On Thu, Jun 27, 2019, Alen Ahmetovic wrote:
>Furthermore, with the cluster I am using, I am unfortunately limited to
>just under a week of simulation time per job, which after some testing, I
>can get a 20 frame subset entropy calculation done (out of a total of 5000
>frames). I am not sure if this subset is enough for an accurate result.
Note that you don't need to do all the entropy caclulations in a single run.
The averaging is just linear, so you can run several shorter jobs and average
the results afterwards.
Since the calculations are so expensive, check often to see if you are getting
significantly different results for different frames. For many systems, the
entropy estimate doesn't change much from frame to frame.
The number of frames needed for convergence can be quite system-dependent, so
it is good that you are running tests on your system.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jun 28 2019 - 06:00:08 PDT
