[gmx-users] REMD, broken simulations and analysis

[Berk Hess]

>From: Luca Mollica <YLU at zurich.ibm.com>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: [gmx-users] REMD, broken simulations and analysis
>Date: Fri, 21 Oct 2005 15:36:39 +0200
>
>Dear users,
>
>I am experiencing the REMD options of mdrun for the GMX 3.3 version on a
>system with 26 replicas, with a protein, water and ions:
>I have to say that it works pretty well and it is really easy going to set
>upa REMD simulation with these new mdrun flags.
>But I have a couple of questions about it before finishing a nanosecond
>simulation and starting to analyze it.
>
>1.  Due to maintenance operations on the cluster onto wich REMD is
>running, the simulation has been interrupted, and I have made it
>restarting using the tpbconv command like in ordinary
>  simulations: is there any consequence on the reliabiliy of Replica
>Exchange method applications in this case ? does the tpbconv command works
>correctly in this case ? Or does it have any different usage mode ?
>
>2. After the end of the simulations, what do I have in my hand ? I can see
>my working directory full of as many trajectories and log and energy files
>as the number of replica I have for my systems: the log files tell me when
>there happened to be an exchange between replicas, but what does the
>trajectory file represents exactly ? Is it the trajectory of replicas
>across temperature space or is it a collection of states at a certain
>temperature ?

Trajectories are written by temperature, so continuation with tpbconv
should work fine, if you manage to continue all simulations from the same
timestep.

grep REPL md0.log gives you all the REMD information, including statistics.

Trajectories can be resorted to be continuous with trjcat, see:
trjcat -h

Berk.