On Wed, Jun 04, 2014, Jason Swails wrote:
>
> Absolute binding free energies from the standard MM/PBSA protocol are
> wholly unreliable.
I want to add a different perspective here:
(1) What is the "standard MM/PBSA protocol"? For better or for worse, the
"mmpbsa" scripts in Amber give users a lot of rope to hang themselves. In
particular, the "advanced" options for non-polar solvation appear to be
rather sensitive, and were calibrated against relative free energies, not
absolute ones. My recommendation is that estimates of absolute binding
energies should use a simpler "SA" model, and pay close attention to
getting good estimates of entropies, and to convergence issues.
(2) The term "wholly unreliable" is subject to interpretation here. If
you are not getting "reasonable" estimates of absolute binding energies
(say within 10kT of an answer that is known to be reliable), then there
are things in the model that are not correct. This *could* have to do
with the approximations made in the implicit solvent modeling, but is more
likely (in my view) to involve problems with construction of the model
(protonation states, binding poses, metal ion participation, ...) and with
convergence issues. Errors in these latter features are likely to have
bad effects on TI/FEP calculations as well.
The MM/PBSA model is a toolkit for thinking about what is going on in
ligand-protein binding (or in other types of conformational change).
Individuals are of course welcome to avoid using the method, judging it to be
unreliable. But useful insights can be gained as well, although they are
rarely in the form of a number that represents a binding affinity that tells
you which ligand to sythesize.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 04 2014 - 12:30:03 PDT
