be needed to generate the odds ratio detected at the genotyped SNP, which depends on the allele frequencies at the two loci (see Box 3). When the frequency of an associated SNP allele is in the range of 0.2 to 0.5, a detected OR of 1.1 implies that a causal variant of frequency 0.01 must have OR 3 to 6 and a detected OR of 1.3 implies an OR at a causal locus of 4 to 16 (Table 2). As noted by Dickson et al., such effect sizes would be detectable in linkage studies of large pedigrees, or of multiple smaller pedigrees if multiple rare variants occur at the same chromosomal locus and so presumably would have been found already. Therefore, if each GWAS associated SNP represented a synthetic association with a single causal variant then, for some traits, we would have already explained the heritability several times over. But what about synthetic associations caused by multiple rare variants?
