Do genomic datasets resolve the correct relationship among the placental, marsupial and monotreme lineages?

Did the mammal radiation arise through initial divergence of prototherians from a common ancestor of metatherians and eutherians, the Theria hypothesis, or of eutherians from a common ancestor of metatherians and prototherians, the Marsupionta hypothesis? Molecular phylogenetic analyses of point substitutions applied to this problem have been contradictory – mtDNA-encoded sequences supported Marsupionta, nuclear-encoded sequences and RY (purine–pyrimidine)-recoded mtDNA supported Theria. The consistency property of maximum likelihood guarantees convergence on the true tree only with longer alignments. Results from analyses of genome datasets should therefore be impervious to choice of outgroup. We assessed whether important hypotheses concerning mammal evolution, including Theria/Marsupionta and the branching order of rodents, carnivorans and primates, are resolved by phylogenetic analyses using ~2.3 megabases of protein-coding sequence from genome projects. In each case, only two tree topologies were being compared and thus inconsistency in resolved topologies can only derive from flawed models of sequence divergence. The results from all substitution models strongly supported Theria. For the eutherian lineages, all models were sensitive to the outgroup. We argue that phylogenetic inference from point substitutions will remain unreliable until substitution models that better match biological mechanisms of sequence divergence have been developed.