Figure 3

Analyses of indicators that could lead to phylogenetic artifacts in RPB1 sequences. The tree at the base is the same as shown in Figure 2, but without ML branch lengths. The CTD-clade is shaded in light blue. From bottom to top, the following are provided for each sequence. 1) CTD: "+" above a sequence indicates that it contains a C-terminal sequence that is consistent with known requirements for CTD function [36, 86, 87]. 2) aa χ²: results of ML analysis of deviation from mean amino acid composition for each sequence. A "*" indicates that the sequence deviates significantly (P < 0.05) from average composition. "" denotes a χ² P value > 0.9, indicating that the sequence deviates little from average composition. 3) Autapomorphies: Blue bars show the number of unique substitutions at sites under strong stabilizing selection across eukaryotic diversity. Numbers tend to be suppressed in taxa with multiple representatives, because synapomorphies uniting the group are not scored. For example, all kinetoplastids share a number of unique substitutions at otherwise invariable sites. If only one kinetoplastid were to be included, the number of autapomorphies would be comparable to Tv and Eh (data not shown, but see figure 5 for examples from animal, fungal and plant clades). 4) Random Sequence Attraction: Bars show the number of equally parsimonious trees on which each of 100 randomly generated sequences attached to specific RPB1 sequences in parsimony analyses. Red indicates a significant attraction to random sequences (in greater than 5% of parsimony replicates). Because many replicates produced more than one equally parsimonious tree, numbers do not add up to 100. A bar extending across multiple taxa indicates that the random sequence attached to the internode supporting that group (for example, the long branch leading to the three kinetoplastids). RPB1 sequences that were significant poles of attraction were removed from the alignment and the analysis repeated. Four separate analyses were performed, each with 100 randomly generated sequences. The lowest graph shows results using the entire data set, with decreasing numbers of sequences in the graphs above. Sequences shaded out in gray were removed based on significant "long-branch attraction" in the analysis shown immediately below.