An evolutionary approach to the inference of phylogenetic networks

Abstract

Phylogenetic networks are models of the evolution of a set of organisms that generalize phylogenetic trees. By allowing the existence of reticulation events (such as recombination, hybridization, or horizontal gene transfer), the model is no longer a tree but a directed acyclic graph (DAG). We consider the problem of finding a phylogenetic network to model a set of sequences of molecular data, using evolutionary algorithms (EAs). To this end, the algorithm has to be adequately designed to handle different constraints regarding the structure of the DAG, and the location of reticulation events. The choice of fitness function is also studied, and several possibilities for this purpose are presented and compared. The experimental evaluation indicates that the EA can satisfactorily recover the underlying evolution model behind the data. A computationally light fitness function seems to provide the best performance. © Springer-Verlag Berlin Heidelberg 2006.