Concordance of shared and private SNPs in P1 and P2 datasets
P1 and P2 SNPs that are shared with the corresponding P1+P2 dataset have much higher levels of concordance than private SNPs in the P1 and P2 datasets (Figure 5). In addition, these shared SNPs show the expected increase in concordance with increasing sequence depth, similar to P1+P2 SNPs, whereas private SNPs show the unexpected increase in concordance at low sequencing depth previously noted in Figure 3. This phenomenon could be due to enrichment with homeo-SNPs, which are truly heterozygous in all individuals but may appear as homozygous when sequencing depth is limited. One limitation of our reduced representation library preparation protocol is that alleles underlying a locus may have different fragment sizes, leading to amplification bias (Davey et al., 2011). In samples with very low sequencing depth, under-represented reads from the larger fragment may be entirely absent, leading to higher concordance at low depth in the private SNPs.