AMS2005-27: Detecting Patterns of Natural Selection in DNA Sequences using Bayesian Generalized Linear Models

Daniel Merl and Raquel Prado
12/31/2005 09:00 AM
Applied Mathematics & Statistics
Drug resistance in pathogens can emerge as a result of the rapid evolution of relatively small regions of certain target proteins experiencing positive natural selection. We describe a new statistical methodology for inferring the rates of molecular evolution for an alignment of protein-coding DNA sequences, with emphasis on the problem of identifying individual amino acids sites experiencing positive natural selection. The method draws upon the theory of Bayesian generalized linear models in order to create a robust and flexible estimation framework, capable of representing the various biological factors influencing selective pressure while accounting for the unique uncertainties that characterize the evolutionary dynamics of rapidly evolving pathogens.

This report is not available for download at this time.