genomewide-significant in the comparison of Icelanders and Scots)—allele frequency differences between Norwegians and Scots were likewise greater than 15% (Table S1), ruling out an effect specific to Icelanders. We also report frequencies of these SNPs in HapMap populations [27] (Table S1). We note that both TLR and NADSYN1 were previously reported to be significantly differentiated among 12 British subpopulations analyzed by the WTCCC (nominal P-values of 10−12 for TLR and 10−8 for NADSYN1) [28]. The WTCCC study has made an important and valuable contribution to research on natural selection by highlighting the potential utility of large sample sizes from very closely related populations for detecting signals of selection. However, the statistical test employed by those authors only evaluated whether frequency differences between the 12 subpopulations were different from zero, and not whether the amount of differentiation was in excess of what would be expected under neutral genetic drift (as inferred from genome-wide patterns). As an illustration of this distinction, we observed that a total of 3,982 SNPs in our data set had frequency differences between Iceland and Scotland that were different from zero at the nominal P-value threshold of 10−7 used for the corresponding test in the WTCCC study. It
