Skip to content

cis-Mendelian Randomization

Summary

cis-Mendelian randomization (cis-MR) is an epidemiological method that uses genetic variants near a gene of interest (typically within $\pm 200$ to $\pm 250$ kb) as instrumental variables to estimate the causal effect of perturbing a specific protein or molecular pathway on disease risk. By restricting instruments to the cis-region of the target gene, this approach minimizes horizontal pleiotropy compared to genome-wide Mendelian Randomization, making it a powerful tool for drug target evaluation and validation.

Core Principles

In drug development, cis-MR acts as a genetic proxy for clinical trials of targeted therapeutics. - Instrument Selection: Variants are selected from the cis-regulatory or coding regions of the target gene (e.g., HMGCR, LDLR, PCSK9, BCKDK). - Assumptions: Like standard Mendelian Randomization, cis-MR relies on the relevance, independence, and exclusion restriction assumptions. By focusing only on local variants, it is highly likely that any effect on downstream outcomes is mediated directly through the target gene product, satisfying the exclusion restriction. - Methods: Methods include inverse-variance weighted MR (IVW-MR), MRLocus, MR-link-2, and MR-PCA.

Applications

In the study by Tambets et al. (2026), cis-MR was used to evaluate drug targets. For instance, the authors estimated the effect of lowering LDL cholesterol via HMGCR, LDLR, and PCSK9 perturbations on type 2 diabetes (T2D) risk. They also investigated whether lowering plasma branched-chain amino acids (BCAAs) by inhibiting the BCKDK kinase reduces T2D or coronary artery disease (CAD) risk using Branched-Chain Amino Acid (BCAA) Catabolism pathway variants. The analysis showed a null effect, indicating that targeting the BCAA catabolism pathway is unlikely to reduce T2D risk.

Study Design and Sensitivity Analysis

Robust cis-MR requires a biologically justified gene region and molecular exposure, explicit validation of variant-exposure associations, and sensitivity analyses tailored to the locus. Correlated non-lead variants can materially improve instrument strength: across haptoglobin and 15 other protein regions, LD pruning, COJO, SuSiE and PCA recovered more cis-genetic variance than a lead-variant-only instrument, although the lead-only estimate remains a useful numerical-stability check.[3]

For two-sample summary data, analysts must account for linkage disequilibrium rather than treating local variants as independent. Appropriate approaches include generalized inverse-variance weighting, principal-component reduction and Bayesian or fine-mapping-based selection; colocalization is complementary because an MR association can otherwise reflect distinct causal variants in LD.[4][5]

  • Damrauer Lab — applies cis-MR and related genetic methods to therapeutic target validation and cardiovascular disease.

Citations