THEORY

Introduction

Methods for inferring evolutionary trees can be divided into two broad categories: those that operate on a matrix of discrete characters that assigns one or more attributes or character states to each taxon (i.e. sequence or gene-family member); and those that operate on a matrix of pairwise distances between taxa, with each distance representing an estimate of the amount of divergence between two taxa since they last shared a common ancestor (see Chapter 1). The most commonly employed discrete-character methods used in molecular phylogenetics are parsimony and maximum likelihood methods. For molecular data, the character-state matrix is typically an aligned set of DNA or protein sequences, in which the states are the nucleotides A, C, G, and T (i.e. DNA sequences) or symbols representing the 20 common amino acids (i.e. protein sequences); however, other forms of discrete data such as restriction-site presence/absence and gene-order information also may be used.

Parsimony, maximum likelihood, and some distance methods are examples of a broader class of phylogenetic methods that rely on the use of optimality criteria. Methods in this class all operate by explicitly defining an objective function that returns a score for any input tree topology. This tree score thus allows any two or more trees to be ranked according to the chosen optimality criterion. Ordinarily, phylogenetic inference under criterion-based methods couples the selection of a suitable optimality criterion with a search for an optimal tree topology under that criterion.