From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Mon Apr 05 2010 - 20:44:59 CDT

Hi Rachel,

Rachel Ruskin wrote:
> However, the results I am getting for internal energy (that is,
> selecting protein only) don't make much physical sense. The energy
> difference between conformations is 2-3 orders of magnitude larger
> than expected.
>
The real problem here is the assumption that you can just follow an
interpolation of the coordinates and expect reasonable results -- this
is rather akin to trying to get a tennis ball from one side of the court
to the other by following a path directly through the net. One should
not be surprised if unphysical intermediates are encountered. To get
what you're after, you need to have something resembling a physically
correct pathway -- there are lots of ways to look for this, but
unfortunately, none of them are cheap. You might try targeted molecular
dynamics (TMD) as an initial method to get a very crude putative
transition path, but really what is needed is something like umbrella
sampling or ABF (if you can define a good reaction coordinate between
the conformations) or transition path sampling (if you cannot).
> My questions:
> Does the protein internal energy take into account the shielding from
> the protein? (are two charges on opposite sides of the protein
> shielded by the intervening charges?)
>
All "shielding" within the protein is really the result of
rearrangements of the charge distribution within the protein; in
atomistic MD with explicit solvent (the usual usage case for namd) the
shielding is explicit because all the charges in the protein will
respond to your hypothetical charges on opposite sides.
> Am I missing some obvious reason NOT to use NAMD Energy for this calculation?
> Additional, possibly off-topic question: should the protein-lipid
> interaction energy and the protein-water interaction energy be
> included in the potential energy landscape (if I'm interested in the
> rate of transition over the potential energy barrier from one protein
> conformation to another)
>
You really want the free energy barrier here, and yes, all of those
contributions should ideally be considered, as they will contribute to
the transition. However, given what you've described, it is unlikely
that you have the resources to properly sample the solvent and lipid
coordinates for any given protein conformation. You may be better served
by using a different program that allows an implicit solvent model,
which will at least reduce the computational burden to get reasonable
intermediates in your pathway.

Best,
Peter
> Thank you very much,
>
> Rachel Ruskin
> Amherst College, senior thesis student
>