Confidence limits on the branching order of phylogenetic trees.

Abstract

We describe a confidence test for branching order that can aid protein phylogeny reconstruction as well as the evaluation of the optimal tree. It is proposed that the process resulting in the observed amino acid residue differences, which is the basis for the identification of the order and relative times of divergence events, is appropriately described by a modification of the negative binomial distribution. The relative total numbers of mutations (accepted and nonaccepted), which result in a given number of amino acid differences, may be obtained as the expectation of this distribution. The associated variances enable significant differences in tree branching order to be established. If the total rates of mutation of the genes encoding the compared proteins are equal, the expected total mutations and their associated variances map identically to their relative times of divergence. In addition, significantly different rates of change (due to differences in total mutation rate and/or acceptance rate) may be identified without the requirement of outlying reference group. The method is equally applicable to phylogenies derived from DNA or RNA sequence information.