FunDi

FunDi is a phylogenetically aware, maximum-likelihood mixture model based prediction program for identifying functionally divergent/specificity determining amino acid residues among/between protein families and sub-families.

Details on the algorithm can be found in the Bioinformatics Publication, currently in Advance Access:
Gaston, D., Susko, E., and Roger, A.J. (2011) A Phylogenetic Mixture Model for the Identification of Functionally Divergent Protein Residues. Bioinformatics (Advance Access)

FunDi as used in the publication can be downloaded here

Current versions of FunDi with bug-fixes are now being maintained at GitHub