Re: ABF reproducibility

From: Jérôme Hénin (jhenin_at_ifr88.cnrs-mrs.fr)
Date: Mon Feb 14 2011 - 06:29:40 CST

Hi Parisa,

Yes, MD simulations of biomolecules are essentially stochastic (even
those that are formally deterministic), and ABF is no less stochastic.
What should be reproducible between ABF runs is their statistical
properties. On this subject, note that the sampling histogram is very
slow-converging (slower than the PMF). I recommend following it on a
log scale (e.g. plot RT times log of sampling, to get something
commensurate to a free energy).

Cheers,
Jerome

On 13 February 2011 22:08, Parisa Akhski <Parisa.Akhshi_at_chem.queensu.ca> wrote:
> Dear NAMD experts,
>
> This may seem simple, but I have tried to run ABF simulations with exactly
> the same input files to test the reproducibility of the results. I know that
> after overcoming the energy barrier, the ion moves under its diffusion (ion
> channel dynamics) and I expect to get the same results using the same input
> files. But, the sampling profile and also the force behavior I get, are not
> the same (sampling at one side of the window is 25K while for the other
> simulation it is 250k). Do we expect to get exactly the same profiles
> (especially the same SAMPLING profile)? If not, what determines this? Is
> there any random factor involved?
>
> Many thanks,
> Parisa

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:56:38 CST