DOCK 3.5

Jump to navigation Jump to search

DOCK 3.5 is no longer available. A substantial modification of DOCK 3.5, now called DOCK 3.6, continues to be used, maintained, and distributed by the Shoichet Laboratory.
DOCK 3.5, released in 1993-4, superceded DOCK 3 and all previous versions of DOCK. It was written in Fortran 77.

General operation changes to DOCK3.5

• Everything that happens after the generation of a rotation-matrix has been completely re-written. This includes applying the rotation-matrix, scoring the atoms, sorting to get the hit list, saving the hit list, and writing output. The way ligands are read in has also been re-written.

• Northwestern DOCK accepts the name of an INDOCK file as argument on the command line. If no argument is provided, INDOCK is assumed.

• Northwestern DOCK saves the atom coordinates for molecules in the hit list instead of the rotation-matrices needed to regenerate the orientation. This was initially done to avoid the time-consuming process of re-reading the database to get the best molecules. An additional benefit to this method is that at any time during the run, the best molecules can easily be visualized.

• Focusing has been removed as implemented in DOCK3.5 and re-implemented based on Brian’s work with DOCK2.5. Coloring with focusing has not been thoroughly tested but seems to work.

• Every conformation and every orientation of small molecule ligands [Shoichet, Leach, & Kuntz 1999] and mutated protein-ligand resides can be completely desolvated at essentially no computational cost.

• Currently the only valid scoring option in Northwestern DOCK is 5. This uses the CHEMGRID generated van der Waals grid, DISTMAP generated contact grid, and a DelPhi generated electrostatics grid of dimensions 65×65×65. Re-implementing CHEMGRID electrostatics should be trivial.

Simplex Minimization: Three values are accepted for the minimize parameter.

• None (or no) -- does no energy minimization, is fast, and frequently produces qualitatively correct hit lists. My tests suggest that while minimizing produces better scoring molecules that make more hydrogen bonds than their non-minimized counterparts, minimizing does not significantly change the types of molecules populating the hit lists.

• Best (or rigid or yes) -- evaluates only the best conformation at each vertex of the simplex minimizer and continues only evaluating this single conformation through all steps of minimization. This is significantly slower (order of magnitude) than not minimizing.

• Hierarchy -- evaluates the entire hierarchy for each vertex of the simplex minimizer and continues to re-evaluate the hierarchy at each step of minimization. This allows the conformation to change as the minimization proceeds. This option is the slowest.

Degeneracy checking: The following degeneracy checking keywords must be removed for Northwestern DOCK to function: degeneracy_wobble, degenerate_save_interval, check_degenerate_children. The current degeneracy checking routine, if turned on, only removes orientations generated from repeated spheres. Degeneracy checking currently does not significantly improve results or significantly reduce run-time. I recommend that degeneracy checking be turned off (i.e., check_degeneracy no).

New/Modified keywords:

bump_max [integer] -- This keyword describes the number of ligand bumps allowed per branch. The definition of this keyword was changed because determining the total number of bumps per molecule is time consuming if reliably implemented in a recombination-docking scheme.

focus_bump [integer] -- Maximum number of bumps (per branch) allowed in an orientation to allow focusing to occur. The implementation of focusing is a modified version of that found in DOCK2.5 (Shoichet, 1992)

focus_type [shape or energy] -- If focusing is turned on, additional spheres will be used to find similar orientations. For the first round of focusing, all orientations under focus_bump are further refined. For all subsequent rounds, only the best scoring orientation will be used for additional focusing. The best orientation is either the highest contact score or lowest energy score.

ligand_desolvation [residue, full, or none] -- This keyword is for the type of ligand desolvation to be used. Currently accepted values are: full (subtract the entire ligand desolvation value), residue (subtract the pre-calculated, hard-coded value for 25 amino acid variations), or none (no desolvation). Only full and residue are currently implemented. Only residues that are in contact with the receptor are desolvated when using residue desolvation. In the future I anticipate this keyword will also have the value ‘partial’, and possibly a counterpart keyword receptor_desolvation.

minimization_max [real value] -- The value for this parameter is the maximum, pre-minimized energy score that will be minimized. This is used primarily to speed calculations when focusing and minimizing are being used together – focus to get low energy structures, and then minimize.

recombine_fragments [yes or no] -- This option allows for recombining fragments from different molecules within a family to generate an overall best scoring molecule. This option is supported by dock (tested with proteins), but the current database is not organized by family so this option has no effect.

split_database_index -- This is a special value for the ligand_atom_file keyword. If this value is used, the filename is read as a plain-text file. Each line of split_database_index should have a complete path to the database files to be used.

timeout [real value] -- Maximum time in seconds allowed per molecule. The default is 86400 (one day in seconds). It is possible that some molecules may take an inordinate amount of time to dock. This is particularly important for molecules containing a large number of rotors in a single branch that contains downstream sub-branches. We find setting this value to 120 seconds is a good compromise. Molecules that timeout will be reported as timing out in the OUTDOCK file. (Added by John Irwin, modified by DL)

truncate_output [real value] This parameter is read only in single mode docking. The value for this parameter is the maximum RMS from the starting point to the docked orientation that will have coordinates written out. RMS values greater than truncate_output will have only remark lines written out