[gmx-users] Re: Water molecules cannot be settled: V-rescale is the cause

Ladasky blind.watchmaker at yahoo.com
Thu Sep 20 09:08:18 CEST 2012

Mark Abraham wrote
> This all sounds much like an issue with the topology or starting 
> configuration. 

And it is for that reason that, during my debugging process, I switched back
from the chimeric protein structures that I was building myself to a
standard PDB structure for which I had previously completed a 5-ns stable
run.  I wanted to eliminate any problems that the starting configuration
might have caused.

Mark Abraham wrote
> I have seen systems where 0.5 fs without bond constraints 
> was necessary to relax issues with the initial conformation, and using 
> position restraints on (say) all heavy atoms would have made this 
> impossible.

I'll keep in mind that even 1.0 fs is sometimes not short enough, when
working far from equilibrium.  That being said, the fact that I saw almost
the same water molecule misbehave at almost the same simulated time point,
whether I used a 2.0 fs or a 1.0 fs time step, gave me the strong hint that
my simulation was both reproducible -- and wrong.  And thus, that my time
step was probably not the critical factor.

Mark Abraham wrote
> Your protocol from July was also asking for trouble by 
> generating velocities and moving straight into an NPT ensemble with 
> position restraints. Starting with an NVT ensemble can be better idea, 
> particularly if the volume is not quite right.

If you read my reply to Justin, you will know that I'm not deliberately
trying to skip any steps.  I have made minimal modifications to the MDP
files that were part of the GROMACS 3.3 lysozyme tutorial shell script. 
That protocol included a position-restrained solvent equilibration step, and
I used it as-is.

I'm not married to the need to go straight to NPT, or even to use position
restraints during solvent equilibration.  If the starting configuration of
my protein of interest shifts while I'm making adjustments to the solvent, I
don't mind.  I'm trying to watch it fold, and the end point is far more
important to me than the starting point.

Did Berendsen himself ever post to this mailing list?  I remember coming
across this signature line on some GROMACS-related post: "Why do YOU use
constraints?"  In my case, the answer is, "because that's what my tutorial

> Now, WHY would I have switched from a Berendsen temperature-coupling
> algorithm to the V-rescale algorithm?  Because of this cautionary note
> that
> I started receiving in GROMACS 4.5 when I started the position-restrained
> mdrun:
> NOTE 1 [file pr.mdp]:
>    The Berendsen thermostat does not generate the correct kinetic energy
>    distribution. You might want to consider using the V-rescale
> thermostat.

Mark Abraham wrote
> Your observations on one system are not enough to reach this conclusion. 
> v-rescale is normally quite appropriate for equilibration. The above 
> hint is one that only matters when you wish to perform proper 
> equilibrium sampling. 

I understand that hint, in hindsight.  It led me down a bit of a primrose

Mark Abraham wrote
> Multiple phases of equilibration are normal, 
> particularly in tricky cases, to gradually approach the conditions under 
> which you wish to sample, via those that help deal with trouble with 
> starting conditions.
> Merely switching to the Berendsen temperature algorithm can be lucky 
> enough to lead to stable equilibration. Pathological starting conditions 
> are sometimes quite sensitive. 

Well, I've been consistently lucky then, up to the point that I switched to
V-rescale.  But I would rather be right than lucky.  I don't care if my
simulation takes an extra day to run, as long as the darn thing doesn't

Thanks again for your help!

View this message in context: http://gromacs.5086.n6.nabble.com/Re-Water-molecules-cannot-be-settled-why-tp4999302p5001134.html
Sent from the GROMACS Users Forum mailing list archive at Nabble.com.

More information about the gromacs.org_gmx-users mailing list