Re: [AMBER] Restart Problem: The results of an integrated simulation and restarted simulations are not equal!

From: case <>
Date: Mon, 28 Feb 2011 13:56:26 -0500

On Mon, Feb 28, 2011, Ali M. Naserian-Nik wrote:
> To find out if splitting a long simulation to some shorter simulations with
> the aid of “irest=1” option, I performed two 40 ps test run for a DNA
> dodecamer: 1) 40 ps integrated simulation and 2) 20 + 20 ps successive runs
> (the 2nd 20 ps simulation was started y restart option “ntx=7” just after
> the 1st 20 ps simulation). When I compared the resulting RMSD curves of the
> backbone and also PDB files obtained at the end of the two 40 ps
> simulations, I found that unexpectedly they are not the same!

This is to be expected, depending somewhat on *how* different the two results
are. There is some truncation of precision with a restart file, and hence
some small initial deviations that eventually grow and dominate the results.
Please see section 1.3.4 of the Amber11 Users' Manual.

If you want to see for yourself what is going on, compare 40 *steps* of MD
(not 40 ps), setting ntpr=1, and then restart in the middle.

This, of course, is a repeat of my usual hobby-horse of suggesting the virtues
of short runs with ntpr=1 when problems or questions arise about what is going
on in MD simulations.


AMBER mailing list
Received on Mon Feb 28 2011 - 11:00:05 PST
Custom Search