CAT-PMSF improves phylogenetic modelling under maximum likelihood and resolves Tardigrada within Panarthropoda, as the sister of Arthropoda plus Onychophora.

Abstract

Tardigrada, the water bears, are microscopic animals with walking appendages, that are members of Ecdysozoa, the clade of moulting animals that also includes Nematoda (round worms), Nematomorpha (horsehair worms), Priapulida (penis worms), Kinorhyncha (mud dragons), Loricifera (loricated animals), Arthropoda (insects, spiders centipedes, crustaceans and their allies) and Onychophora (velvet worms). The phylogenetic relationships within Ecdysozoa are still unclear, with analyses of molecular and morphological data yielding incongruent results. Here we use CAT-posterior mean site frequencies (CAT-PMSF), a new method to export dataset-specific mixture models (CAT-Poisson and CAT-GTR) parameterized using Bayesian methods to maximum likelihood software. We developed new maximum-likelihood based model adequacy tests using parametric bootstrap and show that CAT-PMSF describes across-site compositional heterogeneity better than other across-site compositionally heterogeneous models currently implemented in maximum likelihood software. CAT-PMSF suggests that tardigrades are members of Panarthropoda, a lineage including also Arthropoda and Onychophora. Within Panarthropoda, our results favour Tardigrada sister to Onychophora plus Arthropoda (the Lobopodia hypothesis). Our results illustrate the power of CAT-PMSF to model across-site compositionally heterogeneous datasets in the maximum likelihood framework and clarify the relationships of Tardigrada and the Ecdysozoa.