A hierarchy of approaches supports stronger causal inference regarding the role of modifiable exposures on disease outcomes (see Table 1). Ultimately, what is required is a triangulation of evidence using these different approaches, ranging from whole genome methods to more focused analyses, to determine whether the results obtained using these different methods align consistently [34]. First, genetic correlation [35,36] can be used to identify shared genetic influences (e.g., cannabis use and schizophrenia). This approach allows all genotyped common variants to be interrogated, with correlations with modifiable exposures suggestive of possible causality. Second, conventional Mendelian randomization analyses (using single variants or polygenic risk scores) can be used to establish evidence that genetic proxies for a modifiable exposure of interest (e.g., cannabis use) associate with the outcome thought to be influenced by the exposure (e.g., schizophrenia) [33]. Single variant approaches are appropriate when the genetic variants play a known and relatively specific role in the pathway of interest (e.g., ALDH2 and alcohol consumption), but these will capture a smaller proportion of the variance in the exposure than polygenic risk scores. Third, when
