Tests for admixture

To evaluate the presence of admixture between the parental species and C. bursa-pastoris, we calculated the ABBA-BABA based statistics, D, an estimate of departure from incomplete lineage sorting, and f, an estimate of admixture proportion [41, 42]. These statistics and their significance, which was estimated with a 1Mb block jackknife method, were calculated from population allele frequencies with scripts from Martin et al. [89]. We also used HAPMIX [90] to infer haplotype blocks of admixture with the diploid C. grandiflora, C. rubella, and C. orientalis into the three populations of C. bursa-pastoris for each phased subgenome. We removed sites with more than 20% missing data for each population. The remaining missing data was imputed for the parental populations used in each analysis. As this method determines the probability of ancestry from a diploid progenitor population relative to a non-admixed C. bursa-pastoris subgenome population, we defined regions of the subgenomes as putatively admixed if the probability of ancestry from the progenitor diploid was greater than 50%.