AMBER Archive (2007)

Subject: Re: AMBER: Entropy in MM_PBSA

From: David Mobley (dmobley_at_gmail.com)
Date: Fri Aug 24 2007 - 11:31:28 CDT


Sergey,

These are very good questions, and may be one of the reasons the
quality of results from MM-PBSA can be very system-dependent. In
addition to it being hard to compute entropies using PCA or NMODE,
such methods, even when they work, may not give reliable estimates of
the real entropy of the system. In other words, it's really hard to
calculate the actual total entropy of the system. As a consequence,
it's also hard to test how much leaving out the entropy degrades the
results. As you note, one can hope that entropic changes might be
similar on binding different ligands, meaning that entropic
contributions might cancel in relative affinity calculations --
although I've seen a reasonable amount of experimental (and
computational) data suggesting that this cancellation probably does
not occur.

Anyway, I don't have any great answers for you at this point. However,
Michael Gilson does have a new entropy expansion that he has just or
is about to release; it might be better suited to estimating entropies
than the methods you cite. It may be worth looking into.

Best wishes,
David Mobley

On 8/20/07, Sergey Samsonov <sergeys_at_biotec.tu-dresden.de> wrote:
> Hi AMBERs,
>
> I'm studying the binding of the receptor (~200 aa) with ligands (5 aa),
> and my aim is to calculate the binding energy in the simulation. As for
> MM-PBSA part everything works without any problem. I'd like to ask for
> advice related to entropy component of binding energy. Following the
> tutorial (A3: MM-PBSA) on the AMBER official site, it seems that
> entropic contribution could be neglected. Indeed, it seems not possible
> to calculate the entropy using NMODE tool since it needs extremely
> expensive calculations (moreover it gives segmentation fault because of
> memory problems, which as I found, also was once mentioned and answered
> at the Mailing List). So I tried PCA with 'therm' option but it seems
> that the calculations are also too demanding even for small number of
> frames (though without segmentation faults now). Is there any other way
> to calculate the entropies of binding? Or maybe to obtain the
> quantitative proof that neglecting the entropies doesn't crucially
> influence the calculations? Since the final aim is to compare
> ln(Kbinding) with the binding energies, there would be no problem in
> neglecting entropies if for all the complexes the entropic component
> approximately the same or differs in the range of the data deviation a
> priori.
>
> Thank you in advance!
>
> Sergey
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu