Amino Acid Substitution Matrices Estimated by Maximum Likelihood

Abstract

The present work describes protrates, a program that estimates amino acid substitution matrices and among-site substitution rates based on their likelihood for a given tree topology and a dataset of aligned proteins. The issue of producing maximum likelihood (ML) rate matrices over protein data have been adressed under the framework of general-purpose unbiased substitution matrices [1, 9], since it's believed that parsimony-estimated matrices can be systematically distorted and ignore backward and parallel substitutions - which can affect branch lenghts of ML topologies [