Re: [AMBER] PMF for dimer dissociation

From: Jason Swails <>
Date: Tue, 17 Mar 2015 22:33:44 -0400

On Tue, Mar 17, 2015 at 10:04 PM, Sanmeet Chahal <>

> There are various different types of monomers such as acetic acid,
> pyridine, benzene etc. They can be kept rigid by doing ntr = 1 and applying
> a large force constant to the required residues (usually just the first
> two).

​This will keep them fixed in place, it won't let you scan along the
distance between them.

Benzene is basically already rigid, since it is planar with stiff bonds
that are usually SHAKEn. For small molecules, you can typically apply
distance restraints between all pairs of atoms in each molecule in order to
maintain a rigid geometry, but still permit free rotation and translation.
 (Just don't make the force constant too large, or you will introduce
integration errors due to the high frequency motions.)

> > On Tue, Mar 17, 2015 at 5:18 PM, Sanmeet Chahal <>
> > wrote:
> >
> > > Hello All,
> > >
> > > I need to compute a PMF for the dissociation of a dimer in TIP3P
> water. I
> > > need each dimer to remain 'rigid' from an intramolecular and relative
> > > standpoint. My PMF transformation is simply moving only one monomer
> away
> > > from the other monomer with a simple translation of their atoms. The
> goal
> > > of this calculation is to mimic a PB-type PMF along the separation axis
> > but
> > > now with proper statistics and explicit solvent.
> > >
> > > I thought that thermodynamic integration would be the right formalism
> for
> > > this in AMBER (though umbrella sampling could be more efficient).

​TI scans along an alchemical coordinate -- lambda -- that interpolates
between two Hamiltonians. The PMF you seem to be looking for is a
potential of mean force along a particular degree of freedom (so it's a
free energy averaged over every *other* degree of freedom). Since that
reaction coordinate is a *real* degree of freedom, TI will not help you

There are many methods that you can use to compute a PMF: steered MD with
Jarzynski's equation. Umbrella sampling. Metadynamics. Purely unbiased
sampling (assuming the simulation can sample sufficiently along the whole
PMF). Replica exchange (T-REMD and H-REMD). etc.

The most straightforward conceptually is either Umbrella sampling or
steered MD, IMO. The former has a more helpful tutorial at


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Tue Mar 17 2015 - 20:00:03 PDT
Custom Search