Re: questions about ABF simulations

From: Jérôme Hénin (
Date: Wed Nov 03 2010 - 05:51:48 CDT


On 1 November 2010 15:49, mjyang <> wrote:
> Thanks for the detailed explanation of Jerome and Giacomo. It is helpful to
> my understanding of this method.
> After reading the papers about ABF, I still have some confusions about the
> sampling along the defined order parameter. In ABF theory, the acceleration
> of the order parameter is elminated by the biased force.

Not in the current version of ABF. That was one of the early ideas
published Darve and Pohorille in 2000-2001. In what is now called ABF
(anything published after 2001), only the *time-averaged* force is
canceled. The order parameter still experiences a stochastic

> which parameters will impact the efficiency of the sampling? If a
> dihedral is used as the reaction coordinate from 0 to 180 degree, how long a
> MD simulation is required to sample the whole range?

It is very difficult to tell beforehand, because the simulation is
still stochastic.

> In my protein system, I
> defined the dihedral between the center of mass of a larger immobile domain
> and a smaller flexible domain.

>From this description, you seem to be looking at large-scale motion.
Even if the free energy barriers are smoothed out by ABF, diffusive
sampling might take a while.


> ________________________________
> From: [] On Behalf Of
> Giacomo Fiorin []
> Sent: Friday, October 29, 2010 4:33 AM
> To: mjyang
> Cc:
> Subject: Re: namd-l: questions about ABF simulations
> Hi Mingjun:
> 1) lowerWallConstant and upperWallConstant shouldn't play a role, unless you
> set either boundary on or near the barrier. But the barrier should ideally
> be at the center of your interval, to be sampled correctly.
> 2) No, the ABF method samples forces on the variables that you use, and does
> not compute the free energies from the histogram of the variables' values
> (such as e.g. umbrella sampling with WHAM, metadynamics etc would do).
> 3) Yes, this seems a very sensible starting point.
> Giacomo
> ---- ----
> Dr. Giacomo Fiorin
> ICMS - Institute for Computational Molecular Science - Temple University
> 1900 N 12 th Street, Philadelphia, PA 19122
> ---- ----
> On Thu, Oct 28, 2010 at 2:26 PM, mjyang <> wrote:
>> Dear NAMD users,
>> It is the first time I use the adaptive biased force simulation (ABF) to
>> study the relative free energy change between two distinct conformations of
>> a protein. After reading the corresponding mannual and tutorial, I still
>> have some confusions required your kind help:
>> 1. If the barrier is not known along an order parameter, how large should
>> I set the value for the force constant for efficient sampling, e.g.
>> lowerWallConstant and upperWallConstant?
>> 2. Is WHAM used to calculate the free energy based on the simulation
>> samplings?
>> 3. How large should I set the "width" keyword? I performed 1000,000 MD
>> steps and printed the standard deviation by turning on the "analysis". Is it
>> ok to set the value of "width" to 2 times of the largest deviation?
>> Many thanks.
>> Mingjun

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:56:19 CST