Inferring selection and allelic frequency shifts
The impossibility to test for expression of candidate genes at this point led us to indirectly explore whether neutral evolution could explain genotype and allele frequency dynamics in the population. Molecular signatures of selection often reflect allele frequency shifts from one generation to another. This was possible to investigate with our dataset because of the timestamps associated with breeding pairs and respective offspring and annual monitoring of individual presence in nesting boxes. We first inspected whether genotype frequencies deviated from Hardy-Weinberg equilibrium across years, by calculating year-to-year observed and expected frequencies following the Hardy-Weinberg equation (Wittke-Thompson, Pluzhnikov, & Cox, 2005), considering only offspring, adults and both. We then compared the pairwise frequencies of both alleles and genotypes of candidate loci with Wright’s F-statistics across years. To understand at what point observed values significantly deviate from neutral expected distributions, we randomly selected 100 pairs of loci for among the RADseq dataset, calculated F-statistics, and estimated 97.5% confidence intervals of the F-statistic distribution. F-statistics calculations were performed in Arlequin v3.5.2.2 (Excoffier & Lischer, 2010).