Re: [AMBER] How to decompose PMF into electrostatic and van der Waals contributions

From: Jason Swails <>
Date: Thu, 26 Dec 2013 16:48:16 -0500

On Thu, Dec 26, 2013 at 9:10 AM, zhongqiao hu <>wrote:

> Hi Brian,
> Thank you for your kind reply. I am not not sure what you are talking is
> what I need. Simply speaking, for umbrella sampling, amber writes out
> instantaneous value of reaction coordinate, which is equivalent with force
> on reaction coordinate.

This is not correct. Using this value you can determine the value of the
restraint potential (this feature is used in MBAR and WHAM reweighting
schemes to resolve the unbiased PMF). You can determine the force on the
atoms in the reaction coordinate arising from the biasing potential, but
that is all. The _interesting_ part is the shape of the underlying
potential without the restraint coordinate.

If I can get decomposed components of this force
> from electrostatic and vdw contributions for each writing frame, then I can
> evaluate electrostatic and vdw parts.

This is not correct. A PMF is a potential of mean force along a
coordinate. It is the free energy averaged over all degrees of freedom of
the entire potential _except_ the reaction coordinate, so simply computing
the electrostatic and van der Waals contributions of the atoms involved in
the definition of the reaction coordinate will not help you decompose the
total PMF (since the PMF contains contributions from all of the
interactions between all of the particles).

 What I need is not electrostatic and
> vdw energies of the whole system.

Actually, the only way I can think of actually decomposing the vdW and
electrostatic contributions for a PMF requires the electrostatic and vdW
energies of the whole system and using some type of reweighting technique
like MBAR (as Brian described). I have no idea how this would actually be
done in practice, though.

> If you think cpptraj (or any others like
> MBAR) can postprocess trajectory to give electrostatic and vdw terms
> specifically for reaction coordinate, then please tell me some details so
> that I can follow.

cpptraj has a "pairwise" command that will give you the nonbonded
interactions between two groups of atoms, but this will not help you
decompose the PMF. I think Brian's response was the most helpful in terms
of giving advice about how to artificially decompose the PMF. While what
you are trying to do may sound simple, it is definitely not (and such a
decomposition is not well defined in the first place).

Good luck,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Thu Dec 26 2013 - 14:00:02 PST
Custom Search