DOCK 6.4

For version 6.4, recent improvements to the sampling implementation and
corrections to prevent steric clashes that have been mentioned on the
Dock-fans list [1] are incorporated here.

NEW IN DOCK 6.4

###Internal Energy###

   During growth and minimization, an internal energy scoring function can be
used. The goal of the internal energy function is to reduce the occurrence of
internal clashes during the torsional optimization. This function computes the
repulsive Lennard-Jones term between all ligand atom pairs, excluding all 1-2,
1-3, and 1-4 pairs. (Currently, attractive Lennard-Jones and Coulombic terms
are neglected; the aim is to eliminate internal clashes not to optimize the
internal geometry. In addition, since there is no dihedral term in the force
field, if the attractive terms are included, the molecule might appear less
physical.)  The internal energy can be cut on or off; if cut on, it is
always used and is reported in the final energy.  If it is cut off, it is never
used.  The pruning during growth is done by considering both the internal and
interaction energies.  We recommend its use.

   Note that DOCK 6.3 also had an internal energy function however it was only
used in final minimization and not during growth.  

   The modifications have greatly improved the behavior of the Anchor and Grow
algorithm.

###Growth Tree and Statistics###

   Dock uses Breadth First Search to sample the conformational space of the
ligand. The tree is pruned at every stage of growth to remove unsuitable
conformations. In order to be as space efficient as possible, DOCK only saves
one level of growth at a time unless "write_growth_tree" is turned on. In
order to construct the growth tree it was necessary to do the following: (1)
Retain all levels of growth (before and after minimization) in memory. (2)
Link every conformer to its parent conformer during growth. (3) While writing
out the tree, the traversal starts from a fully grown ligand (leaf), moving up
the branch (parent conformer) until the ligand anchor (root) is reached.
Finally, the growth tree branch is printed as a multi-mol2 file starting from
the anchor to the fully grown ligand, including minimizations. This newly
implemented feature allows visualization of all stages of growth.
Note that the growth trees can easily be visualized using the Viewdock
module in the UCSF chimera program. Extra information regarding conformer
number, anchor number, parent conformer, etc. can also be accessed
directly using this tool.

   The verbose growth statistics feature is useful for debugging incomplete
growths and other possible issues with the growth routines. This feature
is also useful to show progress when docking larger peptide-like ligands
(twenty or more rotatable bonds) which can take several hours.
In the example below, cumulative timings in seconds (e.g., t=16.75s)
are shown at the end of each line to allow quick profiling of
the slowest steps during docking. A separate section is printed for each
anchor sampled when using multiple anchors. For anchor #1, the orienting
routine produces 998 orients, and 41 are retained after clustering and
minimization. The ligand has 7 rotatable bonds. The second line shows the
assignment of layers and segments. For details on the terminology, please
consult the DOCK 4 paper. Subsequently, two lines of information are printed
for each torsion sampled.

    VERBOSE GROWTH STATS : ANCHOR #1
    41/998 anchor orients retained (max 1000) t=16.75s
    Lyr 1-1 Segs|Lyr 2-1 Segs|Lyr 3-2 Segs|Lyr 4-2 Segs|Lyr 5-1 Segs|
    Lyr:1 Seg:0 Bond:8 : Sampling 6 dihedrals C6(C.ar) C4(C.ar) C3(C.3) C1(C.3)
    Lyr:1 Seg:0 21/246 retained, Pruning: 2-outside grid 5-score 218-clustered (62 in growth tree) t=20.92s
    Lyr:2 Seg:0 Bond:5 : Sampling 3 dihedrals C4(C.ar) C3(C.3) C1(C.3) N1(N.3)
    Lyr:2 Seg:0 57/63 retained, Pruning: 1-score 5-clustered (119 in growth tree) t=22.36s
    Lyr:3 Seg:0 Bond:1 : Sampling 3 dihedrals C3(C.3) C1(C.3) N1(N.3) S1(S.o2)
    Lyr:3 Seg:0 105/171 retained, Pruning: 1-outside grid 9-score 56-clustered (224 in growth tree) t=27s
    Lyr:3 Seg:1 Bond:3 : Sampling 6 dihedrals N4(N.am) C2(C.2) C1(C.3) C3(C.3)
    Lyr:3 Seg:1 67/630 retained, Pruning: 16-score 547-clustered (291 in growth tree) t=48.8s
    Lyr:4 Seg:0 Bond:43 : Sampling 3 dihedrals C16(C.ar) S1(S.o2) N1(N.3) C1(C.3)
    Lyr:4 Seg:0 100/201 retained, Pruning: 101-clustered (391 in growth tree) t=56.67s
    Lyr:4 Seg:1 Bond:26 : Sampling 2 dihedrals C11(C.3) N4(N.am) C2(C.2) C1(C.3)
    Lyr:4 Seg:1 154/200 retained, Pruning: 13-score 33-clustered (545 in growth tree) t=73.21s
    Lyr:5 Seg:0 Bond:46 : Sampling 6 dihedrals C17(C.ar) C16(C.ar) S1(S.o2) N1(N.3)
    Lyr:5 Seg:0 99/924 retained, Pruning: 30-score 795-clustered (644 in growth tree) t=179.43s

