As we run larger and larger GWAS, some of the signals that emerge may turn out to reflect the action of modifiable (e.g., environmental or behavioural) exposures, rather than more direct biological effects. At present, what is likely to be required to understand these pathways is a two-step approach in which initial GWAS findings are interrogated further in studies in which detailed phenotype information is available. At present, this is not always possible—for example, a lack of smoking status information in the studies contributing to the recent schizophrenia GWAS means it is not possible to test the possible causal effect of smoking in a stratified analysis. However, as large, richly phenotyped cohort studies (e.g., UK Biobank) emerge, it will become possible to identify modifiable exposures from genetic data and to dissect those pathways within the same cohort. Here, “modifiable” can refer to substance use, but also to factors such as cholesterol or metabolite levels or blood pressure, which are directly influenced by lifestyle choices. A failure to appreciate this point will hamper our ability to translate the results of GWAS
