Skip to main content

Table 2 Comparison of alignment and substitution models.

From: Incorporating indel information into phylogeny estimation for rapidly emerging pathogens

Model ln P(Y) κ ln λ ln ε ω f (10,12,18)
HKY -1555.7 7.2(4.6,11.7) -3.3(-4.1,-2.7) -1.0(-1.3,-0.78) - 0.5 0.96(25)
HKY × 3 -1579.8 7.5(4.8,12.2) -2.3(-3.1,-1.6) -2.7(-3.7,-1.9) - 0.5 0.75(3.1)
M0 -1542.7 7.2(4.6,11.8) -2.2(-3.0,-1.6) -2.7(-3.7,-1.9) 1.0(0.86,1.2) 0.46(0.26,0.65) 0.92(12)
  1. We compare the HKY singlet model, the HKY × 3 triplet model, and the M0 codon model, that forbids stop codons. For the first two models we fix independent nucleotide frequencies but for the M0 model we allow codon frequencies to vary. Continuous parameter estimates are presented as a posterior median followed by a 95% Bayesian credible interval if free, and a single value if fixed. The HKY model has a higher marginal probability than the HKY × 3 triplet model, indicating that not all indels start and end between codons. Removing stop codons and allowing codon frequencies to vary freely increases the marginal likelihood of the M0 model substantially. Despite these increases in marginal likelihood, the M0 model does not support the clade (10,12,18) as well as the singlet model.