Likelihood-Based Inference of Phylogenetic Networks from Sequence Data by PhyloDAG

Processes such as hybridization, horizontal gene transfer, and recombination result in reticulation which can be modeled by phylogenetic networks. Earlier likelihood-based methods for inferring phylogenetic networks from sequence data have been encumbered by the computational challenges related to likelihood evaluations. Consequently, they have required that the possible network hypotheses be given explicitly or implicitly in terms of a backbone tree to which reticulation edges are added. To achieve speed required for unrestricted network search instead of only adding reticulation edges to an initial tree structure, we employ several fast approximate inference techniques. Preliminary numerical and real data experiments demonstrate that the proposed method, PhyloDAG, is able to learn accurate phylogenetic networks based on limited amounts of data using moderate amounts of computational resources.