Re: Computing Potential of Mean Force --SMD-cv

From: Peter Freddolino (
Date: Wed Jul 12 2006 - 02:25:53 CDT

Dear Ramya,
the tutorial on pulling decaalanine
and particularly the section on analysis of CV pulling results
should give you a good template for how to get the PMF out of these
simulations. (b) looks fine to me. For (c),
> c)And, Is W(t) evaluated as:
> W(t)= {/ 0 to t(v*f(t)*dt)} - k/2[R(t)- R0 -v*t]^2
> where,
> / 0 to t =integral 0 to t (sorry!! I had to limit my use of symbols)
> R(t)-R0 = Extension
> v= constant velocity
> k= spring constant
> f(t) = force
Perhaps I'm misreading what you have here, but since the pulling force
is the force that you're measuring the work of, it seems that you have
it twice in the equation (once as f(t) and once by the explicit equation
on the right). The work done by the pulling force at a constant velocity
is just the integral portion of that equation (derived from a
substitution of the basic W = int(f(x) dx) when you know x as a function
of v and t), so you can just plug in the time-based force data you have.

> Is it wise to use a dielectric of 80 for water instead of performing the
> simulation in water box ?
Having water in your simulation does a good deal more for you than just
provide a dielectric constant; it also interacts with your solute
(providing solvation effects), gives you a viscous medium, and provides
realistic screening effects for short range interactions (where the
effective dielectric constant is assuredly not 80). There's a good deal
of ongoing research on implicit solvent models (you can google, for
example, for implicit solvent model and/or generalized born), but NAMD
does not currently feature any of these models.


This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:20 CST