Introduction to Auto-Docking
The ability to predict the conformation of a ligand in a protein or nucleic acid binding site is
of central importance in pharmaceutical design. Detailed knowledge of the molecular
recognition forces is necessary for designing new ligands. Knowing the binding site
conformation helps to show the important interactions that stabilize the complex. Automated
procedures are very helpful in exploring a wide variety of binding modes in complex
systems. The advent of rapid parallel binding assays through combinatorial chemistry and
high throughput screening also motivates a similar need for efficient, rapid computational
screening procedures. The ability to computationally predict possible tight-binding ligands
lowers the time and energy necessary to plan a major combinatorial study.
One approach to auto-docking is to use molecular mechanics techniques to minimize the
energy of a ligand in a binding pocket starting from a series of random positions and
orientations within the binding site. Finding the low energy binding conformations will then
help to highlight the interactions that are responsible for tight binding. For example, the
hydrogen bonds in the low energy complexes can be displayed. Electrostatic surfaces of the
binding pocket and the ligand can also be displayed to observe how the distribution of charge
in the ligand and binding pocket complement each other. Several methods have been
developed to efficiently sample possible binding conformations.
The steps in docking are outlined below. The complete step-by-step instructions follow:
Docking Summary:
1. Select the ligand.
2. Build the Docking box to delineate the region in the binding pocket to place the ligand.
3. The program then does the calculations in two steps:
Step 1: Generate random starting orientations.
Step 2: Use random small moves to find a low energy conformation.
4. Extract the lowest energy conformer (or a few low energy conformers).
5. Minimize the resulting models using normal molecular mechanics.
Placement Methodology
The Placement phase of Dock generates poses from ligand conformations. The available
methods are:
Method
Description
Alpha Triangle
Poses are generated by superposition of ligand atom triplets and
triplets of receptor site points. The receptor site points are alpha
sphere centers which represent locations of tight packing. At
each iteration a random conformation is selected, a random
triplet of ligand atoms and a random triplet of alpha sphere
centers are used to determine the pose.
Alpha PMI
Poses are generated by aligning ligand conformations' principal
moments of inertia to a randomly generated subset of alpha
spheres in the receptor site. This method is most suited to tight
pockets, is quite fast and the search space is greatly truncated.
Flex*
Flex* is a third party docking program, which is implemented in
MOE as if it were a type of placement. A separate license from
BioSolveIT is required to run Flex*. Configuration and
operation of Flex* is described later. Flex* uses an incremental
fragment growth strategy to achieve its function.
Pharmacophore
Information from the pharmacophore is directly used to place
the ligand in the binding site. This is the preferred method when
a pharmacophore is present, but will not work if there is no
pharmacophore.
Proxy Triangle
This method was developed for tackling medium to large sized
ligands which have a very large number of conformers. For
small ligands, the Triangle Matcher method will be called. For
medium and large ligands, conformers are pre-superposed prior
to being placed into the binding site, thus saving computational
time. For large ligands, the score for internal ranking will use
atom representatives instead of all the ligand atoms. More time
is allowed for larger ligands.
Poses are generated by aligning ligand triplets of atoms on
Triangle Matcher
triplets of alpha spheres in a more systematic way than in the
(default)
Alpha Triangle method.
None
This placement method is a special pass-through technique: the
initial coordinates are emitted unmodified. Choose this method
when you wish to use Dock's refinement and scoring tools to
postprocess already docked poses, either from an earlier Dock
run or from other sources.
Note! While the placement stage produces identical results under identical set up conditions,
it is normal for the output to vary from run to run if there is a refinement stage. This is
because of the irreproducible nature of the energy minimization process.
Scoring
For all scoring functions, lower scores indicate more favourable poses. The unit for all
scoring functions is kcal / mol.
ASE Scoring. This score is proportional to the sum of the Gaussians R1R2exp(-0.5d2) over
all ligand atom - receptor atom pairs and ligand atom - alpha sphere pairs. R1 and R2 are the
radii of the atoms in Å, or is -1.85 for alpha spheres. d is the distance between the pair in Å.
