From: Jérôme Hénin (jerome.henin_at_uhp-nancy.fr)
Date: Wed Apr 19 2006 - 16:20:29 CDT
In the spirit of what Peter suggests (using a Tclforces script), you could use
the ABF sampling method with a custom reaction coordinate. This should remove
the issue of reconstructing the PMF from nonequilibrium work data. Coding a
new reaction coordinate for the code distributed with NAMD is a simple matter
of adding a Tcl file defining a few procedures. When writing such a file,
existing reaction coordinate should provide useful templates.
Please let me know if you need help doing this.
Le Mercredi 19 Avril 2006 21:00, Peter Freddolino a écrit :
> You may, however, be able to avoid playing with the source code by using
> tclforces (http://www.ks.uiuc.edu/Research/namd/2.6b1/ug/node32.html)
> instead; you can get it to work along an arbitrary path if you're
> careful. However, Lewyn's comments on calculating energies still apply.
> LEWYN LI wrote:
> > Raul,
> > I have always shied away from playing with the source code,
> > because I never knew if I have inadvertantly introduced errors into
> > the program.
> > On a more theoretical level: since work is not a state function,
> > its value depends on the speed at and the path along which you pull
> > the ligand out of the protein. Therefore, it would be tricky to
> > compare the work value from simulations to experimental data. You
> > could try to apply the "Jarzynski Equality" and calculate the free
> > energy, but this might require many, many trajectories (depending on
> > how far the system is out of equilibrium and other considerations) and
> > its convergence is difficult to verify.
> > Hope this helps!
> > LEWYN
> > On Wed, 19 Apr 2006, Raul Alcantara wrote:
> >> Hello:
> >> I have the following problem: I want to find the average work done to
> >> pull a small ligand through the matrix of a protein... however the
> >> possible escape path is not a straight line. So my question to any
> >> people who have played with the source code is it worth my effort to
> >> modify the code to do a pulling experiment in a curved path (would this
> >> severely hurt performance) or would I just be better off by breaking the
> >> curve into smaller linear segments and doing independent pulling
> >> experiments for each of those.
> >> Thank you
> >> --
> >> Raul E Alcantara
> >> Washington University in St Louis
> >> Center for Computational Biology
> >> 700 S Euclid Ave
> >> Saint Louis MO 63110
> >> Tel (314) 362-8778
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:57 CST