The search for common variants affecting the incidence of a disease has now become possible without making any prior assumptions as to the nature of the variants involved, through the ability to screen a sufficiently large number of well-spaced SNPs providing almost complete genomic coverage. It should then, in principle, be possible to identify the real disease-associated variant by scanning nearby genes for variants that plausibly satisfy the requirement for having an effect on the disease. Most of the common variants found so far in the recent enormous accumulation of new data on WGAS for a wide range of diseases are, however, associated with ORs of only between about 1.2 and 1.5 (Fig. 2). The main challenge to their identification has been to do large enough studies, with replication, to achieve unequivocal statistical significance. The studies must also take into account (see ref. 25 for an example) small overall effects needing large studies for their detection, the potential confounding effects of hidden population substructure, and multiple comparisons, namely the testing of very large numbers of SNPs, which entails using very stringent significance levels—often down to 10−7—to avoid large numbers of false positives.
