Hello,
I'm running a pretty straightforward umbrella sampling simulation with a
restraint between two ions using nmropt=1 and a DISANG file. The first run
for all windows went smoothly and everything was looking reasonable.
However, restarts are bombing out.
Investigating one window with a restraint placed on a short distance (2.375
angstrom) between the ions revealed that restarting from a conformation
where the ions are on opposite sides of the box (i used iwrap = 1 in the
initial runs). I printed out every step for 200 steps of the restart. The
minimum image distance is small, but the NMR restraints: Bond term was
giant. It's as if the NMR restraint is not calculating the minimum image
distance initially and then sticks with it; this observation is at odds
with the initial runs going smoothly (with iwrap, perhaps the NMR restraint
doesn't follow the wrap?). Step by step the trajectory shows the ions
rocketing toward one another as the NMR restraint term drops and then
everything explodes.
I found this old post that lines up well with what I'm observing:
http://archive.ambermd.org/200908/0346.html
I noticed recommendations against using iwrap. I saw one work around was to
effectively recenter the entities of interest.
Is this still an issue?
I'm using Amber20 on gpus.
Demian
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 16 2022 - 10:00:02 PDT
