When all human RefSeq mRNA transcripts (n = 25,556) were aligned to their putative orthologs, the result was a gapped alignment of 36 Mb. However, orthologs for most genes could not be found for all 14 taxa. Therefore, the analyses included only those loci for which our method could assign orthologous sequences for all 14 taxa. This reduced data set consisted of a multiple sequence alignment of 1,698 protein-coding loci with an alignment length of 1,443,825 bp, including insertions and deletions. The mean composition of nucleotide bases in this alignment was as follows: T = 22.8%, C = 23.8%, A = 27.8%, and G = 25.6%. A table showing the nucleotide base composition at each codon position for each taxon as well as the number of nucleotides analyzed is included in supporting information (SI) Table 3. Notably, for each base at each of the three codon positions, there are at most only very small compositional differences among the 14 taxa. A summary spreadsheet of the putative orthologs used in the main analyses (SI Table 4), as well as the accompanying alignment files are available as SI.
Fig. 2 depicts the optimal phylogenetic branching pattern among the taxa whether a Bayesian, ML, or MP phylogenetic tree reconstruction method is used with the coding nucleotide sequence data or the parsimony amino acid data. The branch lengths in Fig. 2 were obtained by ML analysis. Fig. 3 depicts the optimal NJ trees. The MP nucleotide tree has a length of 267,158 steps, and the ML scores for the tree topology are -lnL 7,238,023.807 (PAUP*) and -lnL 7,240,210.130 (PhyML). The branch support values for each of the clades are 100% (MP and PhyML bootstrap percent) or 1.0 (Bayesian posterior probability) at all nodes in the tree with the single exception of the Atlantogenata clade, which had a parsimony bootstrap value of 95% in the nucleotide analysis. The relatively short branch lengths seen in the catarrhine and, more specifically, the ape clade also provide further evidence for the hominid slowdown hypothesis (20, 21).
The NJ tree confirms the presence of the Atlantogenata clade with bootstrap support of 100%, although it differs from the other three methods by not supporting Euarchontoglires or Glires as monophyletic, indicating rodents as the first branching placental clade and depicting Laurasiatheria as the sister group to primates (Fig. 3). Parsimony and NJ results from the translated amino acid sequences both depict a basal split between Atlantogenata and Boreoeutheria. Whereas the parsimony amino acid and nucleotide sequence data results show an identical topology, the NJ amino acid tree depicts a boreoeutherian clade not detected by using nucleotide sequence data and indicates Lagomorpha is sister to a clade that includes the remaining Boreoeutheria (Fig. 3). Notably, with nucleotide sequences, when only first and second codon positions (positions less likely to be saturated by superimposed mutations) are retained, the NJ tree again depicts the basal Placentalia divergence to be between Atlantogenata and Boreoeutheria.
We conducted parsimony and likelihood topology tests on the phylogenetic trees constructed by using the coding nucleotide sequence data (Tables 1 and 2). The likelihood tests clearly indicate the tree depicted in Figs. 1C and 2 is the optimal tree (P Table 1). The trees depicted in Fig. 1 A, B, and D are rejected as being suboptimal. Results of parsimony tests also reject the topologies depicted in Fig. 1 A and D in favor of Fig. 1C (Table 2); however, the topology in Fig. 1B cannot be rejected (P = 0.0897; Templeton test). The parsimony scores from these two topologies differ by only 73 steps.
Answers to all your Biology Questions
