Bioinformatic analysis of protein families for identification of variable amino acid residues responsible for functional diversity

Proteins within a single family usually share a common function but differ in more specific features and can be divided into subfamilies with different properties. Availability of genomic, structural, and functional information implemented into numerous databases provides new opportunities for bioinformatic analysis of homologous proteins. In this work, new method of bioinformatic analysis has been developed to identify subfamily-specific positions (SSPs) - conserved only within protein subfamilies, but different between subfamilies - that seem to play important role in functional diversity. A novel scoring function is suggested to consider structural information as well as physicochemical and residue conservation in protein subfamilies. Random shuffling is performed to rank results by significance, and Bernoulli statistics is applied to calculate p-values. Algorithm does not require predefined subfamily classification and can propose it automatically by graph-based clustering. This method can be used as a tool to explore SSPs with different structural localization in order to understand their implication to structure-function relationship and protein function. Web interface to the program is available at http://biokinet.belozersky.msu.ru/zebra.