[gmx-developers] soft-core atom selection
Shirts, Michael R. (mrs5pt)
mrs5pt at eservices.virginia.edu
Thu Nov 20 15:30:18 CET 2014
Can you upload the files that are causing problems to redmine? Then I can go in a look at the interactions directly. This all need to be documented better.
Department of Chemical Engineering
University of Virginia
michael.shirts at virginia.edu
From: "dmobley at gmail.com<mailto:dmobley at gmail.com>" <dmobley at gmail.com<mailto:dmobley at gmail.com>>
Date: Thursday, November 20, 2014 at 6:37 AM
To: "gmx-developers at gromacs.org<mailto:gmx-developers at gromacs.org>" <gmx-developers at gromacs.org<mailto:gmx-developers at gromacs.org>>
Cc: "michael.shirts at virginia.edu<mailto:michael.shirts at virginia.edu>" <michael.shirts at virginia.edu<mailto:michael.shirts at virginia.edu>>, Berk Hess <hess at kth.se<mailto:hess at kth.se>>
Subject: Re: [gmx-developers] soft-core atom selection
OK, to follow up on this - I think this is incorrect at least in part, and explains a crash I’ve had on my list of things to track down. Specifically, if I simulate multiple carboxylic acids in solution and attempt a hydration free energy calculation (as a specific example, let’s take a case I’ve done where I have neutral benzoic acid at a finite concentration in water, and I am attempting to compute the free energy of turning one of the benzoic acid solutes into dummy atoms) I simply CANNOT get the calculation to run stably, even with reduced time steps, etc. In every respect it behaves like the type of crash I would expect if I had charged sites overlapping one another, but I had been unable to figure out how this could possibly happen. I WAS, however, able to determine that if I added nonzero LJ parameters to the hydroxyl hydrogen on the carboxylic acid (which is a GAFF hydrogen and hence had zero LJ parameters) the crashes go away.
I believe this is explained by your statement about “for atoms without LJ, all perturbed interactions are soft-cored”. Specifically, I am turning off the solute (carboxylic acid) and so its perturbed interactions with the REMAINING, non-perturbed solutes are soft-cored, meaning that charge sites can overlap. At least, that’s my reading of what you’re saying and it would perfectly explain the crashes I’m seeing.
I’ve also seen essentially identical crashes I was unable to understand for binding free energy calculations in a carboxylic-acid-rich host-guest system, presumably again because the REMAINING hydroxyl hydrogens on the host end up being able to overlap with charge sites on the guest which is being turned into dummies. Again I can “fix” it by adding nonzero LJ parameters to the hydrogens on the host, though this is incorrect since it is not “allowed” by the force field.
Unless I’m missing something, this is a substantial mistake, and clearly a better solution is needed.
Thanks so much,
dmobley at gmail.com<mailto:dmobley at gmail.com>
On Wed, Nov 19, 2014 at 3:36 AM, Berk Hess <hess at kth.se<mailto:hess at kth.se>> wrote:
Soft-core is not decided on an atom basis. A pair interaction is soft-cored when at least one of the two atoms is perturbed and at least one of the atoms has c12=0 at lambda=0 or lambda=1. So for atoms with LJ, interactions are only soft-cored when strictly necessary. For atoms without LJ, e.g. many hydrogens in many force fields, all perturbed interactions are soft-cored. For such atoms soft-coring is necessary when connected to a heavy atom that is decoupled. So the only cases where some interactions are soft-cored that are not strictly necessary is in a system where some atoms are decoupled and others are modified, then interactions with hydrogens connected to modified atoms would not need to be soft-cored.
On 11/18/2014 11:45 PM, David Mobley wrote:
I wanted to ask concerning soft-core potentials in GROMACS: How is it determined which atoms are soft core? Specifically, some other code bases allow the option to select specific atoms which are going to be treated with LJ soft core potentials, while other atoms are not treated in this way and maintain their normal potential or a linear mix of their A and B state potentials. In GROMACS, there is no option to select this that I am aware of, so I'm assuming that soft core is applied to all atoms, or to all atoms which have different A and B states or different A and B state LJ parameters?
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the gromacs.org_gmx-developers