation param.
constant 0.000 #A grid map constant energy
map 3ptb.N.map #filename of grid map
nbp_r_eps 3.75 0.0230026 12 6#N-C lj
nbp_r_eps 3.50 0.0237600 12 6#N-N lj
nbp_r_eps 3.35 0.0265667 12 6#N-O lj
nbp_r_eps 3.75 0.0265667 12 6#N-S lj
nbp_r_eps 2.75 0.0084051 12 6#N-H lj
nbp_r_eps 2.75 0.0084051 12 6#N-H lj
nbp_r_eps 2.75 0.0084051 12 6#N-H lj
Appendices
sol_par 0.00 0.0000 #N atomic fragmental volume, solvation param.
constant 0.000 #N grid map constant energy
map 3ptb.H.map #filename of grid map
nbp_r_eps 3.00 0.0081378 12 6#H-C lj
nbp_r_eps 2.75 0.0084051 12 6#H-N lj
nbp_r_eps 1.90 0.3280000 12 10#H-O hb
nbp_r_eps 2.50 0.0656000 12 10#H-S hb
nbp_r_eps 2.00 0.0029700 12 6#H-H lj
nbp_r_eps 2.00 0.0029700 12 6#H-H lj
nbp_r_eps 2.00 0.0029700 12 6#H-H lj

77
 Appendices

sol_par 0.00 0.0000 #H atomic fragmental volume, solvation param.

constant 0.118 #H grid map constant energy
elecmap 3ptb.e.map #electrostatic potential map
dielectric -0.1146 #<0,distance-dep.diel; >0,constant
#fmap 3ptb.f.map #floating grid

_________________________________________________________________

Note how hydrogen bonding is defined for oxygens. If a line in the parameter file contains a
‘10’ in the fourth column, AutoGrid will treat this atom-pair as hydrogen bonding. So in the
example above, the last 3 lines in the “mcp2_O.map” block will be treated as hydrogen bonds.
AutoGrid scans for any polar hydrogens in the macromolecule. The vector from the hydrogen-
donor, along with the vector from the probe-atom at the current grid point, are used to calculate
the directional attenuation of the hydrogen bond. In this example, AutoGrid will calculate H-
bonds between O-H, O-X and O-M.

IV 2. AutoDock DPF

Some examples of commented AutoDock parameter files are given below.

Example 1: Docking using Monte Carlo Simulated Annealing

In this case, the ligand file ‘xk263pm3.pdbq’ has been defined such that it contains 10 rotatable
bonds. The docking will be sampled every 7500 steps, from cycle 45 to cycle 50. Either accepted
or rejected states will be output. The trajectory file ‘xk263pm3.trj’ will hold the state infor-
mation required to generate the coordinates later on. The external grid energy is set to 0.0, which
can allow greater freedom for ligand rotations during docking.

______________________________________________________________________________

seed random
types CNOH # atom type names

fld 4phv.nbc_maps.fld # grid data file

map 4phv.nbc_C.map # C-atomic affinity map
map 4phv.nbc_N.map # N-atomic affinity map
map 4phv.nbc_O.map # O-atomic affinity map
map 4phv.nbc_H.map # H-atomic affinity map
map 4phv.nbc_e.map # electrostatics map
Appendices

move xk263pm3.pdbq # ligand

about -5.452 -8.626 -0.082 # ligand center

tran0 -5.452 -8.626 -0.082 # initial coordinates/A

quat0 1. 0. 0. 0. # initial quaternion:unit-vector(qx,qy,qz);angle/deg(qw)
ndihe 10 # number of rotatable bonds
dihe0 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. # initial dihedrals/deg

tstep 0.2 # translation step/A

qstep 5. # quaternion step/deg

78
 Appendices

dstep 5. # torsion step/deg

trnrf 1. # trans reduction factor/per cycle
quarf 1. # quat reduction factor/per cycle
dihrf 1. # tors reduction factor/per cycle

intnbp_coeffs 1272653.000 1127.684 12 6 # C-C internal energy non-bond parameters