The proportionality constant is default to be 0.035 kcal / mol but can be changed upon
pressing the Configure... button.
Affinity dG Scoring. This function estimates the enthalpic contribution to the free energy of
binding using a linear function:
G = Chb fhb + Cion fion + Cmlig fmlig + Chh fhh + Chp fhp + Caa faa
where the f terms fractionally count atomic contacts of specific types and the C's are
coefficients that weight the term contributions to the affinity estimate. The individual terms
are:
Term
hb
Description
Interactions between hydrogen bond donor-acceptor pairs. An optimistic
view is taken; for example, two hydroxyl groups are assumed to interact in
the most favorable way.
ion
Ionic interactions. A Coulomb-like term is used to evaluate the interactions
between charged groups. This can contribute to or detract from binding
affinity.
mlig
Metal ligation. Interactions between Nitrogens/Sulfurs and transition metals
are assumed to be metal ligation interactions.
hh
Hydrophobic interactions, for example, between alkane carbons; these
interactions are generally favorable.
hp
Interactions between hydrophobic and polar atoms; these interactions are
generally unfavorable.
aa
An interaction between any two atoms. This interaction is weak and
generally favorable.
The coefficients may be adjusted by pressing the Configure... button on the Dock panel.
Alpha HB Scoring. This score is a linear combination of two terms. The first term measures
the geometric fit of the ligand to the binding site. The second term measures hydrogen
bonding effects. Both terms are summed over all ligand atoms. The weights of the two terms
may be adjusted through a panel upon pressing the Configure... button.
The first term has an attractive part and a repulsive part. The attractive part is summed over
atoms in the ligand. Each ligand atom that is within 3 Å of an alpha sphere center contributes
A exp(-0.5d2) to this term, where d is the distance from the ligand atom to the nearest alpha
sphere center, and A assumes the value of -0.6845. The repulsive part is summed over all
pairs of atomic overlap between the ligand and the receptor. For each pair of overlap, the
contribution is between 0 and 1 depending on the severity of the overlap.
The second term measures hydrogen bond effects. For non-sp3 donors and acceptors,
hydrogen bonding sites are projected from the atom. If the site is occupied by a favorable
atom, there is a score of -2 (negative means favorable). Otherwise if it is occupied by some
other ligand atom, there is a score of +1. For sp3 donors and acceptors, all favorable atoms
within 3.5 Å contribute a score of -1 while all other atoms contribute +1. Metals in the
receptor are treated as acceptors but with a three-fold effect.
London dG Scoring (default). The London dG scoring function estimates the free energy of
binding of the ligand from a given pose. The functional form is a sum of terms:
where c represents the average gain/loss of rotational and translational entropy; Eflex is the
energy due to the loss of flexibility of the ligand (calculated from ligand topology only); fHB
measures geometric imperfections of hydrogen bonds and takes a value in [0,1]; cHB is the
energy of an ideal hydrogen bond; fM measures geometric imperfections of metal ligations
and takes a value in [0,1]; cM is the energy of an ideal metal ligation; and Di is the desolvation
energy of atom i. The difference in desolvation energies is calculated according to the
formula
where A and B are the protein and/or ligand volumes with atom i belonging to volume B; Ri is
the solvation radius of atom i (taken as the OPLS-AA van der Waals sigma parameter plus
0.5 Angstrom); and ci is the desolvation coefficient of atom i. The coefficients {c, cHB, cM, ci}
were fitted from ~400 x-ray crystal structures of protein-ligand complexes with available
experimental pKi data. Atoms are categorized into ~12 atom types for the assignment of the
ci coefficients. The triple integrals are approximated using Generalized Born integral
formulas.
Refinement
The Forcefield refinement scheme is more accurate than the GridMin scheme but it also takes
more time.
Forcefield Refinement. Energy minimization of the system is carried out using the
conventional molecular mechanics setup. The current prevailing molecular mechanics
forcefield will be used. However, some of the set up options, like the cutoff settings, will be
overridden during the refinement process. The charges of all atoms will be assigned using the
current forcefield. During the course of the refinement, solvation effects are treated by using
the reaction field functional form for the electrostatic energy term. By default, the final
energy is evaluated using the Generalized Born solvation model (GB/VI) but this can be
changed (see below).
