[gmx-users] how to decouple a solute (a) only from itself and not the solvent, and (b) leaving the LJ repulsive term unmodified

[Christopher Neale]

Dear Users:

I'm trying to induce a protein to unfold and I'd like to use a free-energy-like lambda value to specifically weaken only the attractive part of the LJ interactions within the protein, while leaving LJ interaction between protein and water unmodified. Is there any way to do this in gromacs?

I've had some success with the converse approach of using the free energy code to strengthen the LJ and coulombic interactions between protein and water by removing the grompp errors in the source code when lambda is greater than 1, recompiling, then in the .mdp file setting lambda=0 to none and lambda=1 to vdw-q and then running at lambda=1.5 with couple-intramol=no. One problem with this approach is that when I use couple-intramol=no it kills the parallelization due to the exclusions, so a simulation that could previously run on 8 cores now maxes out on a single core since domain decomposition is impossible on more cores. Another problem is that even though in the thermodynamic limit it should be equivalent to weaken one interaction or strengthen another, I suspect that I'd see much more rapid unfolding by weakening the intramolecular protein attractions than by strengthening the protein-water interactions.

Thank you for your help,
Chris.