Phylogenetic analysis of nucleotide sequences: an algebraic approach.

Abstract

In a previous paper Schmidt and Mueller [Bull. Math. Biosci. 58 (3) (1996) 449.] proposed a new method for calculating phylogenetic trees from dichotomous (present/absent) properties of the taxonomic units. The proposed method is based on a specified distance measure quantifying the phylogenetic information content of the selected properties with respect to the joint evolutionary history. Formally, the method can be applied to arbitrary property patterns and is not based on restricting evolutionary models and optimization principles looking forward in evolutionary time. Instead, its theoretical foundation is the retrospective postulate that properties which are present in any two (closely related) taxa were also present in their most recent common ancestor. The reliability of the derived trees depends essentially on the phylogenetic relevance of the chosen properties. To apply this method to related nucleotide sequences I have defined a set of biologically meaningful nucleotide properties which separate and evaluate the different substitution events. The proposed algorithm provides additive trees and needs no numerical constraints to avoid negative branch lengths or any other meaningless result. In a series of applications the method has been tested and compared with the outcomes of other established tree reconstruction methods. For demonstration, I report and discuss the results of the reanalysis of two families of genomic sequences. Here and in a series of further applications I obtained trees which are biologically acceptable and in good coincidence with the published trees. Thus, it can be concluded that the proposed method will be a useful alternative for the study of phylogenetic relationships.