The relevance of the simulations conducted by Dickson et al. depends on whether the genetic architecture they assumed is a realistic representation of nature. Empirical observations do not generate the pattern of results generated by their genetic model, which implies that their model applied genome-wide does not reflect the typical architecture of complex disease. If their model of genetic architecture is only representative of a small proportion of the genome and represents only a small proportion of the genetic variance, their results cannot be representative of most GWAS results. A genetic architecture of rare variants only will generate a distribution of associated variants very heavily skewed towards genotyped SNPs with low RAF; this does not agree with observation. Multiple rare variants within short genomic regions could generate associations with alleles of higher frequency including major alleles, but the variance explained by the causal variants would need to be so high that firstly, it should have been detected by linkage studies and, secondly, it would imply very few such loci across the genome for the mechanism to be considered important. Results
