realistic amino acid substitution models

[David Roos]

aroger at ac.dal.ca wrote:
> 
> I'm interested in improving empirically-based amino acid
> substitution model for use in phylogenetic analysis. 

[...stuff deleted...]

> Thus it is possible that "averaged" substitution models
> like that of Dayhoff, may be very poor approximations of 
> the actual site by site frequencies simply because they
> are averages of many dissimilar models.
> 
> If what I say is possibly true, it should be of interest
> to develop structure-based amino acid substitution matrices
> for a whole variety of proteins for which one of the homologues
> has been crystallized (making the assumption that the structures
> have not changed too much).  Doing this may allow us to develop
> general substitution matrices for structural elements and to
> test whether the different matrices are significantly
> different from one another (this may in turn help those
> interested in protein engineering) and in what way.
> The application of these models to protein phylogeny would
> only require the user, to provide structural information in addition
> to an alignment.
> 
> Is someone already doing this?  Does anyone know of references
> to any such investigation?

Based on a presentation at the DIMACS phylogeny workshop in Princeton 
earlier this year, Jeff Thorne (at NCState) is doing this kind of thing.  
Good luck -- D.S.Roos