intnbp_coeffs 610155.100 783.345 12 6 # C-N internal energy non-bond parameters
intnbp_coeffs 588883.800 633.754 12 6 # C-O internal energy non-bond parameters
intnbp_coeffs 88604.240 226.910 12 6 # C-H internal energy non-bond parameters
intnbp_coeffs 266862.200 546.765 12 6 # N-N internal energy non-bond parameters
intnbp_coeffs 249961.400 445.918 12 6 # N-O internal energy non-bond parameters
intnbp_coeffs 39093.660 155.983 12 6 # N-H internal energy non-bond parameters
intnbp_coeffs 230584.400 368.677 12 6 # O-O internal energy non-bond parameters
intnbp_coeffs 38919.640 124.049 12 6 # O-H internal energy non-bond parameters
intnbp_coeffs 1908.578 46.738 12 6 # H-H internal energy non-bond parameters

rt0 500. # initial RT

rtrf 0.95 # RT reduction factor/per cycle

runs 10 # number of runs

cycles 50 # cycles
accs 100 # steps accepted
rejs 100 # steps rejected
select m # minimum or last

outlev 1 # diagnostic output level

rmstol 0.5 # cluster tolerance/A

trjfrq 7500 # trajectory frequency

trjbeg 45 # start trj output at cycle
trjend 50 # end trj output at cycle
trjout xk263pm3.trj # trajectory file
trjsel E # A=acc only;E=either acc or rej

extnrg 0.0 # external grid energy

e0max 0.0 # maximum allowable energy to start a run

simanneal # perform automated docking using simulated annealing

analysis # perform a ranked cluster analysis

______________________________________________________________________________

Example 2: Clustering Many Dockings

The next example DPF shows how to use the cluster mode in AutoDock. The PDBQ files contain-
Appendices
ing the final docked conformations have been extracted from the AutoDock log files (using the
UNIX grep command), and stored together in the file “vac1.new.dlg.pdbq”. You can extract the
“DOCKED:” records during the dockings, or after the dockings have finished. For example:

% egrep ’^DOCKED: ’ vac1.*.dlg | sed ’s/^DOCKED: //’ > vac1.grouped.dlg.pdbq

or:

% egrep ’^ATOM |^HETATM|^REMARK|^USER ’ vac1.*.dlg > vac1.grouped.dlg.pdbq

79
 Appendices

The tolerance for the positional rms deviation is set to 1.5Å, so only conformations with this rms
deviation or less will be placed in the same cluster. All conformations will be written out, instead
of just the lowest energy representative from each conformationally distinct cluster.

You may include the rmsnosym command, if you do not wish to use symmetry checking while
clustering. Also, you must finish the DPF with the analysis command, to instruct AutoDock to
perform the clustering and write out the histogram of docked conformations.

______________________________________________________________________________

types CANOH # atom_type_names

rmstol 1.5 # cluster_tolerance/A
write_all # write all conformations in a cluster
cluster vac1.new.dlg.pdbq # structure binning
analysis # do cluster analysis on results
______________________________________________________________________________

Example 3: Docking Using the Lamarckian Genetic Algorithm (LGA)

This DPF shows how to set up a docking using the genetic algorithm (GA) in combination with
the pseudo-Solis and Wets local search algorithm (psw1). This is also known as the Lamarckian
Genetic Algorithm or LGA.

______________________________________________________________________________

seed time pid # for random number generator

types CANH # atom type names
fld 3ptb.maps.fld # grid data file
map 3ptb.C.map # C-atomic affinity map
map 3ptb.A.map # A-atomic affinity map
map 3ptb.N.map # N-atomic affinity map
map 3ptb.H.map # H-atomic affinity map
map 3ptb.e.map # electrostatics map

move benA.pdbq # small molecule

about -1.853 14.311 16.658 # small molecule center
tran0 random # initial coordinates/A or "random"
quat0 random # initial quaternion or "random"
ndihe 0 # number of initial torsions
dihe0 random # initial torsions
torsdof 0 0.3113 # num. non-H tors.degrees of freedom & coeff.
tstep 0.2 # translation step/A
Appendices

qstep 5. # quaternion step/deg

dstep 5. # torsion step/deg
trnrf 1. # trans reduction factor/per cycle
quarf 1. # quat reduction factor/per cycle
dihrf 1. # tors reduction factor/per cycle
intnbp_r_eps 4.00 0.0222750 12 6 #C-C lj
intnbp_r_eps 4.00 0.0222750 12 6 #C-A lj
intnbp_r_eps 3.75 0.0230026 12 6 #C-N lj
intnbp_r_eps 3.00 0.0081378 12 6 #C-H lj
intnbp_r_eps 4.00 0.0222750 12 6 #A-A lj

