Re: Scaling solvent-solute intermolecular interactions

From: Jason Swails (
Date: Wed May 14 2014 - 10:20:57 CDT

On Wed, 2014-05-14 at 13:50 +0000, Michael Bellucci wrote:

> Thank you Jason. Your approach makes quite a bit more sense to me. I
> did see the pairInteraction keyword that NAMD has and I will look into
> using it further. As a quick hack, that is, without modifying the
> source code, I wonder if it would make sense or be possible to use the
> thermodynamic integration portion of the code to do the scaling. That
> is, would it make sense to turn on thermodynamic integration, flag the
> solute or solvent to be perturbed in the pdb, and then run MD with a
> particular value of lambda? Theoretically this should have the same
> effect, but as Axel suggested, accuracy may be a concern. It isn't a
> very satisfying solution, but it might suffice for a quick and dirty
> calculation.

My experience with NAMD has been very limited. I don't know if using
the TI code will work for what you are trying to do (does NAMD use a
dual topology approach? single topology? how are the end states
defined?). If you _can_ use it for your problem, then there should not
be any accuracy concerns *specifically with regards to how long-range
electrostatics are treated* -- they should be rigorously correct (unless
NAMD _does_ neglect long-range contributions to dV/dlambda -- but I find
this unlikely). Of course finite sampling and force field errors
introduce inaccuracies, but that's another topic.


Jason M. Swails
Rutgers University
Postdoctoral Researcher

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:24 CST