Evaluating Our Understanding of Different Limbs of the Metazoan Tree of Life

Different metazoan groups might present different challenges for inferring phylogeny. Reasons for this include both differences in the difficulty in recognizing distinct homologies and differences in underlying evolutionary processes. We investigate this common wisdom using simulations modeled on empirical data sets. We use compatibility to describe the structure of over 200 published matrices. (Compatible character pairs are those that do no necessarily imply homoplasy; for two matrices with the same number of taxa and characters, compatibility decreases as rates of change and homoplasy increase.) We then simulate phylogenies and matrices to match the number of taxa, characters and states per character states until we achieve match the empirical compatibility. Because most published phylogenetic inferences use minimum-steps parsimony (MP), we then find the MP tree(s) for each simulated matrix and contrast it with the correct phylogeny.

MP trees are most accurate when compatibility is high relative to the number of sampled taxa, which in turn tends to be highest among vertebrates, followed by echinoderms; trilobites and mollusks typically rank lower, with brachiopods ranking lowest. Correspondingly, simulations show that vertebrate-based matrices yield the MP trees for which over 90% of the inferred clades match real clades, and for which nearly 20% of matrices have MP trees matching the true tree. Echinoderm-based matrices yield MP trees in which nearly 90% of clades match real clades; this drops to ~80% for both trilobite- and mollusk-based matrices, and 75% for brachiopod-based matrices. Within vertebrates, dinosaur-based matrices yield the most accurate MP trees, followed closely by mammals; fish-based matrices yield the least accurate MP trees. Notably, mammal-based data sets based largely on teeth yield worse MP trees than those based on skeletal characters. These results are consistent both with familiarity breeding accuracy and more challenging underlying evolutionary models for particular groups and character systems.