Long-Branch Attraction Bias and Inconsistency in Bayesian Phylogenetics
We simulated 5,000-nt sequences along randomly-selected four-taxon trees with branch lengths drawn from a uniform distribution on (0,1]. a, Datasets were divided into strong and weak Felsenstein zone (FZ) and inverse Felsenstein zone (IFZ) groups based on the pattern of branch lengths on the true tree and the difference between branch lengths (see Methods). The proportion of replicates in each category from which ML and BI recovered an incorrect phylogeny is shown. Bars indicate standard error. b, The proportion of datasets from which each method inferred the topology with the two longest terminal branches as sister taxa. The label “true” indicates the proportion of datasets for which the true tree has the two longest branches as sister taxa.