Chunk #21 — Using Results from the GWAS of the International Schizophrenia Consortium as an Example, Are the Results of Dickson et al. Supported by Empirical Observation?
substantial excess predicted by a rare variants-only model. Using exactly the same coalescent simulation models and methods as Dickson et al., we repeated the polygenic analysis presented by the ISC. For comparability with the ISC analyses, we sampled variants from the full simulated set to obtain a MAF distribution similar to that observed in the ISC GWAS, and we further restricted analysis to an LD pruned set (no pairwise LD r 2>0.25). Using the same discovery/sample framework described in [12], we stratified variants into quintiles according to the frequency of the risk-increasing allele. In contrast to the results for the observed ISC data (Figure 4a, following Figure 4a of [12]), the results from simulations under Dickson et al.'s model (Figure 4b) show a marked skewing towards the lower quintiles, indicating that lower frequency SNPs on the GWAS platforms do a better job at tagging rare variants than more common SNPs. Therefore, the empirical results from the ISC GWAS are not consistent with the model presented by Dickson et al. being a general explanation of common variant association.
