On Tue, Jul 8, 2014 at 10:12 PM, Him Shweta <shwetahim.gmail.com> wrote:
> Dear all,
>
> I am running entrop calculation in MMPBSA.py for a quadruplex-ligand
> system with an input file as below:
>
> Input file for running entropy calculations using NMode
>
> &general
>
> endframe=100, keep_files=0,
>
> /
>
> &nmode
>
> nmstartframe=1, nmendframe=100,
>
> nminterval=1, maxcyc=10000, drms=0.1,
> nmode_igb=1, nmode_istrng=0.0,
>
> /
>
>
> Here, in this calculation when i am using convergence criteria (drms =
> 0.1), the result is more close to my experimental result, while with
> drms=0.001, the result does not match or is closer to my experimental
> result.
> The thing here which is confusing to me is, can i use a convergence
> criteria for minimized energy gradient as 0.1 (drms).
>
> Please give your input and suggestions.
>
I think that the answer to your question is "no, you cannot use 0.1" as
the minimization convergence criteria. I suspect what is happening in
your calculation is that when you set the convergence criteria to 0.1, the
minimization stops farther away from a true local minimum, meaning that
more of the normal modes will have negative frequencies. At a true
stationary point representing a local minimum, all eigenvalues of the
Hessian will be positive (and these eigenvalues correspond to normal mode
frequencies).
Negative eigenvalues are omitted from the entropy calculation, so if using
drms=0.1 leaves you farther from a local minimum, it's likely that more
modes have negative frequencies (and are therefore omitted) which could
artificially lower the entropy estimate. But this is a consequence of
doing a bad normal mode calculation, _not_ of improving your model. You
can look at the normal mode output files to see how many modes have
negative frequencies. It could be that making drms smaller could also
improve your results, since I think low frequency motions are more
sensitive to proximity to a local minimum (and the low frequency motions
dominate the entropy contribution).
Hope this helps,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jul 08 2014 - 23:00:03 PDT