80
 Appendices

intnbp_r_eps 3.75 0.0230026 12 6 #A-N lj

intnbp_r_eps 3.00 0.0081378 12 6 #A-H lj
intnbp_r_eps 3.50 0.0237600 12 6 #N-N lj
intnbp_r_eps 2.75 0.0084051 12 6 #N-H lj
intnbp_r_eps 2.00 0.0029700 12 6 #H-H lj
outlev 1 # diagnostic output level
rmstol 0.5 # cluster tolerance/A
rmsref benA.pdbq # reference structure for RMS calc.
write_all # write all conformations in a cluster
extnrg 1000. # external grid energy
e0max 0. 10000 # max. allowable initial energy, max. num. retries

ga_pop_size 50 # number of individuals in population

ga_num_evals 150000 # maximum number of energy evaluations
ga_num_generations 27000 # maximum number of generations
ga_elitism 1 # num. of top individuals that automatically survive
ga_mutation_rate 0.02 # rate of gene mutation
ga_crossover_rate 0.80 # rate of crossover
ga_window_size 10 # num. of generations for picking worst individual
ga_cauchy_alpha 0 # ~mean of Cauchy distribution for gene mutation
ga_cauchy_beta 1 # ~variance of Cauchy distribution for gene mutation
set_ga # set the above parameters for GA

sw_max_its 300 # number of iterations of Solis & Wets local search

sw_max_succ 4 # number of consecutive successes before changing rho
sw_max_fail 4 # number of consecutive failures before changing rho
sw_rho 1.0 # size of local search space to sample
sw_lb_rho 0.01 # lower bound on rho
ls_search_freq 0.06 # probability of performing local search on an indiv.
set_psw1 # set the above pseudo-Solis & Wets parameters

ga_run 10 # do this many hybrid GA-LS runs

analysis # do cluster analysis on results

______________________________________________________________________________

Appendices

81
 Appendices

Appendix V: AutoDock References

V 1. Primary References

AutoDock 3.0

Morris, G. M., Goodsell, D. S., Halliday, R.S., Huey, R., Hart, W. E., Belew, R. K. and Olson,
A. J. (1998), J. Computational Chemistry, 19: 1639-1662. "Automated Docking Using a Lama-
rckian Genetic Algorithm and and Empirical Binding Free Energy Function".

ABSTRACT: A novel and robust automated docking method that predicts the bound conformations
of flexible ligands to macromolecular targets has been developed and tested, in combination with a
new scoring function that estimates the free energy change upon binding. Interestingly, this method
applies a Lamarckian model of genetics, in which environmental adaptations of an individual’s
phenotype are reverse transcribed into its genotype and become heritable traits (sic). We consider
three search methods, Monte Carlo simulated annealing, a traditional genetic algorithm, and the
Lamarckian genetic algorithm, and compare their performance in dockings of seven protein-ligand
test systems having known three dimensional structure. We show that both the traditional and
Lamarckian genetic algorithms can handle ligands with more degrees of freedom than the simu-
lated annealing method used in earlier versions of AutoDock, and that the Lamarckian genetic
algorithm is the most efficient, most reliable and most successful of the three. The empirical free
energy function was calibrated using a set of 30 structurally-known protein-ligand complexes with
experimentally-determined binding constants. Linear regression analysis of the observed binding
constants in terms of a wide variety of structure-derived molecular properties was performed. The
final model had a residual standard error of 9.11 kJ mol-1 (2.177 kcal mol-1) and was chosen as
the new energy function. The new search methods and empirical free energy function are available
in AutoDock version 3.0.

AutoDock 2.4

Morris, G. M., Goodsell, D. S., Huey, R. and Olson, A. J. (1996), J. Computer-Aided Molecular
Design, 10: 293-304. "Distributed automated docking of flexible ligands to proteins: Parallel
applications of AutoDock 2.4".

ABSTRACT: AutoDock 2.4 predicts the bound conformations of a small, flexible ligand to a non-
flexible macromolecular target of known structure. The technique combines simulated annealing
for conformation searching with a rapid grid-based method of energy evaluation based on the
AMBER force field. AutoDock has been optimized in performance without sacrificing accuracy; it
Appendices

incorporates many enhancements and additions, including an intuitive interface. We have devel-
oped a set of tools for launching and analyzing many independent docking jobs in parallel on a het-
erogeneous network of UNIX-based workstations. This paper describes the current release, and the
results of a suite of diverse test systems. We also present the results of a systematic investigation
into the effects of varying simulated-annealing parameters on molecular docking. We show that
even for ligands with a large number of degrees of freedom, root-mean-square deviations of less
than 1 A from the crystallographic conformation are obtained for the lowest-energy dockings,
although fewer dockings find the crystallographic conformation when there are more degrees of
freedom.

