SNP genotype data and association analyses
Nestlings that survived to adulthood (recruited) on Hestmannøy and Træna were genotyped on a high-density 200K SNP array (detailed in Lundregan et al., 2018) with median distances between SNPs shorter than 5,000 bp. SNPs were originally identified from whole-genome re-sequencing of 33 individual house sparrows which were mapped to the house sparrow reference genome (Elgvin et al., 2017). DNA was extracted as described in Hagen et al. (2013), separately from telomere analyses. Data preparation and quality checks were performed using the ‘GenABEL’ package (GenABEL project developers, 2013). We removed SNPs or individuals for which there was more than 5% missing data, the minor allele frequency (MAF) was less than 1%, or pairwise identity-by-state (IBS) was more than 95%. After quality control, the genomic relationship matrix (GRM) was computed based on 180,650 [180,666] autosomal markers in 373 [383] individuals (142 [145] males and 137 [142] females from Hestmannøy and 47 [48] males and 47 [48] females from Træna) with numbers in brackets showing sample sizes when individuals with missing tarsus length measurements are included. We then performed two GWA analyses by fitting LMMs for the variation in TL using the package ‘RepeatABEL’ (Rönnegård et al., 2016): The first model included age-standardized tarsus length as a covariate, and the second model did not. Both models included sex, age, hatch day (mean centered), F , and island identity as fixed effects, and brood identity, year, and the GRM fitted as random effects. Finally, we determined if SNPs significantly associated with TL were within 100 kb of any gene within the annotated house sparrow genome, because this is the distance that linkage disequilibrium decays to background levels in this species (Elgvin et al., 2017; Hagen et al., 2020).