Re: adaptive biasing force with RMSD colvar

From: Jérôme Hénin (
Date: Sun Mar 20 2016 - 05:49:54 CDT

Dear Haleh,

Your approach has several issues. First, free energy of conformational
change in proteins is typically very difficult to calculate, and ABF is far
from being a magic bullet in that respect. It will only succeed if applied
carefully by someone who knows the approach well. Second, RMSD is rarely
good enough as a coordinate to achieve that. You may be able to drive the
transformation, but it is probably irreversible (you are stuck in the
nonequilibrium regime of ABF).

My advice is:
1) get a better understanding of ABF and the difficulties that can occur
(see Comer et al JCTC 2015).
2) have a look at other methods, either path-based (string) or using
confinement (
3) if ABF seems like the best choice, look for other coordinates that may
describe the transition better than RMSD


On 17 March 2016 at 17:39, Haleh a <> wrote:

> Dear NAMD user's,
> I am applying ABF for a conformational change between two states of a
> protein. The overall RMSD is 3.15 and I use RMSD colvar to calculate the
> PMF value for such a conformational change. I keep the wall constant,
> Number of samples, rmsd and atom groups constant while changing only the
> simulation time. The whole conformational change under ABF takes place in 1
> ns. However, I observe different PMF values if I prolong the simulation. I
> repeated the simulation for 1, 2,3,4,5,10,15 and 20 ns long simulations and
> for each case I get PMF values within 508-660 kcal/mol where the smallest
> PMF value does not correspond to the shortest simulation. My question is
> why am I getting such a different PMF values for different simulation
> lengths? And how can I optimize the simulation time? If the full
> conformational change is achieved within 1 ns, is that much time enough for
> ABF application ?
> Many thanks,
> Haleh Abdizadeh

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:53 CST