Subsections

4.2 Docking

By default FRED returns a single docked structure for each molecule in the input database. FRED uses the following process to generate this pose within the active site defined by the user.

Exhaustive Docking
1. Enumerate all possible poses of the ligand around the active site by rigidly rotating and translating each conformer within the site.
2. Filter the resulting pose ensemble by rejecting poses that do not fit within the larger of the two volumes specified by the receptor file's shape potential grid and a contour level (also referred to as the outer contour).
3. Filter the resulting pose ensemble by rejecting poses that do not have a least one heavy atom within the smaller of the two volumes specified by the receptor file's shape potential grid and a contour level (also referred to as the inner contour).
4. Filter the pose ensemble by rejecting poses that do not match any user-defined docking constraints.
5. Rank all remaining poses using either the Shapegauss, PLP, Chemgauss2, Chemgauss3 or CGO scoring functions (described in section 4.3). Retain N top scoring candidate poses and discard the rest (by default N is 100).
Perform a systematic solid body optimization of the top ranked candidate poses using either Shapegauss, PLP, Chemgauss2, Chemgauss3, CGO, CGT, Chemscore, OEChemscore or Screenscore.
Rank poses via the Consensus Structure method and discard all but the top ranked poses, unless the user requests that FRED retain alternate poses (in which case as many alternate poses as requested are retained up to the number of candidate poses).
Force Field refinement
1. Do a full coordinate optimization of the pose vs. the MMFF force field. If alternate poses were retained in the previous step those poses are refined as well.
2. Check that the refined pose passed the user-defined constraints (if they are specified), and discard the pose if it does not.
Note The force field refinement step is skipped by default.

These steps are described in more detail in this section.

4.2.1 Exhaustive Docking

The purpose of Exhaustive Docking is to take a multiconformer ligand and generate N candidate poses of the ligand within the receptor site (by default N is 100).

4.2.1.1 Enumerate all possible posess

The first step of the docking processes is the generation of the pose ensemble. The ensemble is constructed by enumerating rigid rotations and translations of each conformer within the active site. The translations are generated by systematically translating the conformer within the active site using a specified step size. Rotations are generated such that a single rotational step does not produce a displacement of any atom greater than a specified rotational step size.

The default translational step size is 1 Angstrom and the default rotational step size is 1.5 Angstroms, which roughly gives a 1 RMSD change for any rotational step. The ensemble size can range from tens of millions of poses, in the case of large active sites and highly flexible molecules, to a few thousand poses for small enclosed sites and rigid molecules.

4.2.1.2 Inner and Outer Contour Filter

The pose ensemble is filtered by an inner and an outer contour filter. These filters reject poses that do not have sufficient shape complementarity to the protein's active site. All heavy atoms of the pose must fit within the outer contour and at least one heavy atom of the pose must fit within the inner contour.

Both complementary volumes are created by creating two isocontours of the same shape potential grid contained in the receptor file at different contour levels. The outer contour level is generally low, resulting in a large volume (typically around 1500 cubic Angstroms), while the inner contour level is high, resulting in a small volume (typically around 50 cubic Angstroms). Obviously more pose atoms will fit within the large outer contour than the inner contour, however to satisfy the outer contour filter every heavy atom of the pose must fit within the outer volume, while satisfying the inner contour filter requires only that one heavy atom fit within the volume.

Note that both the inner and outer contour filters can be disabled by the user. In this case FRED will filter poses based on clashes with the protein structure. You can also enable the clash checking in combination with these filter by using the -clash_scale parameter, although in general this is not required as the outer contour filter has a shape that rejects clashes.

4.2.1.3 User defined constraint filter

FRED now has two types of user defined constraints it will filter the pose ensemble with, protein constraints and custom constraints. Constraint information is stored within the receptor file and there can be any number of each constraint type. Any pose that does not meet all of the user specified constraints is removed from the pose ensemble.

Custom constraints

are user defined constraints that use spheres and associated SMARTS patterns to specify regions within the active site where certain chemical functionality is required. FRED will reject any poses that do not match this functionality. Each custom constraint is referred to as a constraint feature.

Each constraint feature consists of one or more spheres, and optionally a list of SMARTS patterns. A feature without a SMARTS patterns will be satisfied if any heavy atom of the pose falls within one of the feature's spheres. If the feature has SMARTS pattern(s) only atoms which match the SMARTS pattern(s) can satisfy the constraint.

