Using position-dependent diffusion to simulate long colvars trajectories

From: Ajasja Ljubetič (ajasja.ljubetic_at_gmail.com)
Date: Thu Apr 03 2014 - 07:36:28 CDT

Dear all,

When doing *free energy* calculations the *original dynamics of the colvars
is lost* as it is perturbed by the added bias. So it seems one can get
either free energy (by applying a bias) or dynamics (rate constants for
transition between states) using unbiased MD.

Wouldn't it be nice if it were possible to *obtain rate constants from free
energy calculations* as well? It turns out that this is indeed possible
under the assumptions that the colvars undergo Brownian
motion<http://en.wikipedia.org/wiki/Brownian_motion>(i.e. the colvars
behave diffusively). The reason behind this is that the
bias does not perturb the diffusive motion of the colvars.

A few *hundred nanosecond-long biased MD* trajectory can be* extended to
millisecond* lengths in the colvars subspace at a very small additional
computational cost. First one runs the biased MD simulation to obtain the
free energy. From the biased MD trajectory of the colvars, *position
dependent diffusion coefficients *are extracted. The free energy and the
diffusion coefficients are used to simulate very long colvar trajectories.
>From the long colvar trajectories rate constants and other dynamics data
can be obtained.
(Side note: existing biased MD trajectories can be analyzed as well).

The tricky part is obtaining position-dependent diffusion coefficients.
This is the problem we tackled
recently<http://scitation.aip.org/content/aip/journal/jcp/140/8/10.1063/1.4866448>
 (J. Chem. Phys. 140, 084109 (2014); a PDF is available on our
website<about:invalid#zClosurez>).
We also provide an open source
Package<http://scitation.aip.org/content/aip/journal/jcp/140/8/10.1063/1.4866448>
for
calculating position dependent diffusion coefficients and generating long
diffusive trajectories.

Since many people are using NAMD with ABF or metadynamics calculations to
estimate free energies, I hope this e-mail (and the package) are useful.

Best regards,

Ajasja Ljubetič
Young researcher
Laboratory for biophysics <http://lbf.ijs.si>
Institute Jozef Stefan, Slovenia

*Disclaimer*: In case it is not clear -- I'm the author of the mentioned
paper and package.
High level questions (eg. is this diffusive approximation even useful?:)
are probably on topic for the mailing list, but *low-level stuff* (eg.
problems with installations, etc...) should be addressed directly to me *off
the mailinglist*.

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:17 CST