Re: Question about ABF

From: Jerome Henin (
Date: Wed Oct 25 2006 - 19:01:33 CDT

Hi Gaurav,
The profile you obtain without bias looks reasonable, though it may need more
sampling. Even though the barrier is rather low, you will probably find it
difficult to obtain proper sampling at room temperature without any bias.
I suspect that in the second case (biased), the profile is not converged and
you are in a non-equilibrium regime. Does the reaction coordinate increase in
time and never come back? It should be possible to prevent that by using a
significantly bigger value of the fullSamples parameter, better suited to the
timescales of your problem.
Note that the force constant (forceConst parameter) is not related to the
biasing force, but rather defines the stiffness of the harmonic boundary
In that respect, slow degrees of freedom that are not the RC but are coupled
to it are always a problem. In your case case, one thing you probably want to
consider is relative rotation of the two chains. How do your simulations
account for that?

In all cases, I cannot insist too much on the interest of visualizing
trajectories (VMD is fine for that) to get a feeling of what happens in

To answer your question about the deca-alanine example, the constraints we
considered were the lengths of C-H bonds, fixed using RATTLE.

On Wednesday 25 October 2006 19:03, Gaurav Sharma wrote:
> Hi All,
> I am using the ABF method to study the free energy change when two
> chains of my protein move apart during the MD run (this conformational
> change is induced by electrostatic repulsions between the chains). For
> this I used the distance-com RC method by defining a set of atoms on
> each chain as abf1 and abf2. I was not sure if I should use the external
> bias force since the two chains move apart by themselves so I performed
> two simulations one with applyBias on and one with it turned off
> expecting to get the same result in both cases. But the two results are
> markedly different as can be seen from the attached file.
> I need some insight as to how to interpret the results to find out which
> is a more appropriate method. I also read the original Henin & Chipot
> paper on 'Overcoming free energy barriers....' The results for
> deca-alanine (Fig 1) gives free energy profiles for unconstrained and
> constrained simulations. Maybe I missed it but what are the constraints
> in the simulation?
> Thanks in advance for your help.
> Gaurav

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:44 CST