Figure 1.

Causal diagram for Mendelian randomization. The goal is to use genotypes as instrumental variables to infer the causal effect of an exposure on a disease outcome. One critical assumption is that there is no direct effect from the genotype to the disease (no pleiotropy).