Re: Random Velocities Question

From: McGuire, Kelly (mcg05004_at_byui.edu)
Date: Tue Jul 03 2018 - 22:46:05 CDT

Great information, thanks Josh and Axel for the advice.


Kelly L. McGuire

PhD Scholar

Department of Physiology and Developmental Biology

Brigham Young University

LSB 3050

Provo, UT 84602


________________________________
From: Axel Kohlmeyer <akohlmey_at_gmail.com>
Sent: Tuesday, July 3, 2018 9:37:39 PM
To: NAMD list; McGuire, Kelly
Subject: Re: namd-l: Random Velocities Question



On Tue, Jul 3, 2018 at 11:11 PM McGuire, Kelly <mcg05004_at_byui.edu<mailto:mcg05004_at_byui.edu>> wrote:

I've been told that NAMD has randomized velocities built in now. If you run a simulation, and then submit that simulation again, the velocities are automatically randomized, is that true? If so, how do I turn that off? I would like to run the same exact simulation again, but only change the partial charges on my ligand. If the velocities are randomized in the same simulation, then I won't know if the partial charges or the new velocities changed my results. I am looking to see how much the different partial charges has an effect on my results. Thanks!

i see a logic problem with this approach. what you describe would apply only in a case, where you can assume a linear response regime​. however, that doesn't apply for the case where the properties of your system is described by a statistical mechanical ensemble, i.e. you need to average over multiple configurations. here your perturbation is just a small random change, that will cause a divergence simply because MD is a chaotic system. that divergence (same as different randomized velocities) by itself has no meaning, unless it is significant across an ensemble of independent trajectories.

thus to be certain, that a (small) change has a consistent and statistically significant impact on your system, you have to *use* the divergence caused by randomization. i would in your situation create a significant number of de-correlated snapshots of the same (original, unperturbed) system, e.g. a collection of statistically independent frames taken from a long equilibrium simulation and use those as starting points for a collection of simulations with a perturbation. then i would look for statistically significant differences when averaging over all those simulations. only, if you can converge those results beyond the inherent fluctuations and statistical uncertainties, you have proof, that your change has a statistically significant effect.

​axel.​




Kelly L. McGuire

PhD Scholar

Department of Physiology and Developmental Biology

Brigham Young University

LSB 3050

Provo, UT 84602



--
Dr. Axel Kohlmeyer akohlmey_at_gmail.com<mailto:akohlmey_at_gmail.com> http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:21:13 CST