Methods for selecting fixed-effect models for heterogeneous codon evolution, with comments on their application to gene and genome data

Table 1 Fixed-effect models implemented by Yang and Swanson [4].

Model code	Parameters for partitions	Number of Parameters
A	same branch lengths, κ, ω, and π's	b+2+9
B	different but proportional branch lengths, same κ, ω, and π's	b+(g-1)+2+9
C	different but proportional branch lengths, same κ, ω, and different π's	b+(g-1)+2+g×9
D	different but proportional branch lengths, different κ, ω, and same π's	b+(g-1)+g×2+9
E	different but proportional branch lengths, different κ, ω, and π's	b+(g-1)+g×2+g×9
F	different branch lengths, κ, ω, and π's	g×(b+2+9)

The number of parameters is computed under the F3 × 4 method of estimating codon frequencies. b denotes the number of branches in the tree. g denotes the number of site classes. When models employ empirical estimates of each codon frequency (F61 method) the number of model parameters increases by 51 for models with homogenous π's, and by 51 × g for models with heterogeneous π's among partitions.

ISSN: 2730-7182