As GWAS of disease outcomes are carried out on increasingly large samples, more loci will be identified, promising to deliver insights into underlying biological mechanisms. However, as we have seen, it will become increasingly important to also consider whether these associations reflect effects of modifiable exposures. This will require the triangulation of evidence across GWAS of disease outcomes and GWAS of behavioural phenotypes to determine the cases in which signals identified for behavioural phenotypes are the same as those identified for disease phenotypes. Unfortunately, this approach is hampered at present by the relative lack of GWAS of behavioural phenotypes—while we have identified a number of variants associated with tobacco and (to a lesser extent) alcohol use, as well as obesity, this is not yet the case for exposures such as cannabis use. Nevertheless, this situation is rapidly changing—for example, there are now several variants that have been shown to be associated with caffeine consumption [26]. It is also worth noting that both the CHRNA5-A3-B4 and ALDH2 loci were initially identified in candidate gene studies.
