The DOCK runs presented in the subdirectories of the 3dfr directory of the examples directory serve two functions: a check to see that locally obtained results agree with the expected results; and examples of how to implement certain new features. In each subdirectory exists an INDOCK, OUTDOCK, and results file for one or more docking runs utilitizing various features. These are provided for sake of instruction and comparison. To re-run any particular example, copy the desired prefixed INDOCK to INDOCK. Make sure no OUTDOCK or other output files exist and re-run dock3.5.

The 00README file in each directory provides further assistance and descriptions of directory contents. In the main 3dfr directory, please read the 00README file for the procedures used to create each file in this directory.

• mindeg directory

  This directory gives examples of the usage of Force-Field Score Optimization and Degeneracy Checking in SINGLE mode. Please consult these sections in the documentation for further implementation details.

• search directory

  This directory gives examples of SEARCH mode DOCK runs using either force-field scoring or contact scoring. A small, roughly 100-compound section of the Cambridge Structural Database is used.

• colcrit directory

  This directory gives examples of the usage of Chemical Matching and Critical Clusters in SINGLE mode. Please consult these sections in the documentation for further implementation details. This section gives an extensive example of how these features are employed and the power they provide:

Application of Chemical Matching and Critical Clusters

Using DOCK to search for receptor ligand binding modes can prove to be a computationally expensive exercise. To help alleviate this problem and direct DOCK searches, two new functions have been devised to restrict the number of potential node matches that can occur between ligand and receptor.

The first of these functions is called chemical matching (coloring). This technique allows the user to define atom and receptor centers as being of a specific property type e.g. hydrophobic. If chemical matching is utilized, only centers with complementary chemistry are allowed to match, which reduces the number of orientations being tested, and emphasizes chemical complementarity between ligand and receptor.

The second technique is known as critical clustering. This function allows the user to define regions thought to be crucial to the binding process. When critical clusters are defined, the user forces DOCK to include at least one member of each cluster within a valid match. This can dramatically reduce the number of permitted matches, and hence speed up the DOCK run.

• Ligand Chemical Matching

  The methotrexate structure chemistry was defined using the file keymtx as input to the mol2db program (mol2db < keymtx). This file provides a useful example of how to configure your ligand chemical matching definitions for DOCK databases. Note that the chemistry definitions used are not gospel, and can be altered to suit user requirements. Note also the use of the property boring. This is used to define centers with no particular property, and is used because if centers have no defined chemistry, they may be used to match center of any chemistry. Assigning a specific property type to these centers prevents this from happening.

• Receptor Chemical Matching

  The receptor center file (generally referred to as the sphgen output) was created using both the GRID program of Peter Goodford (1985) and sphgen. GRID was used to map 3dfr using both carboxyl oxygen anion and sp3 nitrogen cation probes. The resulting maps were searched in the following manner. Determine the point with the most favorable interaction energy. Exclude all points within 1Å of this point, and find the next best interaction energy. Repeat until no points with an interaction energy better than -7 kcal/mol remain. The resulting points were used to define hydrogen bond donor and acceptor sites. GRID execution and data extraction was accomplished using the hbdata program. Only points within 4Å of a sphgen sphere were retained. The sphgen centers determined for 3dfr found to be further than 2Å from the donor and acceptor points were defined as hydrophobic centers. All other sphgen centers were removed. The choice of points was undertaken within the SYBYL modeling package, and the results were written out as a MOL2 file. The script mol2sph was then used to create the resulting sphgen format file. GRID was used in place of sphgen to define the hydrogen bond regions, since sphgen does not use hydrogen bonding properties when defining its centers, so it is unlikely that the resulting site points will best describe hydrogen bonding regions. Centers can be defined as the user sees fit, e.g. using LUDI, HINT or other such programs. As long as the final center file is in sphgen format, DOCK will not complain.

  Alternatives to sphgen for generating site points.

• Files and Results

  Four sets of files have been created, all prefixed with a descriptive tag indicative of what features were employed:

  Prefix	Minimization?	Chemical Matching (*)	Critical Clusters (**)	Search Time (***)
  dock_min	yes	no	no	1140
  dock_crit	yes	no	yes	155
  dock_col	yes	yes	no	45
  dock_colcrit	yes	yes	yes	15

  (*) chemical matching constraints applied
  (**) centers around the known binding positions of the methotrexate pteridine ring amines defined as separate critical clusters
  (***) CPU seconds spent in docking stage (overhead of reading grids not included), executed on an SGI 50MHz R4000
  

  This data comes from the OUTDOCK files with corresponding prefixes. Note that the rms values given in the OUTDOCK file are meaningless, since when the mtx.db file is created, all the methotrexate coordinates are shifted to the positive coordinate octant.

  The relative times required for this search show the potential value of chemical matching and critical clusters. All four runs generated the crystallographically observed orientation as the highest scoring receptor ligand interaction, but applying constraints produces on the order of a 5 to 100 fold speed increase in DOCK search speed. To run DOCK using one of the above examples, copy the relevant input file to a file with the name INDOCK. Remove the file OUTDOCK if it exists, and then run DOCK.