Re: [AMBER] strategies to speed up MM/GBSA screening

From: Jason Swails <>
Date: Wed, 13 Jul 2016 08:14:24 -0400

On Wed, Jul 13, 2016 at 7:08 AM, Thomas Evangelidis <>

> ​Greetings,
> I want to screen through MM/GBSA a few thousand small molecules that are
> superimposed on the crystal ligand in the binding pocket of a protein. The
> idea is not to calculate absolute binding free energies, but to improve the
> ranking by discarding those that don't fit well into the cavity​. For this
> purpose I impose restraints to the protein atoms and leave flexible only
> the side-chains that are close to the crystal ligand. However, this incurs
> a computational slow down. Since I am not really interested in the dynamics
> of the protein atoms that are not in the binding cavity, is there any way
> to prevent sander from calculating positions and velocities for these
> atoms? I read in the manual that "ibelly" is not compatible with implicit
> solvent and moreover it does not provide any significant speed gains. The
> other option would be to reduce the protein to the binding cavity and
> simulate only that. However, I am not sure if this may lead to significant
> loss of accuracy in some cases.

​The slow part of the calculation is calculating the forces. Compared to
that, the cost of integrating positions and velocities is nearly
negligible. So the question here is "is there any way to efficiently
calculate only the forces on the atoms you wish to propagate"?

Unfortunately the answer to that is "no", especially with GB. The
long-range effect of the nonbonded interactions means you need to compute
the interactions between all atoms you wish to study (near the active site)
with all other atoms in the system. And you also can't restrict the
nonbonded calculation between only the (very small number of) pairs in
which one atom is one of the "atoms of interest", because the GB potential
is not pairwise decomposable.

What this means is that the presence (and location) of atom A can influence
the interaction between atoms B and C (it does this through A's effect on
the Born radii of both atom B and C). Which means that regardless of which
atoms you want to integrate, you effectively need to do the whole force

Moral of the story: If you want to speed up your calculation​, run on GPUs.


Jason M. Swails
AMBER mailing list
Received on Wed Jul 13 2016 - 05:30:03 PDT
Custom Search