Maximum Parsimony Method in Phylogenetics

Abstract

A phylogenetic tree offers a simple way of visualizing evolutionary history. Out of many possible trees one can draw to link different species, we must have a criterion to choose the best tree, just as we need a criterion to choose the best alignment (the highest alignment score given the scoring scheme) or the best substitution model (the smallest information-theoretic indices). Maximum parsimony (MP) is one of the criteria for choosing the best phylogenetic tree, i.e., the tree that requires the smallest number of changes to account for the sequence variation is the best tree. This chapter details two frequently used algorithms used in the MP framework to obtain the minimum number of changes: the Fitch algorithm which assumes all nucleotides or amino acids replacing each other with equal frequency, and the Sankoff algorithm which uses the cost matrix to accommodate differential substitution rate among different nucleotides or different amino acids. Both belong to the dynamic programming algorithm we were first exposed to in the chapter on sequence alignment. The long-branch attraction problem, which is associated not only with MP method but also with other methods that do not correct, or insufficiently correct, for multiple substitutions, is numerically illustrated in detail. Two different ways of reporting statistical support for a phylogenetic tree are presented: the resampling method (bootstrapping and jackknifing) and the statistical test of alternative topologies. Various ways of improving the power of the statistical tests are included and discussed.