[gmx-users] RMSD reproducibility

[Justin Lemkul]

On 2/15/16 12:21 PM, Biplab Ghosh wrote:
> Dear Gromacs Users,
>
> I have the following question:
>
> Suppose I repeat a simulation, keeping everything exactly the same.
> What will happen to RMSD? Will it be exactly same with the previous run?

No.

> If not, then how much variation is expected?
>

Impossible to tell.  RMSD is also a useless metric in this regard.

> To check this, I ran two independent simulations:
> 1) upto t=100ns using 216 processors and
> 2) upto t=200ns using 432 processors.
> Everything else were kept unchanged!
>
> The calculated RMSD for upto t=100ns was quite different.
> Am I doing something  wrong? Please explain.
>

Welcome to the chaotic world of MD.  Deviations occur very rapidly due to a 
number of factors.  This is why it is commonplace to run multiple simulations to 
confirm results.  One trajectory alone is usually insufficient evidence that the 
findings are conclusive.

http://www.gromacs.org/Documentation/Terminology/Reproducibility

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================