mean frequency of the most associated genotyped SNP allele was 0.13 for one rare variant and <0.3 for up to 18 rare variants with a trend that the frequency of the most associated allele increased with more rare variants in the region. This is consistent with a slight trend towards the observed lower R 2 between composite value of the k causal variants and the common SNP allele for increasing k, implying a lower power of detection of synthetic associations with higher RAF. Histograms of the frequencies of the allele coupled with the most number of risk variants of SNP that generates the highest R 2 for k = 1, 9, and 18 (Figure 2d–f); these distributions are quite different from that observed from GWAS results (Figure 2a). Of course, if a region harbours 9 rare variants of frequency 0.01, about 20% of the population are expected to carry at least one of them. If the true genetic architecture of a region was one of many rare variants, then the effect sizes necessarily attributable to each of them must be small, otherwise the region would have been unambiguously identified in linkage studies and would explain a large proportion of genetic
