Bushes in the tree of life

Abstract

A discussion of how homoplasy (the frequency of independently evolved characters) and the spacing of cladogenetic events limit our ability to reconstruct the tree of life using existing phylogenetic methods.

Figure 1. The Shape of a Clade Influences its Resolvability

(A) Early in a clade's history (gray box), the number of cladogenetic events is smaller and the length of stems larger in tree-like (left) relative to bush-like clades (right). (B) In the absence of homoplasy, the number of PICs for a stem is proportional to its time span; many PICs (rectangles) accumulated on the long stem x (left), whereas few PICs accumulated on the short stem y (right). (C) When the stem time span is long, the effect of homoplastic characters (crosses supporting a clade of species A and C and bullets supporting a clade of species B and C) is not sufficient to obscure the true signal (left). In contrast, the same number of homoplastic characters is sufficient to mislead reconstruction of short stems (right), because the number of homoplastic characters shared between species A and C (three crosses in each of the two species) is larger than the number of true PICs (two rectangles).

Figure 2. Four Notable Bushes at Different Temporal Depths of the TOL

(A) The human/chimpanzee/gorilla tree (5–8 million years ago). (B) The elephant/sirenian/hyrax bush (57–65 million years ago). (C) The tetrapod/coelacanth/lungfish bush (370–390 million years ago). (D) The metazoan superbush (>550 million years ago). In each panel, the three alternative topologies for each set of taxa are shown. Below each topology, the percentage and number (in parentheses) of genes, PICs, and RGCs supporting that topology are shown (when available). Numbers of genes supporting each topology in (A), (C), and (D) are based on maximum likelihood analyses; numbers in (B) are based on parsimony. The observed conflicts are not dependent on the optimality criterion used; similar results were obtained by analyses of the data under a variety of widely used optimality criteria (see references below). A fraction of genes in each panel is uninformative: (A), 6 of 98 genes; (B), 9 of 20 nuclear genes; (C), 1 of 44 genes; and (D), 179 of 507 genes. The single-codon indel supporting the tetrapod/lungfish topology [69] could be homoplastic or even the result of lineage sorting (C). Data for each panel are from the following: (A), [24–27]; (B), [3,30,31]; (C), [8,69]; and (D), [9,10].