lengthening time-step with constraints causes RMSD "jump"

From: P.-L. Chau (pc104_at_pasteur.fr)
Date: Thu Jun 02 2011 - 06:53:05 CDT

Could I ask for some help with more RMSD "jump" problems, please?

I have been trying to equilibrate a protein. I used Ewald summation for
the electrostatic interactions. The van der Waals interaction was handled
without a switching function, but with long-range correction. I started
off with strong harmonic constraints of the protein atoms, at 10K and with
time-steps of 0.1fs each. I ran this system for 20ps, and the RMSD during
this period fluctuated around 0.07AA. At the end of this run, I re-started
the simulation and lengthened the time-step to 0.2fs, and I found that the
RMSD jumped to over 2AA and stayed there. I then re-started from the same
point, but used a time-step of 0.15fs, and I found that the RMSD jump
disappeared.

Thinking that the van der Waals interaction could be the culprit, I re-did
the simulations with a switching function (switchdist 10), but I saw the
same effect. 0.2fs always caused an RMSD jump, and 0.15fs never did.

It is well known that r.m.s. deviations fluctuate, but this was an abrupt
increase in the value around which it fluctuated. I also checked that my
boundary energy was finite, at 16991.6855.

I could not think of any reason why lengthening the time-step would lead
to such a dramatic change in protein structure. Could I ask if any other
user has met with a similar problem? How do you deal with it?

The graphs, analyses and NAMD files involved are quite large, since I am
simulating a protein/membrane complex with >330000 atoms. So I am not
including them here. If you would like to see them, please email me, and I
shall put these files on a website and/or email them to you.

Thank you very much!

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 : Wed Feb 29 2012 - 05:24:01 CST