Under evolutionary theory we expect a genetic architecture of many more rare than common variants and a negative correlation between effect size and MAF [13],[14]. Lessons learned from Mendelian disorders lead us to expect that genes involved in the genetic architecture of disease will harbour many causal variants [15]. In association studies it is well recognised that genotyped SNPs tag other, unobserved, variants in the genome and that when a SNP is identified as associated it is unlikely that the SNP itself is the causal variant. Whether synthetic associations with rare causal variants represent a significant proportion of associations detected in GWAS depends on the true, but mostly unknown, genetic architecture.
