Re: Umbrella Sampling Methodology

From: Giacomo Fiorin (
Date: Tue Aug 17 2010 - 10:58:46 CDT

Hi Charlie,

1) Is the methodology that I have outlined above correct?

Unless I'm forgetting something obvious but essential to umbrella sampling,
it should be. Overlapping windows => continuous coverage of phase space.

> 2) I will apply the harmonic restraints via the colvar options, is this the
> correct method and is my colvar script correct?

I suppose the protein is held fixed by position restraints? Otherwise, you
shouldn't use dummyAtom, but define the COM from atomic positions (so that
it can be updated with time). Otherwise the script looks OK: obviously,
you'll need to change the center of the harmonic restraint for the different
windows. To collect the histograms, you can either do it as post-processing
from the file colvars.traj, or use the "histogram" method inside the colvars
module: that counts all values at all time steps, so it has better
statistics: the price is that you have to set the colvar's boundaries and
width from the beginning. Be also aware that changing the "width" will
affect forceConstant.

> 3) Are there any guideline on the optimum length of simulations. I am
> concerned that if the simulations are too long then there will be little
> phase space overlap between windows while if the simulations are too short
> then the molecule would not be equilibrated.

I would look in the literature at umbrella sampling simulations for systems
similar to yours: it's hard do devise general rules for things like this.

> 4) Are there guidelines as to the number of windows per nm displaced or is
> this to be determined afterwards for viewing the generated histograms?

Same as above.

> Finally I would like to note that the NAMD/VMD tutorials currently
> available are exceptionally well written and enable those of us with more
> modest capabilities to preform worthwhile studies, while I believe that an
> umbrella sampling tutorial would be a welcomed addition. I realize that my
> system could also be addressed (more robustly and intuitively) using steered
> molecular dynamics and the Jarzynski equality but he who pays the salary is
> partial to umbrella sampling.


---- ----
  Dr. Giacomo Fiorin
  ICMS - Institute for Computational Molecular Science - Temple University
  1900 N 12 th Street, Philadelphia, PA 19122
---- ----

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:25 CST