Fig. 1.

Diagram showing the presumed relations between each variable. A directed arrow indicates the possibility of a causal effect. For instance, the index genetic instrument represents nearby SNPs with a possible effect on the nearby gene (analogous to cis-eQTLs). A double arrow means the possibility of a causal effect in either direction. The index gene, for example, could have a causal effect on the target gene, or vice versa. We aim to assess the presence of a causal effect of the index gene on the target gene using genetic instruments (GIs) that are free of non-genetic confounding. To do this, we must block the back-door path from the index GI through the GIs of nearby genes to the target gene. This back-door path represents linkage disequilibrium and local pleiotropy and is precluded by correcting for the GIs of nearby genes. Correction for observed gene expression (either of the index gene or of nearby genes) does not block this back-door path, but instead possibly leads to a collider bias, falsely introducing a correlation between the index GI and the target gene