In article <35chan$at9 at quartz.ucs.ualberta.ca>, max at mycroft.mmid.ualberta.ca (Max Cummings) writes:
> Hi,
>> Lately I have been trying to review and summarize what people are doing
> to approximate solvation effects in energy calculations. Our interest
> is in molecular docking so we are looking for computationally cheap
> approximations since ultimately we want to incorporate such a term
> into the energy calculation which is used in our docking procedure.
>> Measuring the changes in buried/exposed surface areas of various atom
> types (for eg. C, N/O, N+, O-, S) and multiplying the sums of these
> changes by an "atomic solvation parameter" which converts these
> areas into an energy contribution seems to be the commonly pursued
> line. Indeed, I have done some analyses of the results obtained in
> docking simulations which did not involve such a correction
> in the energy calculations during the simulation and these
> results indicate that such a correction will yield better
> results in our docking simulations.
>> However...
>> There seems to be a variety of these correction parameters
> floating around and I have not found a set of tests (published)
> which really convinces me about the correctness of a
> particular set of correction parameters. Also, there are surprisingly
> few tests of these kinds of calculations published (see next
> sentence). Online searches of such things as "solvation" or
> "surface burial" yield either very few or way too many
> references.
Vila et. al published an independent set of solvation parameters
a few years back. The trouble is that randomized conformations that
obey excluded volume and are just extended coils have significantly
lower energies than their respective native structures. (Proteins 10:199-218).
This paper serves as a good reference for previous work however such
as that of Ooi et. al (PNAS 84:3086-3090) and others...
We tried to use various sets of solvation parameters as extra energy
terms in the AMBER, CHARMM, and ECEPP potential functions for the
purpose of tertiary structure prediction by conformational search. Under
no circumstances were we capable of making the native structure the
lowest energy conformation. It may be in the presence of explicit
rather than virtual solvent, but it just didn't work for us. In all
cases, we located "inside-out" structures with exteremely low electrostatic
energies which vastly outweighed the solvation energies (by inside out,
I mean structures with most polar groups hydrogen bonded to one
another in the core of the structure wnad with most hydrophobic groups
dangling outside).
Overall, it probably helps to keep you near the native structure if
you start there, but if you attempt to perform a conformational
search starting from randomness, I think you're doomed...
Scott