In order to speed up the calculation, residues (or atoms, depending on the setting) over a
cutoff distance (Pocket Cutoff on the Configure panel, upon pressing the Configure... button)
away from all the pre-refined poses are ignored, both during the refinement and in the final
energy evaluation. However, if after the refinement any poses moved to within 80% of the
cutoff distance of any ignored atoms, the refinement stage will be repeated with an increased
cutoff distance.
In the presence of a pharmacophore defined using the Unified Annotation Scheme, restraints
will be set up between the features on the ligand and the corresponding pharmacophore query
points. If the distance between the feature and the query point exceeds a certain threshold (see
below) by an amount x, there will be a penalty in the potential of the form kx2. The threshold
is the radius associated with the pharmacophore query point minus the Radius Offset
parameter on the Configure panel, or zero if this results in a negative value. The force
constant k is also adjustable via the Configure panel.
The following are the default settings:




Receptor residues over 6 Å (the Pocket Cutoff parameter) away from all of the prerelaxed poses are ignored.
All receptor atoms will be held fixed during the energy minimization.
The final energy will be calculated using the Generalized Born solvation model
(GB/VI , see MOE Potential Energy Model).
For the pharmacophore restraints, the force constant k is set to 10.0 kcal / (mol Å2)
and the Radius Offset is set to 0.4 &Aring.
These defaults may be altered through a panel upon pressing the Configure... button.
If GB/VI scoring is disabled, the final energy will simply be calculated in the same way as
during the course of minimization instead of using the Generalized Born solvation model. If a
flexible receptor is desired, the sidechains of the receptor can be tethered. However,
backbone atoms are always held fixed.
GridMin Refinement. In order to speed up the energy minimization, a grid is used for
electrostatic calculations during the minimization process. The distance dependent dielectric
model is used. In contrast, van der Waals interactions are treated explicitly. However, the
final electrostatic energy is calculated using the explicit Coulomb form instead of the grid. If
an energy-minimized pose lies within 2 Å of any receptor atom or lies outside the site, a huge
energy will be assigned to the pose.
In the presence of a pharmacophore defined using the Unified Annotation Scheme, restraints
will be set up between the features on the ligand and the corresponding pharmacophore query
points. If the distance between the feature and the query point exceeds a certain threshold (see
below) by an amount x, there will be a penalty in the potential of the form kx2. The threshold
is the radius associated with the pharmacophore query point minus the Radius Offset
parameter on the Configure panel, or zero if this results in a negative value. The force
constant k is also adjustable via the Configure panel.
The following are the default settings:




Receptor atoms within 4 Å of any pre-relaxed pose will be included for the van der
Waals energy calculations.
Receptor residues within 5.5 Å of any pre-relaxed pose will be included for the
electrostatic calculations.
The van der Waals interaction formula will assume the standard 6-12 form, as
opposed to a softer 4-8 form.
For the pharmacophore restraints, the force constant k is set to 10.0 kcal / (mol Å2)
and the Radius Offset is set to 0.4 &Aring.
These defaults may be altered through a panel upon pressing the Configure... button.
Duplication Removal
In the Dock application, duplication removal is not performed using the conventional rmsd
cutoff approach. Poses are considered duplicated if they have the same hydrogen bonding
pattern and hydrophobic interaction pattern with the receptor. A pair of ligand-receptor atoms
of the appropriate types are considered to have a hydrogen bond interaction if they are less
than 3.5 Å apart. A hydrophobic atom on the ligand is considered to have a hydrophobic
interaction with a receptor residue if that atom is under 4.5 Å away from a hydrophobic atom
of that residue. Poses are considered duplicated if the same set of ligand-receptor atom pairs
are involved in hydrogen bond interactions and the same set of ligand atom receptor residue
pairs are involved in hydrophobic interactions.