Re: ABF (fwd)

From: Jerome Henin (
Date: Tue Nov 14 2006 - 08:49:03 CST

Dear Matthew,
I'm afraid the case you are studying is far from easy to tackle using ABF, for
two reasons: ABF is about reversibility, and the NAMD implementation is
one-dimensional (for now).
The dissociation of a receptor-peptide complex is very hard to describe by a
single geometric parameter, and it is even harder to simulate that process
reversibly: is the association likely to be observed over MD timescales?
Actually the two problems are very much linked: if all relevant geometric
parameters could be properly accounted for and biased, then reversible
association and dissociation would be more easily observed.

The question is: what information do you want to obtain? Is it really the PMF?
Is it a qualitative description of the process? Is is the free energy of
binding? In the latter case, computing a PMF is probably not the way to go. A
lot of work has been done in that field, e.g. using FEP with or without
restraining potentials. I can give you specific references if you don't have
For a qualitative description, SMD (pulling as slowly as you can afford!) may
be easier to control than ABF.

There is a second ABF-related issue, not that important though: in your
simulations, some non-equilibrium pulling seems to have happened. That kind
of undesirable effect can be avoided by using high values of the fullSamples
parameters - the slower the system is, the higher it should be.


On Tuesday 14 November 2006 08:40, Matthew Davies wrote:
> Dear fellow NAMD users
> We are trying to implement the ABF technique to study the dissociation
> of a receptor-peptide complex. The receptor binds a single
> peptide of 12 amino acids in length. We have applied the technique
> by defining the peptide as one centroid and the binding groove of the
> receptor as the other centroid and in the simulation we have observed a
> clear dissociation of the peptide from the receptor.
> Our problem has been in the calculation of the A(xi) values for the
> simulation, specifically in selecting the values of xiMin and xiMax
> values. At the initial point of our simulation, the two centroids are
> approximately 10.4A apart. When we sample the A(xi) values between 10-20A
> and 15-20A, we get singificantly higher values for the former. We are
> confused by this as we had thought the initial and final points of sampling
> would not significantly affect the determined values. Can you suggest at
> to why this might occur? I have attached both .dat files as examples.
> Best wishes
> Matthew Davies

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:49 CST