>> B-factors (related to RMSF but not quite) are thermodynamic properties,
>> so as long as their value is converged, it could not possibly matter how
>> often you saved frames.
>>
>> Sampling for convergence is of course really hard to do, but in regular
>> MD it depends only on how long you run.
>
> So theoretically a uniform sampling of the two runs (equivalent to
> ntwx=500) should give about the same result as the concatenation of
> ntwx=100 and ntwx=250 runs.
Dear Bill,
Indeed, you should get the same answers in the limit of a long run. You
do not even need uniform sampling ! You just need proper Boltzmann
sampling, intrinsic to properly run MD.
>
>> So, if you believe you have run long enough, you can directly compare
>> RMSF or B-factors for two different simulations with different ntwx
>> without any need for downsampling.
>
> How does one decide what is long enough? Is there a heuristic, other
> than "as long as one can afford"?
Ah !!! THE question...
If anyone knows, please answer to the list !.
Short answer: no one knows how long is long enough. You must look at
trends vs time (B-factors every 10 ps), or cumulative values vs time.
Better yet, you can run two independent simulations from different
initial conformations and velocities and see if your two runs give you
similar values. There is no easy answer, but one MUST look and at least
attempt to ask if what you have done so far 'looks' like good enough.
a.
--
Dr. Adrian E. Roitberg
Associate Professor
Quantum Theory Project
Department of Chemistry
Senior Editor. Journal of Physical Chemistry
American Chemical Society
University of Florida PHONE 352 392-6972
P.O. Box 118435 FAX 352 392-8722
Gainesville, FL 32611-8435 Email adrian.qtp.ufl.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
to majordomo.scripps.edu
Received on Sun Jun 01 2008 - 06:07:17 PDT