Re: [AMBER] Unable to read coords from rst

From: Jason Swails <jason.swails.gmail.com>
Date: Mon, 27 Jan 2014 04:47:59 -0500

On Sun, Jan 26, 2014 at 10:04 PM, Yip Yew Mun <yipy0005.gmail.com> wrote:

> Hi,
>
> I have recently ran a 50 ns MD simulation, where a rst file is written.
> Using this rst file, I plan to continue running the simulation. However,
> the error message "unable to read coords from *.rst" appeared. I read the
> archives and they said that it's due to the protein drifting out of the
> periodic box and I can expect to find **** in my rst file. However, I'm
> unable to spot **** in my rst file, so how can I resolve this problem?
>

Does your restart file have all of the coordinates and velocities that it
should? This should be easy to check. A restart file with N atoms and no
velocities should have N/2+2 lines in the file (1 line is a header line)
if you are running with explicit solvent and N/2+1 lines if you are running
with implicit solvent. If velocities _are_ present, there should be N+2
lines for an explicit solvent restart and N+1 lines for an implicit solvent
one. (If your system has an odd number of atoms, there will be 1 extra
line)

My guess is that your restart file is only partially written, in which case
there is nothing you can do to recover it. The best you can do is extract
the last good snapshot from your trajectory using cpptraj and create a
restart file from it to continue your MD simulations from that. This is
known to happen from time to time if your simulations do not completely
finish and are killed by a scheduling system.

A couple things you can do to avoid this kind of file truncation is to use
NetCDF restart files (ntxo=2 in the &general section of your input file),
set ntwr to a negative number so that each restart written out is unique
and does not overwrite the one before it, or make sure that your simulation
will finish in the given wallclock time.

Good luck,
Jason

-- 
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jan 27 2014 - 02:00:02 PST
Custom Search