[Dock-fans] Amber score

Scott Brozell sbrozell at scripps.edu
Fri Jul 6 15:49:24 PDT 2007


On Fri, 6 Jul 2007, Rima Chaudhuri wrote:

> I am running  2 different sets of ligand conformations ( but of the same 30
> molecules) for rescoring with amber_score.
> A. Using the normal grid score to rank some 30 compound and then use these
> conformations for rescoring with amberscore.I used the movable region as
> 'ligand' in case1 and 'distance' in case2. Case 1 runs fine. Case 2 however,
> stalls at the following point
> Molecule Library Output Parameters
> ------------------------------------------------------------------------------------------
> ligand_outfile_prefix                                        out_dist
> write_orientations                                           no
> num_scored_conformers_written                        1
> rank_ligands                                                    yes
> max_ranked_ligands                                         30
> these are the last few lines of the dock.out file, it is still running and
> has been for the past 2.5 hrs w/o writing anything else. I want to know if
> this is normal? I used the pre-md min steps as 100, md-steps of 5000 and
> post md steps of 100. The movable cutoff was specified as 8.0 angstroms,
> temperature 300K...any suggestions?

The only extra input parameter for distance movable vs ligand movable is
receptor_site_file.  Did you specify it ?
For details see

distance may be a lot slower than ligand.
As usual, the advice is to run in verbose mode when doing something
for the first time: dock -v bla bla
A lot of extra info will be emitted; it will include clues on
the computational intensivity.

> B. Using the same ligand set as in 1, I first ran them in GOLD and used
> these conformations to re-score in dock_amber_score. Almost all the amber
> scores are positive, whereas in the case of using grid score and then amber
> re-scoring the amber scores were all negative values and suggested good
> binding as expected. I understand that using the gold conformations and then
> dock might cause this problem due to differences in energy terms and
> parameters between Gold and dock_amber_score, but to keep things consistent,
> is there a workaround this issue? meaning , if I wanted to make correlation
> plots between the amberscore and experimental data, right now it is quite
> fruitless!
> If I stick to the conformations from grid_score ranking, the do not look so
> good and the experimental correlation is only 18%. I would rather want to
> have the gold conformations as the starting geometry! any suggestions that
> can help me get non-positive binding energy using amber_score?

These are interesting issues perhaps on the bleeding edge of docking;
I am not aware of any study of them.
How do the other DOCK scores fair on the GOLD conformations ?
I cannot think of any cheap black box workarounds.
Increasing the movable region size may improve the amber scores.
One would expect that more receptor preparation would improve the amber
scores.  As mentioned before the error bars on binding energies
from docking are large:

Good luck and keep us posted,

More information about the Dock-fans mailing list