ASTRAL: Fast coalescent-based computation of the species tree topology, branch lengths, and local branch support

Siavash Mirarab
Department of Electrical and Computer Engineering
The University of California at San Diego

Species tree estimation in the face of gene tree discordance due to ILS is a central goal of many phylogenomic projects. In this talk, we introduce ASTRAL, a summary method for estimating species trees from unrooted gene trees. We briefly discuss ideas behind ASTRAL's algorithm for inferring the tree topology, and then focus on ASTRAL's new features for fast computation of branch support and branch length. ASTRAL calculates tree topology, branch lengths, and branch supports using frequencies of quartet trees in gene trees. We describe connections of analyzing quartets to the multi-species coalescent model and provide theoretical justifications for our approach. We further show empirical applications of ASTRAL to real and simulated data. We end by pointing out current limitations and directions for future improvements.