[gmx-developers] modifying replica exchange to allow Hamiltonian exchanges

[Shirts, Michael (mrs5pt)]

Hi, Ilja-

> Ideally, one would like to have a single generalized HREX ensemble,
> yet with a possibility of decoupling vdW and Coulomb (up to three
> independent lambdas (I only need two at the moment)). This is in
> contrast to having to do (up to three) independent HREX runs.

This is already in the code scheduled for inclusion.

> Right
> now I am doing both changes (Coulomb and vdW) at the same time, which
> benefits from adjusting sc_alpha and sc_sigma slightly to get better
> acceptance ratios. I am mostly using your p=1 potential form.

I have discovered lots of problems doing this.  When one moves to different
systems with different charges/LJ terms, the sc_alpha and sigma required to
get good exchange significantly.  It is almost always more statistically
efficient to do coulomb linearly, then do vdw. p=1 is a good form to use.

> I
> would like to have an option of having multiple independent lambdas
> for different residues when I have to mutate several in one structure.

We are planning to do this, but this is HIGHLY nontrivial. This will take a
while to get in.  But it's definitely in the works.

> I would like to help test the implementation that is capable of
> nonlinear HREX against my script. Let me know if that would be of
> value.

I'll email you off list with more information.

Best,
~~~~~~~~~~~~
Michael Shirts
Assistant Professor
Department of Chemical Engineering
University of Virginia
michael.shirts at virginia.edu
(434)-243-1821