AMBER Archive (2002)

Subject: Re: equilibrium problem about cut and belly

From: Thomas Steinbrecher (thomas.steinbrecher_at_physchem.uni-freiburg.de)
Date: Tue Nov 05 2002 - 11:24:54 CST

Greetings amber users,

I have encountered the same problem with the
biotin/streptavidin tutorial using AMBER7 on a beowulf
linux cluster with MPICH (See my previous post).

When I minimize the system with cutoff 999 sander crashes
and complains that the parameter MAXINT is to small and
advises me to recompile.

Minimization seems to work with a cutoff of 12 A.

If I start dynamics with cutoff 999 or 36 A, sander crashes
and tells me that MAXPR and the LIST OF NONBONDED PAIRS are
exceeded.

Dynamics with cutoff 12 A or 24 A work, but exhibit large
fluctuations of ETOT and EPOT, as reported by Dr. Sxzheng

When I look into the old mdout files provided in the
tutorial, the memory consumption of AMBER4.1 seems to be
smaller for the same system.
Is this behaviour to be expected or does it come from using
a PC cluster instead of a single machine?

Do I have to recompile sander with greater values in
sizes.h to treat proteins the size of strepavidin?

Kind regards,

Thomas

On Tue, 5 Nov 2002 08:14:51 -0800
 "David A. Case" <case_at_scripps.edu> wrote:
> On Tue, Nov 05, 2002, sxzheng wrote:
>
> > While I use cut=9.0 then the Etot of the system will
> be fluctuant strongly
> > ( about 1000 kcal/mol) even for a 1ns's dynamics.
>
> This sounds like something is wrong. But 9.0 is too
> short (generally) for
> a cutoff for free energy use, and 999 may be too long.
> Try something in the
> 12-14 Ang. range, and equilibrate carefully.
>
> ..good luck...dac