AMBER Archive (2005)

Subject: Re: AMBER: large std error in entropy estimation

From: David A. Case (case_at_scripps.edu)
Date: Thu Jun 23 2005 - 13:25:03 CDT


On Wed, Jun 22, 2005, Ye Mei wrote:
>
> I am trying to calculate the binding free energy of a drug molecule to a
> certain peptide. I used the mm_pbsa module in amber8. But the vibrational
> contribution to entropy(times T) often has very large std error, like
> -3.00+/-13.00Kcal/mol. Even though I pick out 20 snapshots from last 5
> picoseconds when the system already in equilibrium.

I think this is kind of unusual, and you will probably have to examine the
results from individual snapshots to learn more. Are one or two structures
out of line with the rest? Is the result dominated by a very small number of
low frequency modes? It could be that your system is more variable than most
(you say "peptide", which may imply lack of a well-defined structure); if that
is the case, you may need a lot more sampling, or a completely different
method. Use of a very small number of snapshots (like 20) is something that
seems to work on well-folded proteins.

...good luck...dac

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu