From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Mon May 01 2017 - 10:10:18 CDT
Hi Jim,
On Fri, Apr 28, 2017 at 9:25 PM, Jim <jim.jim.strong_at_gmail.com> wrote:
> Hello,
>
> I want to obtain free energy profiles of passages of molecules through
> nanopore systems immersed in an external electric field. I can use
> different free energy sampling schemes (cv-SMD/Jarzynski, (*)ABF, US,
> US-REMD, etc) to obtain PMFs. I have a couple questions related to that and
> to new NAMD releases:
>
> 1. Is there a method that is preferred (aiming at quantitative accuracy
> while possibly ignoring computational costs)?
>
>
> - a) For example, in regard to ABF, what can (and should) be used in
> the current NAMD version 2.12 (or CVS 2.13): ABF vs eABF vs on-the-fly eABF
> vs gABF vs egABF vs ...?
>
> I get confused between all these flavors and their applicability to the
> particular problem/system of interest.
>
>
You should read the associated papers to understand the differences between
free energy methods. You do not need to know each method in depth to
realize that they differ either in the way of enhancing sampling of rare
states (e.g. by a biasing potential, constraint, or external force) or in
the method of measuring the relative free energy (a histogram, or the
integration of the work performed, or the biasing potential itself).
All methods will converge in the thermodynamic limit of large sampling, but
the rate of convergence depends on the particular case, and also on *your*
criterion for quantitative accuracy! (A PMF is not a single number.)
Very loosely speaking, the rate of convergence will be around 1 over the
square root of the number of independent/uncorrelated samples for all
methods.
>
> 2. Release notes of NAMD 2.12 in regard to Collective Variable Module:
>
> http://www.ks.uiuc.edu/Research/namd/2.12/features.html
>
> claim
>
> http://colvars.github.io/totalforce.html
>
> while a message from the mailing list from Mar 2015 reads:
>
> http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l.201
> 5-2016/0433.html
>
> This confuses me. If I want to go ahead with *new* ABF simulations under
> external electric field (that will affect *all* atoms in the simulation box
> regardless of CV), what should or should I not do to ensure accurate final
> results?
>
If you want to do a new calculation, there is nothing special to do for
that issue. Both the old and the new conventions are correct. In the
older convention, the electric field will not appear in the system force,
while in the new convention it will. (Hence the reason why you shouldn't
switch conventions during a simulation).
My message in March 2015 only applies to the older convention, of course.
>
> 3. Again from the release notes:
>
>
> - "Scripting command "cvcflags" to optimize performance of complex
> colvars"
>
>
> I don't understand where and how the command "cvsflags" is and can be
> used. Is there an example usage?
>
This is a subcommand of the Tcl scripting interface that allows you to turn
on/off individual components. If you define a colvar with a single
component (i.e. a single function, the typical case) you won't find use for
that command. If you are in the context of doing a scripted superposition
of components, the documentation of this subcommand is sufficient:
- *colvar <name> cvcflags <flags>: for a colvar with several cvcs (numbered
according to their name string order), set which cvcs are enabled or
disabled in subsequent evaluations according to a list of 0/1 flags (one per
cvc).*
It sounds, however, like you are looking for general examples of uses of
the Colvars module and/or free energy calculations. Try looking for
examples here:
http://colvars.github.io/
Giacomo
>
> Thanks.
>
>
>
>
>
>
>
>
-- Giacomo Fiorin Associate Professor of Research, Temple University, Philadelphia, PA Contractor, National Institutes of Health, Bethesda, MD http://goo.gl/Q3TBQU https://github.com/giacomofiorin
This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:21:15 CST