Datasets were simulated as described in "Recoding amino acids to a reduced alphabet may increase or decrease phylogenetic accuracy" by Peter G. Foster, Dominik Schrempf, Gergely J. Szöllősi, Tom A. Williams, Cymon J. Cox, and T. Martin Embley Dataset lengths are 10 million sites, which were divided into 100 alignments of 100,000 sites each for analysis.