Re: AMBER: MM_PBSA problem

From: Xin Hu <hux.mail.rockefeller.edu>
Date: Wed, 13 Apr 2005 13:49:05 -0400

> I think people may be getting sidetracked here. The nmode program is
> always
> compiled in double precision (as is all of Amber). You may be thinking
> about
> compiling for 64-bit architectures, to allow very large memory use. But
> you
> should only consider doing this after you are thoroughly familiar with the
> operation of these programs on smaller problems. I would guess that most
> attempts to do normal mode calculations on systems with more than 10,000
> (even 5,000) atoms end up being more frustrating than scientifically
> illuminating.
>
> ...dac
>


Hi Dr. Case,
Thank you for your note. Actually I run the example in ~/amber8/src/mm_pbsa,
the "ras_raf_II_wat_lig" (for nmode analysis) is about 1200 atoms, I guess
it should work fine with most of systems.

I run the example on a node of linux cluster (2G memory each node), I notice
that there are actually more than 1G memory free while the nmode calculation
stucking there, seems not a memory problem. The Amber compiled as serial,
and it does not related to the MPI.

I also tested the example on my SGI workstation (1.6GHZ CPU, 2G memory), and
the nmode works fine, it took several mininus for the entropy calculation.
However, when I try my protein system consisting of about 4600 atoms (about
260 residues) on the SGI, the same problem happened with the nmode
calculation. It is taking 3 hours now and still not finish the first
calculation (and also there is a large free memory).

Does anyone have an experinece with MM_PBSA (nmode) calculation on a problem
more than 5000 atoms? or what's the reasonable system that nmode can work
fine? As a general method on protein/protein or protein/ligand calcuation, I
do not expect the MM_PBSA (nmode) should has difficulties with 5000 atoms
(around 300 aa) calculation.

Thanks for suggestions (still my problems with the test running on linux)



Xin


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Apr 13 2005 - 18:53:01 PDT
Custom Search