AMS2003-4: Bayesian Methods for Phylogeny Independent Detection of Positively Selected Amino Acid Sites

Daniel Merl, Raquel Prado, and Ananias Escalante
12/31/2003 09:00 AM
Applied Mathematics & Statistics
A positively selected amino acid site is one for which natural selection encourages diversification. The identification of such sites is of biomedical importance, as diversifying sites cannot act as reliable binding sites for location-specific drugs. We introduce a new method for detecting positive selection based on a class of Bayesian generalized linear models (GLMs). This method does not require explicit assumptions about phylogeny and offers relatively reduced time to Markov chain Monte Carlo (MCMC) convergence. We compare our Bayesian GLM approach with three current methods for detecting positive selection: Nei and Gojobori’s ADAPTSITE, Yang’s PAML, and Huelsenbeck and Ronquist’s MrBayes.