Generalizing Phylogenetic Parsimony from the Tree to the Forest

Abstract

Phylogenetic analysis is usually limited to considering descent relationships structured as a tree. Descent patterns with structures other than a tree characterize evolutionary systems in which entities can inherit genetic material from more than one source, i.e., hybrid ancestry. Descent patterns in a phylogenetic system with a single hybrid event can be described as the sum of two gene trees, each describing the history of part of the genetic material composing the system. Systems with more than one hybrid event will require a larger set of trees. This set of gene trees is called a phylogenetic forest. A potentially reticulate data structure, the hypertree, can represent a forest-structured phylogenetic hypothesis and simplify the calculation of parsimony costs. A workable parsimony criterion for hypertrees is the simultaneous minimization of mutation costs and the complexity of the forest. A method is presented to perform data-directed permutations on hypertrees in a heuristic search for parsimonious solutions. For any given data set, parsimonious hypertrees will range from zero hybrid events and a certain amount of homoplasy to multiple hybrid events with no homoplasy. Choosing from among this set of parsimonious hypertrees is a conceptually difficult problem that will depend on how likely hybrid events are in the particular system. This multitree approach is a simple generalization of cladistic parsimony to reticulate systems of common ancestry.