Estimating evolutionary rate of local protein binding surfaces: a Bayesian Monte Carlo approach

To infer protein function by matching local surface patterns, an effective scoring matrix for evaluating surface similarity is critical. In this study, we develop an evolution model of binding surfaces using a continuous time Markov process. We develop a Bayesian Markov chain Monte Carlo method to estimate the substitution rates of amino acid residues with specialized move sets. We then develop scoring matrices of residue similarity specific to a functional site and show how they can be used to identify similar binding surfaces, and how such information can be used for predicting biological roles of proteins. Our method is especially effective in extracting evolutionary information from the phylogeny of sequences homologous to a protein structure, all of which may be of unknown functions.