Drug Target Mendelian Randomization

A study design that restricts genetic instruments to variants within or near the gene encoding the drug target (e.g., a protein or transcript). This minimizes horizontal pleiotropy by leveraging the direct biological relationship between a gene and its product. For example, using PCSK9 gene variants to predict the effect of PCSK9 inhibitors on LDL cholesterol. Cis-MR is the foundational design for drug target validation because it mimics the specificity of a pharmaceutical agent.

A statistical method that assesses whether two traits—such as a gene expression level and a disease outcome—share the same causal genetic variant at a genomic locus. In drug target MR, colocalization distinguishes a shared causal mechanism from mere genetic correlation due to linkage disequilibrium. A high posterior probability of colocalization (e.g., H4 > 0.8) strengthens the evidence that modulating the target will affect the disease.

Genomic loci where genetic variants are statistically associated with the expression level of one or more genes. eQTLs serve as instrumental variables in drug target MR by proxying the effect of modulating gene expression. Tissue-specific eQTL datasets (e.g., GTEx for 49 tissues) are critical because a drug target's causal effect may only manifest in the disease-relevant tissue.

A violation of the MR exclusion restriction where a genetic variant influences the outcome through pathways independent of the drug target. In drug target MR, this can occur if a variant affects multiple genes or has off-target biological effects. Methods like MR-Egger regression and MR-PRESSO detect and correct for pleiotropy, but cis-MR designs inherently reduce this risk by using variants physically linked to the target gene.

A hypothesis-free approach that systematically tests the causal effect of a drug target proxy on hundreds or thousands of disease outcomes simultaneously. Phe-MR is used in drug development to predict both on-target efficacy and potential adverse effects before clinical trials. For instance, genetically proxying IL-6 receptor inhibition against the full phenome can reveal both cardioprotective benefits and infection risks.

A design where genetic variant-exposure associations and genetic variant-outcome associations are estimated from two independent, non-overlapping study populations. In drug target MR, this often means using eQTL data from one cohort (e.g., GTEx) and disease GWAS summary statistics from another (e.g., FinnGen or UK Biobank). This separation prevents winner's curse bias and enables leveraging large, publicly available summary-level data.