Gene ontology and species-specific adaptations
To gain insight into the ecological and biological functions of putative adaptive loci, we identified candidate loci found within genes and the Gene Ontology (GO) terms associated with such genes (Primmer et al. 2013). We used the Ensembl variant effect predictor (VEP) to perform annotation of the candidate loci (McLaren et al. 2016) and the software SNP2GO (Szkiba et al. 2014) to identify cellular component, biological process, and molecular function GO terms associated with the candidate loci using an FDR of 0.05, and following the annotations of the thirteen-lined ground squirrel genome. We tested for enrichment considering all the candidates identified by the four analyses combined, as well as for each analysis individually. To evaluate the impact of the different method assumptions on GO term enrichment, we further tested for enrichment considering the identified candidate loci divided in several categories: population structure outlier approach (pcadapt ), representing those candidates uniquely identified for the pcadapt analysis; genotype environment associations (‘GEA’), considering the candidates uniquely identified for the LFMM, RDA and pRDA analyses; candidates identified including population structure (‘POP’), considering those loci uniquely identified for the pcadapt and RDA analyses; and candidates identified when excluding population structure (‘noPOP’), considering those loci uniquely identified for the LFMM and pRDA analyses. For all three datasets (NIDGS, SIDGS, and the combined IDGS), we further considered the candidate SNPs resulting in non-synonymous substitutions and used the online databases Ensembl and UniProt to identify the genes and proteins involved (Bateman 2019; Yates et al. 2020).