Re: restart with constraints causes RMSD "jump"

From: Jim Phillips (jim_at_ks.uiuc.edu)
Date: Wed May 25 2011 - 11:31:29 CDT

The coordinates used for the constraints are being overwritten by the
bincoordinates file. You can see this because the BOUNDARY energy where
constraints are reported is zero at restart. This only happens if the
coordinates and consref filenames are the same, so you can work around the
issue by giving the files different names (or make one a symlink to the
other). I just checked in a fix that moves the bincoordinates read
to after constraints reference coordinates have been copied.

Thanks for bringing this up.

-Jim

On Wed, 25 May 2011, P.-L. Chau wrote:

> I would like to re-ask the question which was posted on the NAMD mailing
> list last year:
>
> http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l/13197.html
>
> This question still has not received a reply. I personally emailed the
> scientist who posted it and he said that he still had not found an answer.
>
> I note that in a previous publication (Sotomayor and Schulten (2004)
> Biophysical Journal, 87, 3050-3065) the authors use constraints in an NPT
> ensemble. Their constraints act only on the main-chain atoms. I am running
> a similar system where the constraints act on all the protein atoms, but
> not on the surrounding hydrated membrane.
>
> I found that on stopping the run and restarting, there was a "jump" in the
> r.m.s. deviation, as was reported previously on the NAMD mailing list. It
> is well known that r.m.s. deviations fluctuate, but there was an abrupt
> increase in the value around which it fluctuated.
>
> If I cancelled the constraints, then this "jump" disappeared. If I
> shortened each run to only 200 time-steps, this "jump" also disappeared.
> Lastly, if I used a much smaller protein, then even if I put constraints
> on the whole protein and run long simulations, this "jump" was not
> observed. It appeared that this "jump" came from doing long simulations on
> a large system (>330000 atoms) with constraints.
>
> I would like to ask if any of you have met with this problem, and if so,
> whether there is any fix for it. I am sorry I have not included any of the
> files, graphs and analysis results because they take up too much space. I
> shall be happy to email those to anybody who is interested, or put them on
> a website so they can be downloaded.
>
> Thank you very much indeed.
>
> P-L Chau
>
> email: pc104_at_pasteur.fr
> Bioinformatique Structurale
> CNRS URA 2185
> Institut Pasteur
> 75724 Paris
> France
>

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:20 CST