Re: FEP error "atoms moving too fast"

From: jz7_at_duke.edu
Date: Mon Jun 05 2006 - 12:25:08 CDT

Thanks so much for the reply. I have been using 1fs as the timestep in my
simulation. And the system is already equilibrated for about 0.5ns before
doing the FEP calculation. So I guess my problem is b) and the related psf
file problem. Here is how I build my dual-topology psf file and fep_input
file.

During the FEP calculation, I am changing one cofactor (A) to the other
(B), as well as changing two Cl ions to two water molecules. The initial
(A) and final (B) cofactors share most common part, and the only
difference is that there is an extra CH3 group in the initial cofactor A.
Unlike the FEP tutorial, I change the whole cofactor A to B during FEP
simulation, instead of just changing the CH3 group.

In the fepinput coordinate file, I have both coordinates for cofactors A&B
with different B column value. And the psf file is build for the whole
system, where cofactor A and B both exist.

Is there anything wrong with the way I set up my system? Thanks so much!

Jeny

On Sun, 4 Jun 2006, Chris Chipot wrote:

> Jeny,
>
> there are at least three possible diagnostics for the problem you are
> encountering:
>
> a) your system has been improperly equilibrated;
>
> b) the list of excluded atoms is incorrect, so that the topologies of
> the initial and the final states do see each other when they never
> should. I would also check the internals common to the two states;
>
> c) the time step is too large;
>
> and any combination of the above.
>
> There is a clear misunderstanding on the choice of delta lambda here.
> An improperly chosen increment of the coupling parameter is unlikely
> to cause a simulation to crash, unless there is an obvious flaw in the
> definition of your system. Too large a delta lambda will simply result
> in an absence of low-energy configurations of the target state that
> are also configurations of the reference state - hence an absence of
> convergence of your free energy calculation, but certainly not a crash
> thereof.
>
>
> Chris chipot
>
>
>
> jz7_at_duke.edu wrote:
>
> > Dear all,
> >
> > I was running FEP calculation for a system about 50000 atoms
> > (inclusing protein, DNA, cofactor, ions, and water). However, the
> > calculation didn't even go through the first lambda change step (I set
> > the change for each step to be 0.01). The error message I got is
> > something about "atoms moving too fast" and after some iterations the
> > energy change goes to inffinity. Is it because the change of
> > lambda for each step is still too large for my system or something is
> > wrong with the setup of the calculation? How should I check it?
> >
> > Thanks a lot!
> >
> > Jeny
>
> _______________________________________________________________________
>
> Chris Chipot, Ph.D.
> Equipe de dynamique des assemblages membranaires
> Unité mixte de recherche CNRS/UHP No 7565
> Université Henri Poincaré - Nancy 1 Phone: (33) 3-83-68-40-97
> B.P. 239 Fax: (33) 3-83-68-43-87
> 54506 Vand?uvre-lès-Nancy Cedex
>
> E-mail: Christophe.Chipot_at_edam.uhp-nancy.fr
> http://www.edam.uhp-nancy.fr
>
> Science without management is worse than management without science
>
> N. G. van Kampen
> _______________________________________________________________________
>

-------------------------------------------------------------
JIERU ZHENG Phone: (919) 660-1634
Department of Chemistry Fax: (919) 684-4212
Duke University Email: jz7_at_duke.edu
Box 90348
Durham, NC 27708

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 05:19:31 CST