[Dock-fans] amber_score

Scott Brozell sbrozell at scripps.edu
Thu Nov 1 22:39:49 PDT 2007


I answered some of these in the latest 'Combining DOCK with AMBER'
but reproduce them for simplicity.

On Wed, 31 Oct 2007, Francesco Pietra wrote:

> I open a new thread (amber_score) because previous thread (Combining Dock with
> Amber) was becoming too complex, in the hope to get some general guidelines.
> To summarize
> (1) With a pore protein model and a flexible 150-atoms ligand (working with
> Dock6.1 and Chimera 17 Oct build) I could carry out both rigid_docking and
> flex_docking along the default lines of the tutorials. The ligand is seen
> inside the pore, roughly the same view in both cases.


> (2) As prepared with Chimera, the protein did not allow to get (via dock-prep)
> the prmtop and inpcrd files needed for amber_score. Therefore, I prepared the
> protein with Leap in Amber9, getting prmtop and inpcrd, which allowed a correct
> graphical representation in both Chimera and VMD.

This is not surprising since Amber's LEaP (which is the vehicle for
amberization under the hood of DOCK's prepare_amber.pl)
is the sole mechanism for Amber input preparation.

Here is the usual sequence:
Do docking (along the lines of tutorials
Structure Preparation,
Sphere Generation and Selection,
Grid Generation,
Docking );
Then rescore with Amber_score (along the lines of tutorial
Structure Preparation & AMBER Score ).
Note that Amber_score has receptor input files distinct
from DOCKing receptor input files.

Typically, one has an original pdb which is the starting point for
dock prep, and the resulting mol2 file (from step 2f in the
Structure Preparation tutorial) also can be saved as a pdb;
it is this later pdb that is usually the starting point for
prepare_amber.  However, one can imagine several ways to
obtain the actual receptor input files; for example, one might
take the pdb file from step 2g and use that for prepare_amber.
This would be a simple consistency test for protonation
between the DOCKing receptor and the Amber_score receptor.

It seems to me that you are performing various consistency checks
between DOCKing inputs and Amber_score inputs.
That is a sound approach, but human inspection will play a
dominant role since the DOCK and the Amber packages are
not input compatible (despite the appearance through the
magic of amberization).

[Note to Amber/DOCK superexperts: there are several caveats
that I have ignored for pedagogical purposes.]

> (3) The pdb file for the protein created in Amber9 from prmtop and inpcrd with
> ambpdb proved unsuitable for Chimera dock-prep (standard amino acids were
> mis-viewed as non-standard residues and assigned to Antechamber, which wrongly
> calculated fractional charge for the residues). I changed strategy, loading the
> protein to Chimera from the Leap-created prmtop and inpcrd files. That also
> failed (dock-prep), but for a different reason, Value Error: underlying C++
> Molecule object is missing (I have documented that with an error.log on
> previous thread), which I do not understand.

Please post this small episode to Chimera-users.
This may be an issue of operator error, but maybe not and
anyways it may help them improve robustness.

> Now the files from Leap (2 above) allowed amber_scoring. I carried out both a
> rapid (few minutes) "ligand" option and a longer (4 hours) "everything" option.
> In both cases, from the complex "final_pose.amber.pdb", the ligand is immersed
> into the protein, though not along the pore walls, unlike with the rigid and
> flex docking mentioned in (1). The bad point is that the ligand was broken into
> two major fragments, and minor CH2 fragments. At a rough inspection, the
> protein does not appear to have suffered any major damage, in both options. I
> must add that breakage of the ligand is surprising because it is a thermally
> very stable molecule.

Something is amiss - covalent bonds should not break.
Did you run prepare_amber.pl ?

> I am much confused about the line to follow for amber_score. As carried out
> above, the receptor has no spheres, while the ligand mol2 file was taken from
> the tutorial-like procedures. Does the receptor pass through the stages of
> spheres/grid also for amber_score?

Amber_score only uses a sphere representation of the receptor
for the distance-movable-region.  Otherwise Amber_score does not use
any spheres, and it never uses any grids.
Perhaps my comments to (2) are useful here ?


More information about the Dock-fans mailing list