Rooting the eutherian tree: the power and pitfalls of phylogenomics

Table 1 Comparison of the log-likelihood for the three hypotheses with each model

Concatenate or separate model	Substitution model	Tree	< ln L > (Δ ln L ± SE)	KH	wSH	BP	#p	AIC
Concatenate model	GTR + Γ₈	1	-117.2 ± 31.1	0.000	0.000	0.0
		2	-147.3 ± 29.7	0.000	0.000	0.0
		3	< -4,076,316.3 >			100.0	26	8,152,684.6
	Codon + Γ₄	1	< -3,828,351.7 >			88.1	81	7,656,865.4
		2	-77.8 ± 64.5	0.112	0.185	11.3
		3	-142.7 ± 65.0	0.014	0.026	0.6
	JTT-F + Γ₈	1	< -1,905,933.9 >			51.6	37	3,811,941.8
		2	-84.1 ± 37.4	0.014	0.028	0.2
		3	-1.7 ± 41.9	0.478	0.637	48.2
Separate model (among 2,789 genes)	GTR + Γ₈	1	< -3,963,489.9 >			86.2	72,514	8,072,007.8
		2	-117.4 ± 72.3	0.050	0.092	4.1
		3	-91.4 ± 72.7	0.104	0.174	9.7
	Codon + Γ₄	1	< -3,621,322.1 >			89.6	225,909	7,694,462.2
		2	-128.0 ± 103.2	0.107	0.164	10.4
		3	-527.9 ± 96.3	0.000	0.000	0.0
	JTT-F + Γ₈	1	< -1,799,245.4 >			93.4	103,193	3,804,876.8
		2	-134.9 ± 88.5	0.064	0.112	6.6
		3	-317.6 ± 85.5	0	0.000	0.0

Maximum likelihood (ML) trees varied depending on the substitution model used for the concatenate analysis, whereas the separate model analyses consistently supported tree 1. The log-likelihood of the ML tree is given in angled brackets, and the differences in the log-likelihoods of alternative trees from that of the ML tree ± 1 standard error were estimated using the formula of Kishino and Hasegawa [28]. Numbering of the trees corresponds to that shown in Figure 1. KH and wSH denote P values derived using by the test of Kishino and Hasegawa [28] and the weighted test of Shimodaira and Hasegawa [27], respectively, calculated by the CONSEL program [47]. AIC, the Akaike Information Criterion [29]; #p, number of parameters of the model.

ISSN: 1474-760X