METHODS

Field sampling of Bermuda’s feral chickens (G. Gallus )

Field samples of fresh blood were collected on Whatman Filter cards from chickens that were culled by the Bermudian government in 2015. Individuals were chosen for sequencing (see below) in order to include 1) a range of sublocalities spanning the small archipelago’s main islands and 2) a range of microhabitats including developed and undeveloped areas. In total, n=15 males and n=6 females were selected for sequencing. Phenotypic data was not available for the individuals that were used for Whole Genome Sequencing, and was instead collected from individuals sampled in the field in Bermuda during 2018 (table 2). This consisted of body weight and comb weight measures taken post-mortem (collected from 95 Bermudian feral individuals, 55 males and 40 females), as well as comb morphology recordings (presence of pea comb, rose comb and duplex comb morphs), and the number of different leg colorations present in the population (yellow and/ or grey leg colours), taken from 134 Bermudian feral individuals. In addition, comb and body weight, and morphological measures were also taken from 36 male and 36 female Kauai feral birds. For the Kauai birds, 25 birds were blood sampled and sequenced, with samples once again taken from a variety of sublocalities.

Publicly available sequence data for domestic and wild G. gallus

Whole-genome sequence data for domestic and Red Junglefowl contrasts were obtained by from data published by Qanbari et al (Qanbari, Rubin et al. 2019). We downloaded the reads from the European Nucleotide Archive (ENA, http://ebi.ac.uk/ena), and called variants using the same workflow as for the Bermuda samples. The accession numbers and sample labels used are listed in Supplementary Table 10.

DNA Sampling and whole genome sequencing

We extracted genomic DNA using the DNEasy Blood and Tissue Kit (Qiagen), according to the manufacturer’s protocol. We sequenced 21 Bermuda chicken samples at 30X coverage on the Illumina HiSeq X platform at SciLifeLab, Stockholm, and 25 Kauai chicken samples on an Illumina novoseq X6000. In the Bermuda samples, we trimmed residual adapter sequences with Trimmomatic version 0.36 (Bolger, Lohse, & Usadel, 2014), aligned the sequences to the chicken genome (version Galgal6) using bwa mem version 0.7.17(Li, 2013) and processed the alignments using a workflow inspired by GATK best practices (DePristo et al., 2011; McKenna et al. , 2010; Van der Auwera et al ., 2013), including the use of the functions MergeSamFiles, SortSam, BaseRecalibrator, ApplyBQSR, GenomicsDBImport and GenotypeGVCFs from GATK version 4.3.0.0. Sequencing duplicates were removed with the MarkDuplicates function from picard versions 2.27.5 (https://broadinstitute.github.io/picard/index.html). In the Kauai samples, poly-G tails were trimmed using fastp (Chen, Zhou et al. 2018), after which the workflow mentioned above was followed, but with bwa mem version 0.7.15, GATK version 3.8.1.0 and picard version 1.118. Finally, we filtered the resulting SNPs with the GATK VariantFiltration tool. We retained variants that passed filters QD < 2.0, FS > 60.0, MQ < 40.0, MQRankSum < -12.5, and ReadPosRankSum < -8.0. In a second filtering step using bcftools version 1.14 (Danecek, Bonfield et al. 2021), SNPs with up to 40% of missingness and a minor allele frequency larger than 5% per population were kept. All SNPs for the selective sweep analyses were phased using 50 iterations in Beagle 5.4 (Browning, Tian et al. 2021).