AMBER Archive (2009)

Subject: Re: [AMBER] unexplained blow-up at restart

From: case (
Date: Mon Jul 13 2009 - 11:51:46 CDT

On Mon, Jul 13, 2009, Sally Pias wrote:

> The restart coordinates (input for the second NPT run) look fine. The
> protein has simply translated a couple of Angstroms, in comparison to
> the reference coordinates (structure prior to MD). Otherwise, the
> structures are nearly identical. Note that the restart coordinates
> are from a run with nscm set to 0.

I'm coming in late to this thread, so maybe this is off-base, or has been
resolved, but here goes:

A constant pressure simulation expands or contracts the coordinates as the box
size changes. It also does that internally to the reference coordinates.
However, if you do a restart, and try to restrain to the original (un-scaled)
reference coordinates, you will get a high initial restraint energy.

I don't know if there is any simple way to avoid such behavior, or even one
that could be added to the code. I use the following, "good enough for
government work" approach: just use the restart file after the first NTP run
for both the -c and -ref flags. Of course the protein restraints won't be
exactly those of the true initial structure, but this seems to be about the
best one can do.

A better solvatebox procedure, which would give a density closer to the
desired final one, could minimize the approximations in this scheme. Try a
closeness parameter of 0.7 or 0.75 in solvateBox as a first step.


AMBER mailing list