Re: [AMBER] SCEE/SCNB zeros

From: Ross Walker <ross.rosswalker.co.uk>
Date: Wed, 22 Nov 2017 08:29:00 -0500

Hi Stefan,

If memory serves it was added as a 'backup' check for the mechanism by which 1-4 interactions are handled. Sander and PMEMD handle 1-4's by traversing the dihedral list but obviously that means only the first dihedral in the series in each case should be used. The sign of the third index is used to handle this. Originally SCEE and SCNB factors were never in the prmtop - a fixed value from the &cntrl namelist was used for the entire calculation. When I added support for variable SCEE and SCNB values to properly support mixing force fields, for example Glycam with FF14SB, as an extra check during debugging things were setup such that additional SCEE and SCNB scaling factors beyond the first occurrence for a given dihedral would be set to zero. If the code is working correctly these values should never be referenced. The thought process being that if they were getting referenced for some reason then, since the code divides by SCEE and SCNB, this would trigger a divide by zero and the resulting simulation would
yield NANs rather than what look to be potentially valid, but in reality incorrect, results.

Hope that helps.

All the best
Ross

> On Nov 22, 2017, at 05:11, Stefan Doerr <stefdoerr.gmail.com> wrote:
>
> Hi, I wanted to ask if there is any documentation on why tleap sets some
> scee and scnb factors to 0. I could not find anything in the AMBER manual
> about it. Jason Swails told me that it was used as a convention for
> omitting 1-4 interactions but afaik the sign of the third index of a
> dihedral already is used to indicate that:
>
> http://ambermd.org/formats.html
> "If the third atom is negative, this implies that the end group interations
> are to be ignored."
>
> Best,
> Stefan
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 22 2017 - 05:30:03 PST
Custom Search