GB vs. explicit solvent calculations

From: Thomas Steinbrecher <thomas.steinbrecher.physchem.uni-freiburg.de>
Date: Thu, 20 Mar 2003 14:53:44 +0100

Dear AMBER users,

I am conducting MD simulations on a 200 aar protein
complexed with a ligand using AMBER7 on a Beowulf linux
cluster.

I recently tried to perform a generalized Born simulation
following the guidelines in the GB tutorial on the AMBER
homepage instead of my previous approach with appr. 6000
molecules of TIP3P solvent.

I am using the standard pairwise GB model with IGB=1 and
saltcon=0.2 with a 24A cutoff. My ligand is parametrized
with the gaff-forcefield and I have applied bugfix 24

>From my output files I see that 1ps of equilibration takes
1300 sec. with explicit solvent (cutoff 12A), while it
takes 2700 sec. using the GB model, both calculated on a
dual processor PC.

I rather expected the GB calculation to be faster, even
with the inclusion of salt effects.
Did I do something wrong or is this behaviour to be
expected?
Can it be due to the size of my system or due to the
varying cutoff?
Is it possibly related to the fact that I run parallelized
sander?

I was wondering if anyone can share experiences regarding
the performance of the different GB models in terms of
speedup compared to explicit solvent simulations.

Many thanks in advance,

Thomas Steinbrecher

My sander input file is:
-----
  equilibrate structure
 &cntrl
   imin=0,
   ntc=2, ntf=2,
   cut=24.0, igb=1, saltcon=0.2, gbsa=0,
   ntpr=50,
   nstlim = 5000, dt=0.002,
   ntt=1, tempi=0.0, temp0=300.0, tautp=1.0,
   ntx=1, irest=0, ntb=0,
   nscm = 1000,
   ntr=1,
 &end
  keep all atoms frozen
  1.0
RES 1 255
END
END
Received on Thu Mar 20 2003 - 14:53:02 PST
Custom Search