Estimating the history of mutations on a phylogeny

Evolution is a historical process that has left its signature on the molecules and morphology of living organisms. Attempts to better understand the specific and general features of evolution involve making inferences about the past from these tell-tale signs (Felsenstein, 1985; Brooks and McLennan, 1991; Harvey and Pagel, 1991; Pagel, 1999). Ancestral state reconstruction is a powerful tool in this endeavor, as exemplified by its application to a wide array of questions (e.g. Langley and Fitch, 1974; Gillespie, 1991; Templeton, 1996; Messier and Stewart, 1997; Bishop et al., 2000). Traditionally, ancestral reconstruction has relied on well-understood approaches such as parsimony. However, the last decade has seen an excited flurry of research into statistical approaches as exemplified by the contents of this volume and the primary literature. In a general sense, most methods for ancestral reconstruction have focused on reconstructing the ancestral states at the internal nodes of a phylogeny. Often we are not interested in particular nodes of the phylogeny but the whole history of a character. In this chapter we focus on a Bayesian method for estimating these histories on phylogenies (we refer to a complete description of a character’s history as its mutational path). Mutational path methods differ most notably from other approaches in their ability to estimate not only the ancestral states at the internal nodes of a phylogeny but also the order and timing of mutational changes on the phylogeny. Our goal here is to provide a concise introduction to the statistical tools necessary for estimating mutational histories and making inferences from these histories, and to provide some examples of the power of this recent approach.