[gmx-users] Re: Launching Replica Exchange MD
bobjohnson1981 at gmail.com
Tue Dec 19 17:57:07 CET 2006
>They need to be named remd0.tpr, remd1.tpr, remd2.tpr etc. Then mpirun -n
>4 mdrun -np 4 -s remd will do the job. "man mdrun" has the details.
Don't I have to use the -multi and -replex flags somewhere? When I
just give -s remd (i.e. the prefix for the .tpr files), I get an error
message saying that it can't open "remd.tpr".
>Sure, just grab a contiguous subset, copy them somewhere safe and run your
>test job. Be aware that depending on your system, the acceptance ratio can
>vary with temperature... see a JCTC paper (I think) from earlier this
>year. So don't stress too much about matching up the acceptance ratio to
>some pre-planned value.
I first ran some test jobs to get the potential energy of my system at
various temperatures. I then fit a polynomial to this data to get
E(T). I then iteratively solved the exchange probability assuming that
all replicas swap temperature with a 0.2 acceptance ratio. Provided
that my E(T) function well represents my actual energies, I think my
acceptance ratio should stay pretty constant between all replicas.
More information about the gromacs.org_gmx-users