No. 1022

Title:

Quantitative Assessments of Genome-Wide Indels Support Atlantogenata
at the Root of Placental Mammals

Author(s):

Waddell, Peter J (ISM, University of Tokyo);
Umehara, Satoshi (University of Tokyo);
Griche, Karim-Cyril (C/- University of Tokyo);
Kishino, Hirohisa (University of Tokyo)

Key words:

Indel phylogenetics; Atlantogenata; Boreotheria; Supraprimates; Laurasiatheria; Eutheria; Placentalia; phylogenomics; ancestral polymorphism.

Abstract:

Recently sequenced genomes of mammals and other vertebrates potentially contain many rarely occurring indels. These should provide an abundant source of phylogenetic characters with which to test difficult to resolve parts of the tree of placental (eutherian) mammals. One of the hardest and most important parts of the tree is the location of the root. While the hypothesis Boreotheria (also called Boreoeutheria) is consistent with LINE insertion data, it is unclear which of the three hypotheses, Epitheria (Xenarthra diverging first), Atlantogenata (Xenarthra with Afrotheria) or Exafricoplacentalia (Afrotheria first) is correct. Here we test these hypotheses using a log likelihood ratio test and highly conserved well-aligned indels that are five base pairs or larger in size. Finding a data configuration of 51 for Boreotheria with just 2 and 1 supporting alternatives, and finding 4 15 3 for Epitheria, Atlantogenata and Exafricoplacentalia, respectively, our tests reject the alternatives to Boreotheria (p << 0.05) and Atlantogenata (p < 0.05), so the root of placentals appears to fall between these two main lineages of placental mammals. Exploratory analyses of large collections of indels reveal some important possibilities. One of these is that the duration of the two main lineages of placentals may have been very brief, possibly ~ 1 million years or so. Another is that indels, which are located some distance from other indels, indeed do seem to show a much lower homoplasy (C.I. e 0.8) than sequence data (C.I. ~ 0.5-0.6). On a more cautionary note, there are also indications of unexpected biases in the homology and alignment of the genomic data. Given these caveats, it is important that maximally independent types of low-homoplasy molecular data corroborate all major clades of mammals. Estimating the ancestral population sizes of mammals at such an ancient age is shown to be particularly difficult, but potentially soluable. Taking account of biases in estimating the proportion of characters involved in ancestral sorting and in the lengths of deep internal edges, are essential. The resolved root of placentals is important to reconstructions of ancestral characters and genomes, and should yield important insights in these areas.