AMBER Archive (2009)

Subject: Re: [AMBER] Optimization problem

From: Gustavo Seabra (gustavo.seabra_at_gmail.com)
Date: Thu Feb 12 2009 - 08:48:23 CST


Hi Daniel,

> After that I extract, with ptraj, the average structure of the last 50ps of
> the MD.
> I try to optimize this structure and I give a drms at 0.05. But the RMS
> don't converge to this value. I have a minimum value at something like
> 0.2...

Apart from the technical problems with minimization that have already
been addressed by Dr. Case and Dr. Simmerling, I wanted to mention
that you may also be suffering from using a *very bad* starting
structure for your minimization. Average structures may be useful for
some things, but I believe few people would advise you to use them as
a real (physical) representation of the molecule structure.

Imagine you have a molecule that interchanges between two
conformations during this 50ps time. If you take the average
structure, you will have a beast that is, in fact, none of the two.
There's no guarantee that your average has any physical meaning, let
alone be a structure that can be used in subsequent calculations...

You would be much better off if you do some sort of analysis of your
final data (clustering, for example), then extract a *representative*
structure of the most populated cluster. That is a real structure you
can then use for further calculations.

Gustavo.

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber