[gmx-users] Re: Resplica exchange with solute tempering (REST)

[Mark Abraham]

On 29/03/2011 4:54 AM, Sai Pooja wrote:
> There was a typo in the mail:
> E_protein-protein = Potential - E_water-water - E_protein-water
>
> On Mon, Mar 28, 2011 at 11:55 AM, Sai Pooja <saipooja at gmail.com 
> <mailto:saipooja at gmail.com>> wrote:
>
>     Hi,
>
>     I have tried to implement a variant of Replica exchange called
>     REST using gromacs. In REST, the potential energy is scaled in the
>     following manner:
>
>     For the nth replica:
>
>     E_n = E_protein-protein + a_n * E_protein-water + b_n * E_water-water
>
>     I used tables to implement this. Separate tables for the three
>     components were generated with the corresponding scaled
>     potentials. Then for replica exchange I use gromacs to run MD for
>     1ps then extract the Energy corresponding to the three components
>     using g-energy and exchange replicas based on the acceptance
>     criteria outlined in the paper.
>
>     E_water-water = LJ_water-water + Coulomb_water-water
>     E_protein-water = LJ_protein-water + Coulomb_protein-water
>     E_protein-protein = Potential - E_protein-protein - E_protein-water
>
>     Is there some obvious error in this implementation? I use CHARMM
>     forcefield with tip3p water and alanine dipeptide. The paper uses
>     OPLS-AA with tip4p and is able to use 3 replicas -300k, 420k,
>     600k. I am not able to reproduce these results i.e. the same
>     acceptance ratio with 3 replicas. 5 replicas gives something close
>     to the desired acceptance ratio.
>

The exchange ratio depends on the overlap integral of the distributions 
of the potential energy. To have a chance of  reproducing their work, 
you need to use similar numbers of degrees of freedom (i.e. number of 
atoms) so that the ratio of width/average of your distributions is similar.

Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110329/cc913a49/attachment.html>