Genomic response to selection:
We investigated if the phenotype-associated SNPs identified in our
analyses showed signals of a response to selection and if those signals
were replicated across ecomorph pairs. To test this, for each ecomorph
pair we compared FST and DXY values for
phenotype-associated SNPs to a random background subset of SNPs. This
random subset was 100 SNPs randomly selected from the whole dataset and
the mean FST and DXY values for those
SNPs were calculated. This was repeated 10,000 times and the means for
FST and DXY were taken across all
permutations. These permuted values were then compared to the empirical
mean FST and DXY values for the
phenotype-associated SNPs using the t.test function in R.
Analyses of recombination rate
variation:
To test the effect of the recombination landscape on phenotype-genotype
association, we first estimated recombination rates using the published
Arctic charr linkage map (N=3,636) using MareyMap v1.3.6 . RAD
loci from the linkage map were aligned to the Salvelinus sp.
reference genome with Bowtie2 using the -very-sensitive setting.
Loci were kept if they uniquely mapped to one location, mapped to the
same chromosome as all other loci on their linkage group, and followed
the orientation of the linkage map (i.e., not reversed). The filtered
dataset was used to estimate the recombination rate across each
chromosome using a spline algorithm. Spar values were varied for each
chromosome from 0.5 to 0.9, depending on chromosome size, to best fit
the data . Subsequently, WindowScanR v0.1 (available at:
https://github.com/tavareshugo/WindowScanR) was used to summarize
recombination rate values in 1 MB windows along the genome. All SNPs
were assigned to these windows using BEDtools . A random subset of
100 SNPs was then selected and the mean recombination rate for those
SNPs was calculated based on their windows. This was repeated 10,000
times to generate a background mean recombination rate, which was then
compared to the mean recombination rate of the phenotype-associated SNPs
(based on their windows) using a t-test.