I find that attempting restart a constant-pressure MD sim that uses
harmonic restraints to the refc coordinates blows up. Here's what
seems to happen: The first NPT run starts from the restrt file of an NVT run
with T equillibrated to 300, and also uses that same NVT restrt file
for the refc coordinates. During this NPT run the volume of the box
decreases and so the positions of the refc atoms are moved with it.
Upon restarting the NPT run from the NPT run's restrt file and using
the same refc file as before (that is, the restrt from the NVT run)
the refc coordinates are not adjusted for for the difference between
the NVT box size and the (interim) NPT box size, so the protein in the
refc file is translated 1 or 2 Angstroms relative to the NPT runs
restrt file (more precicely relative to where the previous NPT run
"thought" the refc coords were). The result is a large initial
harmonic restaint energy followed by spurious heating, and finally a
vlimit crash, as the restraints try to drag the protein through the
solvent to the old refc position.
I'm using Amber8, patch-level 43, and subsequent patches don't seem to
address this. Is it fixed in Amber 9? I suppose the fix would be an
option to compare the box sizes of the refc and restrt file and scale
refc accordingly.
Don Bashford
Saint Jude Children's Research Hospital
-----------------------------------------------------------------------
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 05 2006 - 21:28:16 PDT