82
 Appendices

AutoDock 1.0

Goodsell, D. S. and Olson, A. J. (1990), Proteins: Str. Func. and Genet., 8: 195-202. "Auto-
mated Docking of Substrates to Proteins by Simulated Annealing".

V 2. Reviews of Applications

Goodsell, D. S., Morris, G. M. and Olson, A. J. (1996), J. Mol. Recognition, 9: 1-5. "Dock-
ing of Flexible Ligands: Applications of AutoDock".

V 3. Selected Applications and Citations of AutoDock

Minke, W.E., Diller, D.J., Hol, W.G., and Verlinde C. L. (1999), J. Med. Chem., 42: 1778-
1788. "The role of waters in docking strategies with incremental flexibility for carbohydrate
derivatives: heat-labile enterotoxin, a multivalent test case".

Laederach, A., Dowd, M.K., Coutinho, P.M., and Reilly, P.J. (1999), Proteins:Structure,
Function and Genetics, 37:166-175. "Automated Docking of Maltose, 2-Deoxymaltose, and Mal-
totetraose into the Soybean beta-Amylase Active Site".

Matias, P. M., Saraiva, L. M., Soares, C. M., Coelho, A. V., LeGall, J., and Armenia Car-
rondo, M. (1999) JBIC, 4: 478-494. "Nine-haem cytochrome c from Desulfovibrio desulfuricans
ATCC 27774: primary sequence determination, crystallographic refinement at 1.8 Å and model-
ling studies of its interaction with the tetrahaem cytochrome c3".

Bitomsky, W. and Wade, R. C. (1999), J. Am. Chem. Soc., 121: 3004-3013. "Docking of
Glycosaminoglycans to Heparin-Binding Proteins: Validation for aFGF, bFGF, and Antithrombin
and Application to IL-8".

Lorber, D. M. (1999), Chemistry & Biology, 6: R227-R228. "Computational drug design".

Rao, M. S. and Olson, A. J. (1999), Proteins:Structure, Function and Genetics, 34: 173-83.
"Modelling of factor Xa-inhibitor complexes: a computational flexible docking approach".

Heine, A., Stura, E.A., Yli-Kauhaluoma, J.T., Gao, C., Deng, Q., Beno, B.R., Houk, K.N.,
Appendices

Janda, K.D., and Wilson, I.A. (1998), Science, 279: 1934-1940. "An antibody exo Diels-Alderase
inhibitor complex at 1.95 Å resolution".

Coutinho, P. M., Dowd, M. K., Reilly, P. J., (1998), Industrial & Engineering Chemistry
Research, 37: 2148-2157."Automated Docking of -(1,4)- and -(1,6)-Linked Glucosyl Trisaccha-
rides in the Glucoamylase Active Site."

Lozano, J. J., Lopez-de-Brinas, E., Centeno, N.B., Guigo, R. and Sanz, F. (1997), J. Com-

83
 Appendices

puter-Aided Molecular Design, 11: 395-408. "Three-dimensional modelling of human cyto-

chrome P450 1A2 and its interaction with caffeine and MeIQ".

Mahmoudian, M. (1997), J. Molecular Graphics & Modelling, 15:149-153. "The cannabinoid

receptor: Computer-aided molecular modeling and docking of ligand".

Coutinho, P. M., Dowd, M. K. and Reilly, P. J. (1997), Proteins: Str. Func. and Genet., 28:162-
173. "Automated Docking of Glucosyl Disaccharides in the Glucoamylase Active Site".

Coutinho, P. M., Dowd, M. K. and Reilly, P. J. (1997), Proteins: Str. Func. and Genet., 27:235-
248. "Automated Docking of Monosaccharide Substrates and Analogues and Melthyl alpha-
Acarviosinide in the Glucoamylase Active Site".

Neurath, A. R., Jiang, S., Strick, K. L., Li, Y.-Y., and Debnath, A. K. (1996), Nature Medicine,
2:230-234. "Bovine beta-lactoglobulin modified by 3-hydroxyphthalic anhydride blocks the CD4
cell receptor for HIV".

