From: Jérôme Hénin (jerome.henin_at_uhp-nancy.fr)
Date: Tue Dec 02 2003 - 13:50:38 CST
Well, two points about "randomness" in simulated trajectories.
First, if you use the standard NAMD way of keeping temperature constat (i.e.
Langevin dynamics), NAMD introduces stochastic forces that are pseudo-random.
The seed used by the generator is somewhere in the output, so that you can
re-run the simulation and specify the seed as a parameter. The other
constant-temperature algorithms available in NAMD are not stochastic.
The second point is that, even with a deterministic algorithm, trajectories
above a certain length are not reproducible, because of the very strong
instability of the dynamic system : even truncation errors of the floating
point arithmetic lead to differences in the trajectory that grow
exponentially with time.
Hope it answers your questions,
Jerome
> Before asking this question, I have to say that I'm not a chemist or a
> biologist. I have been running several example simulations available at
> namd homepage just because of my interest in parallel processing framework
> of namd2 (charm++).
>
> What I don't quite understand is why some of the simulations are not
> reproducible even for several sequential runs. (for example, bpti). Is
> namd2 introducing some randomness in it's own code? Is it possible to set
> on or off such randomness in input parameters.
>
> Thanks,
> Hansang.
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:37:11 CST