Nearly 3,000 disease associations from GWAS have been published in the past few years [Hindorff et al., 2010]. Relatively few of these are known to follow specific, nonmultiplicative models. It may be that testing for deviations is not done routinely, although even in studies where such investigations have been carried out, few SNPs have shown convincing evidence of recessive or dominant effects [e.g. Wellcome Trust Case Control Consortium, 2007]. Our simulations have shown that such effects will often be detectable, and therefore it is worth explicitly testing associated loci for deviations. As noted above, real disease effects may not deviate as much as fully recessive and dominant effects, and small deviations from multiplicativity will be relatively hard to detect, and easily disguised with only a slight amount of distortion.