    Lyr:1 Seg:0 indicates that this is Layer #1 and Segment #0. Layer and
segment numbers start from zero and correspond to the array indices used
internally. Bond:8 refers to the bond number in the mol2 file. "Sampling 6
dihedrals C6(C.ar)  C4(C.ar)  C3(C.3)  C1(C.3)" specifies the exact torsion
being sampled. Six dihedral positions are being sampled in this case, as
determined by the drive_id in flex_drive.tbl. 21/246 retained means 21
conformers were retained from the 246 conformers generated during growth (41
conformers x 6 dihedral positions = 246 new conformers). The Pruning section
demonstrates how these 246-21=225 conformers were pruned: 2 conformers
were outside the energy grid, 5 conformers exceeded the score cut-off (see
pruning_conformer_score_cutoff) and 218 conformers were clustered. Typically
clustering removes the greatest number of conformers during each torsion grown
as controlled by the pruning_clustering_cutoff parameter. The reader is
encouraged to verify that the number of conformers retained can be calculated
as above at each stage of growth. If the growth tree is turned on, the total
number of conformers stored in the growth tree are also reported.

###Database Filter###

   The Database Filter is designed for on-the-fly filtering of small molecules
from the database during docking. Filtering small molecules by heavy atoms,
rotatable bonds, molecular weight and formal charge is currently supported.
This routine is designed to be modular so that other descriptors can be easily
added. The default values are deliberately set to allow most small molecules
to pass through. One use of this routine would be to partition a database into
subsets such as "0-7 rotbonds" or "300-500 molwt" or "neutral charge". Another
use would be to exclude ligands that are too small (<200 amu) or too large
(>500 amu) for a particular target. This routine can also be used to filter a
database without performing any docking.

###Pre-minimization###

   Pre-minimization, which explores only the torsional (N) degrees of freedom,
occurs prior to standard minimization.  Note that if 500 iterations of
standard minimization and 20 iterations of torsion pre-minimization are
specified, at most 520 steps of minimization will be performed. The motivation
for using the torsion pre-minimizer is to optimize the torsions before
translating. This parameter should only be used in cases where the growth tree
reveals that the minimizer is translating a correctly oriented scaffold to
relieve clashes instead of adjusting the torsions first. Alternatively, the
restraint minimization routine could be used (see below). 

###Restrained Minimization###

   We have implemented an "RMSD tether": Erestraint = k * RMSD ^ 2.
If simplex_restraint_min is yes then every time the minimizer is called the
conformer is restrained to the initial conformation passed to the minimizer.
Only the active heavy atoms are used in the calculation.  If this function is
used during growth, the entered conformer grown so far is restrained including
the most recently grown torsion.  This ensures considerably less movement
during minimization.  

###Miscellaneous###

   The test suite now contains reference outputs from a 64 bit platform
(x86_64).  Several configuration files have been added; thanks to
Carlos P Sosa et al. for work on the ones for the Blue Genes. 

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

DEPRECATED FEATURES

1.  Elimination of the final minimization functionality from the code.  We
believe that it is no longer needed with the better behavior of Anchor and
Grow. 

2.  The Internal Energy function was moved from flexible growth. The Internal
Energy function is able to be used for all docking and minimization
calculations. 

3.  For a variety of reasons still under active development,
amber_score cannot effectively be used as a secondary_score.
This function was temporarily deprecated in 6.1, and using input parameter
amber_score_secondary causes program termination.
The recommended protocol is to perform two DOCK runs with the second run
specifying amber_score as the primary_score.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

BUG FIXES

Some bug reports have not yet been processed, but several important bugs have
been fixed since the last release.

For DOCK:

   For the uses of multiple anchors fragments (choose of different portions of
the molecule) a C++ vector was not properly cleared. This was resolved

   The orienting bug fix has been added and validated through visualization.  

   The resolution of the clash problem for highly flexible ligands (Rotatable
bonds > 7) has been accomplished by inclusion of an internal energy function
at every stage of growth.  The internal energy function can be cut on or off
and will always or never be used, respectively.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

INPUT PARAMETER CHANGES AND THEIR DEFAULT VALUES

Added:
   use_database_filter [no] (yes, no)
   dbfilter_max_heavy_atoms [999] (int)
   dbfilter_min_heavy_atoms [0] (int)
   dbfilter_max_rot_bonds [999] (int)
   dbfilter_min_rot_bonds [0] (int)
   dbfilter_max_molwt [9999.0] (float)
   dbfilter_min_molwt [0.0] (int)
   dbfilter_max_formal_charge [10.0] (float)
   dbfilter_min_formal_charge [-10.0] (float)

   write_growth_tree [no] (yes, no)

   simplex_tors_premin_iterations [0] (int)
   simplex_grow_tors_premin_iterations [0] (int)
   simplex_restraint_min [yes] (yes, no)
   simplex_coefficient_restraint [10.0] (float)

Removed:
   internal_energy_att_exp
   internal_energy_dielectric

   simplex_final_min [no] (yes, no)
   simplex_final_max_iterations [0] (int)
   simplex_final_min_rep_rad_scale [1.0] (float)
   simplex_final_min_add_internal [no] (yes, no)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

FILE FORMAT CHANGES

For DOCK:
   The label for Grid Score changes from  
##########    Grid Score:
##########           vdw:
##########            es:
to 
##########    Grid Score:
##########      Grid_vdw:
##########       Grid_es:

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

REFERENCES

1. http://mailman.docking.org/pipermail/dock-fans/2009-May/002058.html