Figure 1.
Scheme of the species tree used for computer simulations.
θ, present and ancestral population size parameters (here, assumed to be constant). τ, divergence times. x and y, length of internal branches. The MCcoal notation of this species tree corresponds to: “(((sp1_# θ1, sp2_# θ2): τ1 # θ12, sp3): τ2 # θ123, sp4): τ3 # θ1234;”. Both τ and θ are measured in substitutions/site.
Figure 2.
Expected levels of gene tree discordance under different multilocus strategies and different combinations of mammalian evolutionary parameters.
δ, % difference between the tMRCA of the sampled genes and the species split time. mit, mitochondrial. nuc, nuclear. θmit, population size parameter for mitochondrial genes. υmit, substitution rate for mitochondrial genes. The corresponding population size parameter for nuclear genes was 2.5 times smaller. The nuclear substitution rate was tenfold lower in each case.
Figure 3.
Performance of the Bayesian (MB) approach under different multilocus strategies and different combinations of mammalian evolutionary parameters.
Sample size, n = 5. The line with circles shows the average percent error among the 50 replicates. Solid lines indicate the lowest and the highest estimate still within the range IQ±1.5 IQR. Columns show the different diversity parameter combinations and rows correspond to the different multilocus strategies. mit, mitochondrial. nuc, nuclear. θmit, population size parameter for mitochondrial genes. υmit, substitution rate for mitochondrial genes. The corresponding θ for nuclear genes was 2.5 times smaller. The nuclear substitution rate was tenfold lower in each case.
Figure 4.
Performance of the Bayesian (MB) approach under different sample sizes.
θmit = 0.025. υmit = 0.01. The population size parameter for nuclear genes was 2.5 times smaller. The nuclear substitution rate was tenfold lower. The line with circles shows the average percent error across replicates. Solid lines indicate the lowest and the highest estimate still within the range IQ±1.5 IQR. n, number of sequences sampled per locus.
Figure 5.
Performance of the Bayesian (MB) approach without fixing mutation rates and heredity multipliers.
θmit = 0.025. υmit = 0.01. The population size parameter for nuclear genes was 2.5 times smaller. Nuclear markers were simulated with two different average substitution rates: υnuc = 0.001 (A) and υnuc = 0.0005 (B). n = 5. The line with circles shows the average percent error across replicates. Solid lines indicate the lowest and the highest estimate still within the range IQ±1.5 IQR. MB, substitution rates estimated by the Multilocus Bayesian (MB) software.
ML, substitution rates estimated from Maximum Likelihood (ML) individual gene trees.
Table 1.
Results of the *BEAST analysis with different multilocus strategies and species radiations.