The most suitable evolution model for the alignment was determined by using ProtTest3 [ 41] as JTT + G + F. The JTT model was used in all subsequent analyses.
More precisely, energy contributions can be added up in a columnwise manner to yield an effective energy model for the alignment as a whole [ 49].
Amino acid sequences were aligned with MAFFT v6.717b (L-INS-i), and ProtTest3 was used to select the appropriate model for the alignment, in this case, the LG+G model.
ModelGenerator [37] was used to deduce the best-suited amino acid substitution model for the alignments.
PartitionFinder [ 85] was used to detect accurate models for the final alignment, using each protein as a partition.
This suggests that the scoring model used for the alignment procedure does not bias the selection of the optimal substitution matrix.
Although the Hidden Markov Model used for the alignment is based on an extensive training set, the alignment did require some manual optimisation.
The likelihood of the phylogenetic tree T and the model Θ for the alignment at each site can be calculated as (1) P (A i | T, Θ ) = ∑ μ ∑ ν δ A ℓ i a ν P (ν | μ, t ℓ, Θ ) f μ P v ℓ (A i | v ℓ = μ, T, Θ ) where v ℓ is the ancestor node connected to a leaf node ℓ with branch length t ℓ, μ and ν denote the type of codon, and f μ is the equilibrium frequency of codon μ.
Models for the different alignments were as follows: (i) Nematode ribosomal sequences: GTR + I + G; (ii) Nematode GH45 DNA sequences: 010234 + I + G + F; (iii) Nematode GH45 aminoacid sequences: WAG + I + G + F; (iv) Fungi ribosomal sequences: TIM3 + G; (v) Fungi GH45 DNA sequences: TVM + I + G; (vi) Fungi GH45 aminoacid sequences: LG + I + G.
The alignments of U1, U2, U4, U5, and U6 were downloaded from Rfam [44] and covariance models for these alignments were built using the INFERNAL-cmbuild programme.
