Re: free energy perturbation

From: Jérôme Hénin (
Date: Wed Jan 05 2011 - 07:39:37 CST

Hi Bin,

The sentence is slightly ambiguous. What it means is this:

The convergence analysis requires "backwards" energy differences.
Suppose we have three lambda-points, a, b, and c. The forward
calculation gives us samples of deltaU_ab at point a and deltaU_bc at
point b. For convergence analysis, we need a histogram of deltaU_ba
sampled at point b, which we don't have. That is the job of the
backwards calculation. (In the old linear paradigm, that was not
necessary because there was a simple, linear relationship between
deltaU_ba and deltaU_bc).

So while it is true that the free energy difference can be computed
from a one-way run, the two-way calculation is needed for more
elaborate analyses (which include overlap sampling and Bennett's
acceptance ratio method).


On 5 January 2011 09:07, BIN ZHANG <> wrote:
> Hi, All:
> I have a question about on comment in the FEP tutorial (Page 14), which I
> quote as following:
> " Since a soft-core potential [12, 13] has been introduced to avoid the
> so-called “end-point catastrophes”, the potential energy no longer varies
> linearly with the coupling parameter . It is, therefore, necessary that the
> reverse transformation be carried out explicitly to access G+!, as is
> proposed in the above NAMD scripts."
> I do not fully understand the logic here. Why the reverse transformation has
> to be carried out to access the dG in nonlinear case? One should be able to
> calculate the energy difference, which is needed to calculate dG, in either
> case.
> Any input is greatly appreciated.
> Best,
> Bin

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:56:31 CST