Re: reconstructing PMF

From: Ioana Cozmuta (
Date: Thu Apr 13 2006 - 13:58:09 CDT

Hi Jerome,

I used harmonic constraints for the MD simulations. Thus, what you are
suggesting is that using the umbrella sampling technique to reconstruct
the PMF from the MD data would be appropriate in this case?!

About the atoms being pushed/pulled.
1. If we were talking about a point mass then there would be no
difference between pulling and pushing.
2. Even for the case of a chain of particles if all of them would have
the same mass then again there is no difference between pushing and pulling.
3. If the pulling force is applied to the center of mass of a molecule
and not distributed between atoms with a weight that is the atom mass
then pulling and pushing will have different effects.

I should have simply asked how does the force applied to the center of
mass of the molecule act? Or better, I should have checked the source
code first and see how this is implemented. The way it is implemented
indeed it does not make a difference if the atoms are pulled or pushed.


Jérôme Hénin wrote:
> Hi Ioana,
>> I have a trajectory in which I pull a DNA molecule through a 40A long
>> channel (nanopore) using steered molecular dynamics (harmonic potential,
>> constant velocity pulling). I did reconstruct the PMF from these
>> simulations using Jarzynski's inequality.
>> Alternatively, however I would like to use a technique that uses
>> equilibrium simulations to reconstruct the PMF. For this reason I've
>> taken frames from the SMD trajectory corresponding to a 0.1A movement of
>> the center of mass of the molecule along the reaction coordinate and
>> I've run for each of these frames full (equilibrium) MD.
> DId you apply a bias on these simulations? If not, it is very likely that the
> system has drifted away from its initial position, at least in some cases,
> and I would be surprised if you achieved a good sampling of the pathway along
> the pore in these runs. In the traditional way of doing this, umbrella
> sampling, you would add harmonic biases that confine the system in a
> particular region of the channel, and then stitch the resulting histograms
> together to compute the PMF.
>> Is there any standard methodology that I could use to reconstruct the
>> free energy profile/PMF combining the SMD data with the additional
>> information from the equilibrium MD simulations? Initially I thought
>> that I could calculated the potential energies from each of the
>> individual MD simulations, use a Boltzmann weight (so that states with
>> low energy are emphasized more) and then input these energies into the
>> PMF calculations.
> As far as I can see, the only information you can easily extract from there is
> an average potential energy profile.
>> My second question is related to the SMD simulations and the way the
>> external potential is distributed over the pulled atoms. It is my
>> understanding that in the constant velocity or constant force pulling
>> the external potential is applied to the center of mass of the molecule.
>> However it would seem that the atoms that only the atoms following the
>> center of mass are being pulled while the atoms that precede the center
>> of mass are being pushed. Is this really the way SMD is implemented or
>> is the force acting on COM being distributed such that the effect on all
>> atoms is that of being pulled?
> What is the difference between atoms being "pushed" and "pulled"?
> Jérôme

Ioana Cozmuta, Ph.D
Eloret Corporation
NASA Ames Research Center
Mail Stop 230-3
Moffett Field, CA  94035
phone (650) 604-0993

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:55 CST