Data analysis

Allele frequencies and genetic diversity measures were calculated using VCFtools and TASSEL, pairwise F_ST values were calculated using VCFtools, genetic distance and kinship matrixes were computed using TASSEL and visualised with the R gplots package [40], and principal components analysis (PCA) was performed with TASSEL. An unrooted neighbour-joining (NJ) tree was produced in TASSEL. Population structure was examined with the model-based clustering algorithm STRUCTURE 2.3.4 [41], with K values between 1 and 12, 20,000 burnins, 40,000 MCMC repetitions and ten independent runs for each value of K. Q-matrixes were displayed on geographical maps using ArcMap v.10 of ArcGIS [42]. Graphs were plotted in Excel and histograms constructed with XLSTAT 2018.

A genome-wide association study (GWAS) was performed to identify SNPs associated with the domesticated phenotype, using a mixed linear model with the first five components of a PCA as input. A kinship matrix (Q+K model) was computed using TASSEL, and a Bonferroni correction with α = 0.01 was used to highlight significant marker-trait associations [43]. Manhattan plots were obtained with the R package qqman [44].