of the WTCCC data where cases [11] and controls [8],[12] have been genotyped for the three low-frequency coding mutations, testing for association while conditioning on carrier status of one of these mutations completely ablates the signal at the common SNPs (minimum unconditional P-value = 1.2×10−6, conditional P-value = 0.52) (Figure 2). Thus, NOD2 fulfills two important predictions of the synthetic association model: a cluster of low-frequency, high-effect variants can create a GWAS signal, and that signal vanishes when the causal alleles are taken into account.
