Identifying signatures of convergent evolution
The simplest approach for detecting of patterns of convergent evolution
is to look for genes that are FST outliers in multiple
species, however this approach may miss some true signals as it is very
stringent (Storey and Tibshirani, 2003; Fraser and Whiting, 2019). As a
more sensitive test, the Null-W approach (Yeaman et al. , 2016)
was used to detect signatures of convergent evolution, by identifying
top candidate genes in one species, and then comparing the
FST scores of orthologs to the top candidate genes to
a null distribution of randomly chosen genes from the genome. This was
done using a standard set of 10,000 randomly chosen control SNPs and
comparing both the orthologs and the null distribution genes to the
control SNPs using Wilcoxon ranked sum test W-scores (Wilcoxon, 1945;
for more details see Reeve, 2019 or Supp. Mat. of Yeaman et al., 2016).
These W-scores were normalized into Z-scores using a formula from
Whitlock and Schluter (2009, p. 342), and empirical P-values for the
orthologs were calculated based on their position in the null
distribution using the empPvals function of the qvalue R
package (Storey et al. , 2015). Empirical P-values were corrected
to reduce false discoveries using a Bonferroni correction. Any gene
pairs that remained significant were considered signatures of convergent
evolution.