Salmonid Ortholog Sequence Analysis

Putative orthologue sequence sets were collated with Best Reciprocal Blast (BRB) protein matches. For salmonid species the top-two BRB-hits were assigned to putative orthologue groups. Multiple codon sequence alignments were constructed using MAFFT^{62 (link)} and quality trimmed with Guidance in an iterative framework where sequences were re-aligned after identification of poorly aligned codons.
Maximum likelihood (ML) gene trees were calculated by the R-package Phangorn^{63 (link)} using codon alignments, the GTR+G+I model, and 100 bootstrap replicates. Branch specific GTR+G+I substitution rates were estimated functions from the R-package ape⁶⁴ , while branch specific synonymous (dS) and non-synonymous (dN) substitution rates were estimated with non-negative least squares regression in the Phangorn R package^{63 (link)} using pairwise dN and dS distance matrixes from codeml^{65 (link)} and the ML gene tree topologies as input.
Branch-site specific test for positive selection was carried out by a likelihood-ratio test on the ML-likelihood estimates for sequence evolution under different models in codeml. The smallest likelihood estimate from four omega starting values (0.5, 1, 1.5, and 2) was used in the likelihood ratio test (LRT). False discovery rate adjustments of p-values were done with the p.adjust function in R.