On Wed, Jun 11, 2025, Rahul Singal via AMBER wrote:
>I am currently using Amber24 to perform normal mode analysis (NMA) through
>MMPBSA.py on a large dataset of approximately 10⁵ structures. However,
>as NMA is CPU-dependent and not parallelized in the same manner as
>other components of the workflow, the calculations are turning out to
>be extremely time-consuming. I was wondering if there are any ways to
>accelerate this step, given the scale of the dataset. Specifically, I would
>appreciate any suggestions on how to efficiently handle such a large number
>of structures—whether through parallelization strategies, offloading the
>NMA portion outside of MMPBSA.py, or any best practices that have worked
>for similar large-scale analyses. Any guidance or recommendations would be
>highly appreciated.
The variation in (estimated) entropy from one snapshot to the next is
generally quite small. So, you can run many snapshots without NMA, and just
run a small fraction of them (say 100?) with NMA. You can check the
assumption that the variation of results among snapshots is small.
[As an aside, 100,000 structures is a *very* large number: if you are just
interested in an average and a standard deviation, you might check to see if
running the entire dataset is really needed.]
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 11 2025 - 08:00:03 PDT
