[gmx-users] g_msd, REMD, trestart...

David van der Spoel spoel at xray.bmc.uu.se
Sat Mar 15 22:37:25 CET 2008

Julián GC wrote:
> Thanks for your help, Mark.
> I cannot demux the trajectories, because I'm running each box at with a different temperature
> and with a different initial configuration (appropiate to that temperature)
> (This, combined with the replica exchange, is usually called "parallel tempering").

Of course you can demux. The question is only how to determine the 
diffusion constant in view of the change temperature. You can probably 
do it however, by slightly modifying the kinetics algorithm we 
introduced for protein folding (Phys. Rev. Lett. 96 pp. 238102 (2006)).

An alternative would be to use the velocity autocorrelation function, in 
between exchanges only.

> Checking the g_msd -h I saw that I can save some time using g_msd -e and -b (end and begin 
> time for the MSD measurement), but that's the best I can see right now.
> I do see the discontinuities you mention, thats what I would like to avoid. And, I agree: if I 
> realize I need statistics on blocks longer than 100ps, I'm gonna have to try something else,
> at least reducing the replica exchange frequency
> Thanks again, bye,
> Julian
> Mark Wrote:
>  GC wrote:
> />>/ 
> />>/ Hi all,
> />>/ 
> />>/ I'm studying small water clusters with Gromacs 3.3 (I'm also performing 
> />>/ MC with a self-writed MC program).
> />>/ I run 24 boxes in paralell, with -replex 100000 (it is important to use 
> />>/ replica exchange).
> />>/ 
> />>/ I need to get MSD results, but I don't think g_msd takes into account 
> />>/ that every 100ps the boxes are being exchanged.
> /
>>Constant-T REMD trajectories are intrinsically discontinuous, so g_msd 
>>will also have discontinuities. If all of your replicas started from the 
>>same structure, then the MSD calculation is measuring from the same 
>>starting point. If you need statistics for longer continuous chunks, 
>>then you need to reconsider your approach. If you can adjust suitably 
>  temperature, then generating demux-ed constant-replica REMD 
>>trajectories will at least eliminate the discontinuities. (demux.pl is 
>>either in the source distribution or on the contributions section of the 
>>GROMACS webpage, I forget which)
>>>/ Does anyone knows how to fix that? Have anyone had to deal with 
> />/> something like this before?
> />>/ (The obvious "easy" solution is to chop the trajectory in blocks of 
> />/> 100ps, calculate MSD in each block, and then average all the results, 
> />>/ but would be nice if this annoying "shortcut" could be avoided)
> /
>>See g_msd -h... I think -trestart saves you some legwork here.
> Tarjeta de crédito Yahoo! de Banco Supervielle. Solicitá tu nueva 
> Tarjeta de crédito. De tu PC directo a tu casa.
> Visitá www.tuprimeratarjeta.com.ar
> ------------------------------------------------------------------------
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php

David van der Spoel, Ph.D.
Molec. Biophys. group, Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205. Fax: +4618511755.
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se

More information about the gromacs.org_gmx-users mailing list