[Dock-fans] Amberscore after MD
sbrozell at scripps.edu
Sat Aug 16 15:27:46 PDT 2008
On Fri, 15 Aug 2008, Francesco Pietra wrote:
> For a protein and a ligand I carried out rigid docking, flex docking
> and amberscore with DOCK6. Then, the complex protein-ligand was
> embedded in a lipidic membrane for MD simulation with AMBER 10.
> At this point, I would like to carry out amberscore again. Is it wise
> to do that for the lowest_energy_structure captured from MD
> trajectories with ptraj? If so, working out manually the pdb file,
> extracting the parts related to the protein and the ligand for
> "Prepare" with Chimera.
> The alternative (which I tend to disfavor) is the averaged structure
> from the MD simulation.
> Or is anything better?
The simple approach is to use the lowest_energy_structure, because the
averaged structure might not correspond to reality, and to eyeball
any structure as a basic check on reality.
However, more complicated approaches may be better.
In general, one can usually find at least one metric for judging the
quality or appropriateness of the starting structure.
For example, sometimes a particular receptor functional group
interacts with a ligand. One could form a histogram of the group-to-
ligand distance over the MD trajectory. The shape of the histogram
would determine the set of starting structures: one big peak - pick
a structure from the peak bin; two separated but similar peaks -
now two starting structures would be necessary; etc. Other metrics
might be ligand-centric or receptor-centric.
Note that any MD geometry might not be very good, but that might not
matter. Amber score is probably more sensitive to the geometry
than the other DOCK scores because MM interaction energies are
more forgiving of clashes, etc.
On the other hand, Amber score provides many levels of relaxation
of poor geometries, and using a real Amber MD snaphot would be
already relaxed. So Amber scoring before and after DOCK docking
might be useful.
On Sat, 16 Aug 2008, Francesco Pietra wrote:
> I forgot to ask about the number of steps for amberscore that may
> hopefully lead to meaningful outputs. After flex docking, I used in
> limit_max_ligands no
> skip_molecule no
> read_mol_solvation no
> calculate_rmsd no
> orient_ligand no
> flexible_ligand no
> bump_filter no
> score_molecules yes
> contact_score_primary no
> contact_score_secondary no
> grid_score_primary no
> grid_score_secondary no
> dock3.5_score_primary no
> dock3.5_score_secondary no
> continuous_score_primary no
> continuous_score_secondary no
> gbsa_zou_score_primary no
> gbsa_zou_score_secondary no
> gbsa_hawkins_score_primary no
> gbsa_hawkins_score_secondary no
> amber_score_primary yes
> amber_score_secondary no
> amber_score_receptor_file_prefix protein (with
> all hydrogen atoms in place)
> amber_score_movable_region everything
> amber_score_before_md_minimization_cycles 100
> amber_score_md_steps 3000
> amber_score_after_md_minimization_cycles 100
> amber_score_gb_model 5
> amber_score_nonbonded_cutoff 18.0
> amber_score_temperature 300.0
> ligand_outfile_prefix output
> write_orientations no
> num_scored_conformers_written 1
> rank_ligands no
> In order to re-run amberscore after MD in the membrane, how many
> (minimum) steps may be suggested?
That is an open question. My feeling is that your receptor preparation
is extremely good compared to the typical for docking. So the above
protocol may be ok. As usual, to guage the protocol will require more
calculations: systematically increase the cycles, steps, and movable sizes
and look for convergence.
More information about the Dock-fans