Fast and accurate evaluation of phylogenetic clade support
ETH Zurich, Switzerland
I will discuss fast approximate likelihood-based methods for evaluating branch supports of inferred phylogenies (aBayes, aLRT and SH-aLRT), and how these compare to more traditional Bayesian and maximum likelihood bootstrap supports. Model violations make it almost impossible for any support measures to be interpreted as probabilities of inferred clades to be true. An interpretation of branch support based on frequentist-like error rates is more informative. The concepts are illustrated in simulations and with examples from real data.more than 7000 gene families in 37 whole genome sequences from mammalian species.