Gamper AM, Winger RH, Liedl KR, Sotriffer CA, Varga JM, Kroemer RT, Rode BM. (1996),
J. Med. Chem., 39, 3882-3888. "Comparative molecular field analysis of haptens docked to the
multispecific antibody IgE".

Sotriffer, C. A., Liedl, K. R., Winger, R. H., Gamper, A. M., Kroemer, R. T., Linthicum, D. S.,
Rode, B.-M. and Varga, J. M. (1996) Molecular Immunology, 33: 129-144. "Heteroligation of a
mouse monoclonal IgE antibody (La2) with small molecules, analysed by computer-aided auto-
mated docking".

Zhang, T. and Koshland, D. E. (1995 ), Protein Science, 4: 84-92. "Modeling substrate binding
in Thermus thermophilus isopropylmalate dehydrogenase".

Kedishvili, N. Y., Bosron, W. F., Stone, C. L., Hurley, T.D., Peggs, C. F., Thomasson, H. R.,
Popov, K. M., Carr, L. G., Edenberg, H. J. and Li, T.-K. (1995) J. Biol. Chem., 270: 3625-3630.
"Expression and kinetic characterization of recombinant human stomach alcohol dehydrogenase".

Stone, C. L., Hurley, T. D., Peggs, C. F., Kedishvili, N. Y., Davis, G. J., Thomasson, H. R., Li,
T.-K. and Bosron, W. F. (1995) Biochemistry, 34: 4008-4014. "Cimetidine inhibition of human
gastric and liver alcohol dehydrogenase isoenzymes: identification of inhibitor complexes by
kinetic and molecular modeling".
Appendices

Tummino, P. J., Ferguson, D., Jacobs, C. M ., Tait, B., Hupe, L., Lunney, E. and Hupe, D.
(1995) Arch. Biochem. Biophys., 316: 523-528. "Competitive inhibition of HIV-1 protease by
biphenyl carboxylic acids".

Friedman, A. R., Roberts, V. A. and Tainer, J. A. (1994) Proteins: Str. Func. and Genet., 20:
15-24. "Predicting molecular interactions and inducible complementarity: fragment docking of
Fab-peptide complexes".

84
 Appendices

Lunney, E. A., Hagen, S. E., Domagala, J. M., Humblet, C., Kosinski, J., Tait, B. D., Warmus,
J. S., Wilson, M., F erguson, D., Hupe, D., Tummino, P. J., Baldwin, E. T., Bhat, T. N., Liu, B. and
Erickson, J. W. (1994) J. Med. Chem., 37 : 2664-2677. "A novel nonpeptide HIV-1 protease inhib-
itor: elucidation of the binding modes and its application in the design of related analogs".

Vara Prasad, J. V. N., Para, K.S., Ortwine, D. F., Dunbar, Jr., J. B., Ferguson, D., Tummino, P.
J., Hupe, D., Tait, B. D., Domagala, J. M., Humblet, C., Bhat, T. N., Liu, B., Guerin, D. M. A.,
Baldwin, E. T., Erickson, J. W. and Sawyer, T. K. (1994) J. Am. Chem. Soc., 116: 6989-6990.
"Novel series of achiral, low molecular weight, and potent HIV-1 protease inhibitors".

Stoddard, B. L. and Koshland, Jr., D. E. (1993) Proc. Natl. Acad. Sci. USA 90: 1146-1153.
"Molecular recognition analyzed by docking simulations: The aspartate receptor and isocitrate
dehydrogenase from Escherichia coli".

Jeffery, C. J. and Koshland, Jr., D. E. (1993) Protein Science 2: 559-566. "Three-dimensional

structural model of the serine receptor ligand binding domain".

Goodsell, D. S., Lauble, H., Stout, C. D. and Olson, A. J. (1993) Proteins: Str. Func. and
Genet., 17: 1-10. "Automated Docking in Crystallography: Analysis of the Substrates of Aconi-
tase".

Stoddard, B.L. and Koshland, Jr., D.E. (1992) Nature, 358: 774-776. "Prediction of a receptor
protein complex using a binary docking method".

V 4. Web publications

Lunney, E. (1995) Network Science, 1: http://www.netsci.org/Science/Cheminform/

feature01.html. "Structure-Based Design and Two Aspartic Proteases".

Appendices

85
 Appendices
Appendices

86