[gmx-developers] Hamiltonian replica exchange

[Matt Wyczalkowski]

The underlying functionality needed for certain free energy
calculations -- that is, the ability to evaluate the "native"
configuration (Xm) at a "foreign" Hamiltonian (En) -- is the same
functionality you need for Hamiltonian replica exchange.  This is why
the two topics are discussed together.

It actually is possible to implement Replica Exchange with GROMACS as
it is, though with a lot of overhead.  At the end of a simulation, you
can extract "foreign" energies with the -rerun flag to mdrun, with the
TRR file from the "native" simulation and a TPR file specific to the
"foreign" hamiltonian you wish to evaluate.  Based on these energies
you can perform your swaps (in some "server" wrapper script), create
new TPR files based on your swap outcomes, and start your simulations
anew.  In this way, you can implement Replica Exchange.

Its not pretty, and its not clear if the overhead will be worth it,
but its doable.

Cheers,

Matt

On Feb 19, 2008 10:51 AM, servaas michielssens
<servaas.michielssens at student.kuleuven.be> wrote:
>
>
> Forgive me to start a new thread about this but in the other the discussion
> was meanly about free energy caculations.
>
> What I wanted to do is just simple hamiltonian replica exchange to increase
> sampling. For every replica I have a different topology with a different
> hamiltonian. So the swapping criterium is:
> -betam[Em(Xn)-Em(Xm)]-betan[En(Xm)-En(Xn)]
>
> So two extra energy evaluations needed are: Em(Xn) and En(Xm) respectivly
> the energy of configuration Xn with hamoltonian m and the energy of
> configuration Xm with hamiltonina n.
>
> So my question is there an exixting function that I can use to calculate the
> energy of a configuration with another topology file? If not what would you
> be the best strategy to create one?
>
>
> kind regards,
>
> servaas
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-developers-request at gromacs.org.
>