AMBER Archive (2002)

Subject: Sander Recompilation for the Biotin-Streptavidin Tutorial?

From: Thomas Steinbrecher (thomas.steinbrecher_at_physchem.uni-freiburg.de)
Date: Wed Oct 30 2002 - 09:46:30 CST

Dear AMBER users,

at the moment I try to do the Biotin-Strepavidin Tutorial
from the AMBER Homepage using AMBER7.
It is a protein in which one binding site is surrounded by
a water cap. The total system contains about 8000 atoms.

During the minimization step, sander failed before starting
the calculation giving:

sander .out file:

----------
     Static Integer Memory requirement of: 2823807 exceeds
MAXINT of 2000000

| Memory Use Allocated Used
| Real 2000000 620512
| Hollerith 400000 47441
| Integer 2000000 2823807

| Max Nonbonded Pairs: 5400000
  ** Redimension and recompile
----------

I used the input file provided by the tutorial, plus ntb=0
(turning off the periodic box seemed necessary to me for
this system)

----------
 &cntrl
  imin=1, maxcyc=200, ntpr=20,
  scee=1.2, nsnb=999, ntmin=2,
  ntb=0,
 &end
----------

My first question is, is there anything wrong with my input
leading to so much memory consumption? In the original
tutorial files (using AMBER 4.1) a value of MAXINT = 800000
was sufficient for the calculation.

I presume that I have to recompile sander with different
parameters in sizes.h. Please correct me if this is not
necessary.

My second question is, if I have to recompile sander I want
to use values as big as possible in sizes.h to avoid
problems like this.
How big can I make the parameters on a normal PC (512 MB
RAM) without running into problems?

Kind regards,

Thomas

PS: Sorry for the long post, I was not sure what
information is important and what not.