Consensus Feral Sweeps and Overlaps
For the feral populations, we considered that if a selective sweep was detected by at least two selective sweep finding methods it represented stronger evidence of a true selective sweep. Note that this may sometimes not be the case, as any given detection technique may be better tuned to detecting a specific sweep type that is ‘missed’ by other methods. Nonetheless, we chose to minimalize false positives by focusing on consensus selective sweeps (though individual selective sweep types are also presented in the Supplementary Tables). In the case of the Bermudian population, 59 regions were detected in such a manner (Supplementary Table 8), whilst in the case of the Kauai population 106 selective sweep regions were detected (Supplementary Table 9). When we check for overlaps between these two sets of selective sweep regions, three are detected in both population samples (Table 1), with this being a significant enrichment over the null distribution (permutation test, p < 0.001). Of these, one selective sweep also overlapped with the domestication sweeps identified by Qanbari et al (Qanbari, Rubin et al. 2019)(see Table 1). An overview figure of the locations of selective sweeps identified by the different methods is presented in Supplementary Figure 3.

Selective sweeps origin identification.

To identify the origin of the three shared Hawaiian and Bermudian feral selective sweeps (i.e. are they domesticated haplotypes that have become fixed vs. wild Red Junglefowl haplotypes) Chromopainter software was used (Lawson, Hellenthal et al. 2012). This takes the feral genomes and ‘paints’ on the domesticated and wild haplotypes to ascertain which are most likely to be the donor population of a given sweep haplotype. Two separate layer and broiler populations were used as donors, as well as two Red Junglefowl populations (from Thailand and India, respectively). These were then used as donors for the Bermuda and Hawaii populations, whilst an additional run with Bermuda was also performed, also including Hawaii as an additional donor. In the case of the Bermudian feral selective sweeps, domesticated haplotypes were the most likely donor populations, with broilers and layers donating similar amounts for two of the shared selective sweeps (chr1@17.5Mb, chr1@32.8Mb). In the case of the third sweep (chr2@78.3Mb), the broiler donors were the strongest donor for the Bermudian population. A similar pattern was also seen for the Hawaiian population, though in this case the broiler population was the largest donor for each selective sweep (see Figure 2 A-D). The degree of donorship from Red Junglefowl alleles was broadly similar for the two feral populations.
By further breaking down the domesticated populations into the individual donor sub-populations, we once again see a similar pattern, but also that it is often only one of the two candidate donor sub-populations that contributes the most to a particular sweep. In particular, the layer donors are actually more prevalent when sub-populations are used, with the brown layer donorship high for Hawaii at chr1@17.5Mb, the white layer a high donor for Hawaii at chr2@78.3Mb, and the brown layer a high donor for Bermuda at chr1@32.8Mb (see figure 2B). When focusing on the Bermuda population, with the addition of Hawaii as a donor, we can see a strong overlap between the two feral populations, with Hawaii the largest donor for the sweep at chr2@78.3 and the joint largest for the sweep at chr1@17.5Mb (see figure 2C, D).
Of the three shared sweeps, one also overlapped with a domestication sweep (chr1@32.8Mb). In this case, the domesticated donor is much more evident than either the Hawaiian or Red Junglefowl donors.

Gene annotation and Function

The 59 selective sweeps regions detected by both methods for the Bermudian population contain a total of 61 genes when using the Ensembl genome annotation browser, though a large percentage are long non-coding RNA and other gene-free regions (see Supplementary Table 8 and below). A total of 9 fully annotated genes were identified in selective sweeps (ADCY1, CALU, CRAMP1, DRD3, MYST/Esa1-associated, PTPRB, TAFA5, TSNARE1, ZC3H12A ). These genes are involved in anxiety, schizophrenia, depression and related behaviours (ADCY1, DRD3, PTRB, TAFA5, tSNARE1 ) bone remodelling (MYST/Esa1-assctd ), eye development/ vision (PTRB ), the immune system (ZCH3H12a ) and metabolism (CALU ). There was no enrichment for these types of gene functions, however. Of the 61 genes, only 21 were able to be assessed in the PANTHER over-representation analysis (using complete GO processes setting), with the above mentioned nine genes plus a further 12 unannotated genes still included. The only enrichment was for unclassified processes/ genes in this set (p<0.0001). Of the 61 genes, 40 were long non-coding RNAs (lncRNA). Similar results were found with the Hawaii consensus sweep regions – in this case 171 genes were present in these sweeps, with 131 being useable by Gene ontology software PANTHER (Protein Analysis Through Evolutionary Relationships) (Mi, Muruganujan et al. 2012). The only enriched GO process was once again unclassified (p<0.0001), whilst 25 of the genes were lncRNAs, 10 were micro, miscellaneous, or sno RNAs, 25 were unknown, and the remainder protein coding (see Supplementary Table 9). Gene function was very diverse in the case of the Hawaiian gene set, though of note is the presence of SEMA3A(chr1@9.4Mb), which was one of the three most significant improvement/ domestication-related genes detected by Rubin et al (Rubin, Zody et al. 2010). Similarly, the genetSNARE1 was also found in a selective sweep identified by Rubin et al. 2010.