Re: [AMBER] excluding atom sets from MMPB/GB-SA calculations

From: Jason Swails <jason.swails.gmail.com>
Date: Wed, 17 Dec 2014 14:06:31 -0500

On Wed, Dec 17, 2014 at 1:06 PM, Hallel Freedman <hfreedma.ualberta.ca>
wrote:
>
> Hi Ray,
> Thank you for your suggestion about changing the dielectric constant of the
> solute. I will definitely try this.
> Since I am interested in determining relative binding free energies, only,
> for a set of nucleotides, I believe it may be justified to ignore the
> contributions from the phosphate atoms since these belong to all ligands
> whose binding free energies are being compared. So I was thinking in terms
> of going into the AMBER MMPB-SA code and introducing a parameter, similar
> to what you have with residue decomposition, where the sum over atomic
> sites involved in determining binding free energy is only taken over
> selected atoms. Does this sound feasible to you?
>

​No. That's not how MM/PBSA works -- it parses the total energies computed
for the whole system from sander (or NAB) output. In any case, the
energies are not decomposable that way (deleting the phosphate group *also*
modifies the dielectric boundary which in turn modifies the interactions
between otherwise unrelated pairs of atoms; this is true for both GB and
PB).​ [1] Apologies if it's clear that I'm misunderstanding what you are
proposing here.

​Two more approximate things you could do is to zero-out the charges of the
phosphate group (using the "change" command in ParmEd) or just strip those
atoms out entirely. It may be a challenge to defend either approach,
though, unless you can somehow quantify the error you are adding to your
calculation by ignoring the phosphate electrostatics.

Good luck,
Jason
​​
[1] In GB, the "dielectric boundary" is wrapped into the effective GB
radii, which is computed as an integral over nearby atoms, so it is also a
non-local property that can't be decomposed.
​​
​--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Dec 17 2014 - 11:30:02 PST
Custom Search