Date: Mon, 5 Apr 2004 10:37:00 -0600 (Mountain Daylight Time)
From: "Thomas E. Cheatham, III"
Subject: Re: AMBER: question about image


> Could you please tell me How I can find out if the solute drifts over time
> during MD? In what situation imaging is needed?

...since this came to the whole list, I will respond to the whole list.
In summary, by looking at the energy, RMSd, etc. in general you cannot
infer whether the solute has drifted or if imaging is needed... Some
indicators of large translational motion are increases in EKCMT (the
kinetic energy of the center of mass in translation) or of the need
to image sudden/instantaneous spikes in the RMSd plot vs. time (as a
molecule jumps across the periodic boundary).

Although those two bits of information are indicators, they are not
sufficient and it is likely best to routinely *look* at a movie of the
data, PRE-imaging, even including water. Your eye is rather excellent at
discriminating motion, holes in the water box, and other common problems.
A number of excellent programs exist such as MD Display by Lybrands group
(discussed in recent posts to the reflector by Chris Moth), VMD from
Shulten's group, MoilView from Simmerling's group, among others. To
visualize systems with water, I usually color the water a dark color (like
blue)...

I will admit that in my early days we got burned a number of times by NOT
looking at the graphics, such as when we saw the flying block of ice
problem. What happened is that we only looked at the imaged trajectories
and didn't understand why the fluctuations kept dropping over time in
longer runs. When we viewed the un-imaged trajectory, it became readily
apparent that the whole system was simply translating through the box.
Vacuum bubbles can also be a problem if the system was not
well-equilibrated at constant pressure. There are other issues as well
that can be easily visualized, so in summary, look at the movies...

--tom