Protein constraints

are new to FRED 2.2 and requires use of the receptor setup GUI to use. Protein constraints are placed on individual protein atoms and come in three basic types:

Hydrogen Bond: constraints tell FRED that a pose must make a hydrogen bond interaction with the specified protein atom to pass the constraint filter. These constraints have more geometric specificity than custom constraints (they are inherently directional), and can be specified as either acceptor or donor constraints (or both). They function by recognizing "lone pair" and "polar hydrogen" positions around acceptor and donor atoms respectively and require that one of the acceptor's "lone pair" positions is within 1.0 Angstrom of the donor's "polar hydrogen" position. Note that the actual position of a donor's polar hydrogen is not used, rather FRED generates its own set of likely polar hydrogen positions.
Metal: constraints tell FRED that a pose must make a coordinating interaction with the metal the constraint is placed on to pass the constraint filter (metals are treated as being part of the protein by FRED). These constraints have more geometric specificity than custom constraints. Similar to hydrogen bond constraints, metal constraints work by defining "coordinating positions" around a coordinating atom, and requiring that a "coordinating position" be within 1.0 Angstrom of the metal.
Contact: constraints tell FRED that a pose must make a contact interaction with the atom the constraint is placed upon to pass the constraint. Contact is defined as having a heavy atom within 4 Angstroms of the protein atom the constraint is placed on. These constraints have the same geometric specificity as custom constraints (i.e. making a custom constraint with a 4 Angstrom radius centered on the protein atom will perform the same function).

Mini Constraint F.A.Q.

Which constraint type should I use? In general protein constraints are designed for simplicity and constrain the geometry of hydrogen bond and metal constraints more realistically than custom constraints can. Custom constraints on the other hand are more flexible in the sense that they allow users to specify their own chemistry required to satisfy the constraint, via SMARTS patterns.

Is there a maximum number of constraints I can use? There is no direct limit on the number of constraints that the user can specify, however there is a modest increase in memory requirements for each additional constraint specified. Fred is very efficient at restricting the search space of possible poses based on user defined constraints. Accordingly adding constraints will generally decrease run time.

Why do some of the poses FRED is generating violate my constraints? The constraint spheres are mapped onto a grid during the docking process, and the resulting interpolation error can allow atoms slightly outside a sphere (approximately 1/2 the translational stepsize or 0.5A by default), to satisfy a constraint.

4.2.1.4 Ranking

All poses of the ensemble that pass the previous filtering steps are scored by the exhaustive scoring function (see flag -exhaustive_scoring), which can be either Chemgauss3 (the default, see section 4.3.5), Chemgauss2 (see section 4.3.4), Shapegauss (see section 4.3.2), PLP (see section 4.3.3) or CGO (see section 4.3.10). The poses are then ordered by score and the top N scoring poses are retained (N is 100 by default, see also the flag -num_poses). Accordingly this list of N poses may contain several different poses for the same conformer and does not necessarily contain a pose for any given input conformer.

4.2.2 Optimization

The top ranked poses from Exhaustive docking (by default 100) may be optimized using a systematic solid body optimization against either Shapegauss, PLP, CGO, CGT Chemgauss2, Chemgauss3, Chemscore, OEChemscore or Screenscore. Alternatively the optimization may be skipped (by default Chemgauss3 optimization is used). The systematic solid body optimization is done by rigidly rotating and translating the molecule at half the stepsize used in the Exhaustive Docking, 4.2.1. One positive and negative step is taken in each translational and rotational direction, so 729 (i.e. 27x27) poses are tested from which the optimal one is selected.

4.2.3 Consensus Structure

The poses returned from exhaustive docking (and optional optimization) are scored by one or more scoring functions (Shapegauss, PLP, Chemgauss, Chemgauss2, Chemscore, Screenscore, CGO or CGT). For each scoring function a list of the poses is created ordered by rank. Each pose is then assigned a consensus structure score equal to the sum that pose's rank in each list. The pose with the top consensus structure score is then retained and all other poses are discarded, unless the user has requested that alternate poses be saved (see flag -num_alt_poses) in which case poses are ordered by consensus structure score and a number of poses up to one plus the number of alternate poses requested are passed on.

Note that the user can specify the weights that different scoring functions are given in this calculation. See the flags -pose_select_weight_xxx, where xxx is the name of the scoring function.

4.2.4 Force Field Refinement

Optionally after consensus scoring, poses can be refined using the Merck Molecular Mechanics Force Field. The refinement is full coordinate optimization of all ligand atoms (the protein is held rigid). This step is optional and very CPU intensive, and it is only recommended when using the Zapbind scoring function which is very sensitive to small atom-atom clashes.

User defined constraints, see section 4.2.1, are ignored during the refinement process. However, by default after the refinement process any poses that violate the constraints are discarded (if this is the only pose the entire molecule will be discarded and not appear in the output hitlists at all).

Note by default Force Field Refinement and the constraint re-checking are skipped, see the flag -refine if you wish to turn this step on.