From: "Xavier Periole"
Subject: Re: AMBER: modelling a Ca binding site
Date: Thu, 14 Aug 2003 13:07:26 -0400


Hi Karsten,

first of all simulations of calci-proteins is a subject in its own and
lots of people have tried and stoped because the parametrization
of the calcium interacting with the protein is kind of very sensitive.
Some collegues (from Toulouse) have published a paper about the
parvalbumin and determined free energy of changing a Ca in Mg
in different sites of the protein (Allouche D. J Mol Biol. 1999;285(2):857)
The results are quite good and done with CHARMM. Note that
only the binding site was allowed to adjust to the modification. After
I was develloping new potentials for calcium and others cations
and we just got to the conclusion that we could have a better
description of the structure of the binding site of the calcium/magnesium
when compare to the crystal structures but the energetics was out of
range. The reason is that calcium is extremely polarizable and so
the two body description of its interactions is not accurate enough.
Basically it is very delicate to try to study the specific effect of a
cation
on a protein.

This work was done with a very high resolution x-ray structure. MD on
homology model is also something very delicate to manage. It has been
observed that with less than 50-60% identity it is almost impossible to
rich the quality of a NMR structure compare to a x-ray structure. I
guess that if you restrict your system on the binding site you could do
some ab initio simulations. For the rest of the protein it is gonna to be
difficult to differentiate between the effect of cation presence/absence
and the fact that you have a model contructed by homology.

Hope it'll help,
XAvier