FIGURE 2 Workflow of the whole genome re-sequencing-based design of PCR-RFLP markers (GB-RFLP). A. Markers were designed based on genetic differentiation (FST) or genome-wide genotype-phenotype association (GWA) data of a 453-genome dataset using pairwise ingroup–outgroup comparisons. B. Variants were screened for RFLPs with one allele (but not the other) being cut by a restriction enzyme. C. Target variants were filtered based on the presence of additional restriction sites and additional SNPs within the restriction site. Quality control was performed by plotting allele frequencies and haplotype networks. D. Likelihoods that a genotype corresponds to a population (or not) were calculated based on population-specific allele frequencies. Percentage of correct assignments together with false negative and false positive rates were based on bootstrapping of genotypes (informed by empirical allele frequencies in the genomic dataset). E. Primers were designed based on 801 bp sequences (core SNP +/- 400 bp) in a way that the restriction enzyme would generate two fragments with a ~1:2 length ratio. F. PCR conditions were tested and optimized. G. Restriction digest was performed on 8 ingroup samples and 5–8 outgroup samples. Genotypes were determined by visual inspection. These data were used to calculate the number of correct assignments as well as rates of false positives and negatives (as for the bootstrapping dataset in D).