Phylogenetic Hypothesis Testing

Abstract

Abstract Probabilistic methods of phylogenetic reconstruction (maximum likelihood and Bayesian inference) provide a robust framework for the statistical inference of phylogenies and for phylogenetic hypothesis testing. The likelihood ratio test is a powerful statistical tool for comparing the explanatory power of alternative phylogenetic hypotheses that strongly relies on the likelihood function. In the Bayesian framework, phylogenetic hypothesis testing is usually performed using Bayes' factors. Statistical tests involving phylogenies are widely used for comparing alternative phylogenetic hypotheses (tree topologies), studying evolutionary rate constancy (clock‐like behaviour) of deoxyribonucleic acid (DNA) and protein sequences, estimating changes in selective pressures at the sequence level, inferring ancestral character states at internal nodes on phylogenies, estimating historical biogeography, examining patterns of trait evolution along the tree or applying phylogenetic comparative methods, among others. In addition, molecular dating analyses allow formulating and comparing hypotheses on absolute divergence times of cladogenetic events. Key Concepts: Probabilistic methods of phylogeny reconstruction rely on explicit models of evolution, and provide a powerful framework for testing evolutionary hypotheses. The likelihood principle provides a direct approach to hypothesis testing in terms of probability. The likelihood ratio test is a powerful tool for comparing two alternative hypotheses. In the Bayesian framework, hypotheses are usually compared using Bayes' factors. Topology tests evaluate whether the analysed data, besides supporting a particular phylogenetic hypothesis, could support or reject alternative tree topologies. Molecular dating analyses, by converting evolutionary distances into absolute ages, allow comparing alternative scenarios of temporal divergence on phylogenetic trees. The effect of natural selection can be detected by measuring the relative rates of synonymous and nonsynonymous substitutions in multiple nucleotide sequence alignments of protein‐coding genes. Explicit models of trait evolution support probabilistic reconstructions of ancestral character states at internal nodes on phylogenies and testing of hypotheses regarding the process of trait evolution. Phylogenetic comparative methods aim to identify adaptation through correlations between species traits and other variables, while accounting for phylogenetic nonindependence of data.