Re: [AMBER] Old problem with restart

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 16 Aug 2011 17:10:22 -0400

Hi Bruno,

On Tue, Aug 16, 2011 at 4:27 PM, Bruno Rodrigues <bbrodrigues.gmail.com>wrote:
>
>
> Well, sure, but I'm looking at solvation shells, and not having the solue
> completely filled with water might be a problem. I also observed a strange
> behavior when the DNA comes out of the 1st box. You can see from my last
> email an image where the DNA is shifted to border.
>

This is expected during wrapping. What happens is that sander/pmemd looks
to see if the COM (or maybe center of geometry) of a "molecule" (a closed
set of bonded atoms) has moved outside of the primary unit cell (which is
arbitrarily defined!). If it has, it moves it *into* the other side which
just means that all of those molecules are moving one box over, so the
molecule in the "displayed" box has moved in from the box on the other side.
 This is strictly a visualization artifact. The actual calculation takes
into account every periodic image, using the minimum image convention (that
is, we calculate where all of the periodic images of a particular atom are,
and use the shortest one for any direct space calculation). Thus, it
doesn't matter how good a particular restart file "looks", which is affected
by how the unit cell is defined (we typically like to define it with our
system of interest at the center of the box and the waters surrounding it,
but that doesn't have to be), the energy of those restart files will be
exactly the same, and the simulation taken from either of those restart
files will be indistinguishable (of course the results won't be 100%
reproducible over long time periods due to the limitation of machine
precision and restart file precision, but the point remains valid).


> > I'm comparing the heating rate, one was heated up from 300K to 400K in
> 4ns
> and the other in 40ns. The restart crashed at 395K and 36ns, just a few
> steps from the goal!
> There are "***" characters in the restart file. If I generate new restart
> files from the last frame of the simulation without velocity, does it
> change
> a lot my dynamics?
>

Yes, the dynamics will change significantly. This is true even if you keep
the same velocities and just change the random seed on a stochastic
thermostat or barostat, since such systems are chaotic by nature (so it is
not necessarily a problem). The problem with using a ptraj-generated
restart file (which doesn't *really* have no velocity information, all
velocity components of all atoms are set to 0), is that your temperature
drops to 0 at that point, in which case you have to wait for your system to
heat back up before you can claim that collected snapshots are members of
your desired ensemble. If you have a restart file from a previous step (the
last restart file that worked), just use that one and set iwrap = 1. Don't
worry that your box looks "weird" with your system sticking out of half of
the box. The results are generated as though your system is periodic (and
replicated infinitely in every direction), and you can always postprocess
snapshots with ptraj afterwards to get them centered the way you want.

Hope this helps,
Jason

-- 
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Aug 16 2011 - 14:30:02 PDT
Custom Search