These data suggest that SNPs are a cost effective and informative replacement for STRPs when used to detect population structure. Based on these results, a set of approximately 100 SNPs with MAFs of 40–50% can resolve population structure. If SNPs with lower MAFs are randomly chosen, then more than 250 SNPs may be required to obtain reliable results. However, these results identify a subset of 20 SNPs that also reliably resolve population structure in this sample. These results suggest that a small "genomic control subset" selected based on allele frequency differences in the two populations could be quite useful. Although the regression indicates that any SNP, on average, shows a 5% allele frequency difference between the two populations, our results show that the SNPs with higher MAF are more useful for the STRUCTURE analyses. Thus, it is better, in general, to choose a SNP with MAF of 0.45 and 0.5 in Caucasians and African Americans than to choose a SNP with MAF of 0 and 0.05.
