Cis-Mendelian Randomization

Instruments are selected from a strict genomic window around the target gene, typically within 1 megabase (Mb) of the transcription start site. This cis-regulatory region ensures variants directly influence the gene product's expression or function. Common sources include expression quantitative trait loci (eQTLs) and protein quantitative trait loci (pQTLs). By restricting to this local window, the design leverages the direct molecular link between the genetic variant and the encoded exposure, maximizing the biological validity of the instrument.

The primary advantage of cis-MR over standard MR is the drastic reduction of horizontal pleiotropy, where a variant affects the outcome through pathways independent of the exposure. Because the instrument is physically linked to the gene encoding the exposure, it is far less likely to influence a complex outcome through alternative biological routes. This strengthens the exclusion restriction assumption, a core requirement for valid instrumental variable analysis, making causal estimates more robust than those from trans-acting variants scattered across the genome.

Cis-MR is a cornerstone of drug target Mendelian randomization, directly simulating the effect of a pharmaceutical intervention. A genetic variant that alters the function or concentration of a protein mimics a drug that agonizes or antagonizes that same target. Key applications include:

• Predicting on-target efficacy before clinical trials
• Identifying potential adverse side effects through phenome-wide outcome analysis
• Prioritizing drug targets with human genetic evidence, which are twice as likely to succeed in clinical development

Cis-MR is frequently paired with colocalization analysis to distinguish a true causal relationship from mere genetic linkage. Colocalization tests whether the same single causal variant drives both the exposure (e.g., protein level) and the outcome (e.g., disease risk). A high posterior probability for a shared causal variant (e.g., PP.H4 > 0.8) provides strong evidence that the genetic association is not due to linkage disequilibrium confounding, thereby reinforcing the validity of the cis-MR causal estimate.

Cis-MR enables highly granular, tissue-specific causal inference by utilizing eQTL data from specific cell types or tissues. For example, an analysis can isolate the causal effect of a gene's expression in the liver versus adipose tissue on metabolic disease. This specificity is achieved by using instruments derived from tissue-specific gene expression datasets like GTEx. This allows researchers to pinpoint the exact anatomical context of a causal mechanism, a level of resolution unattainable with standard, non-cis MR designs.

A critical challenge in cis-MR is the potential for weak instrument bias. Since the analysis is restricted to a small genomic region, there may be few variants, and they may explain only a small fraction of the exposure's variance. To mitigate this, analysts calculate the F-statistic (with a threshold typically > 10) to ensure instruments are sufficiently strong. Methods like allele score aggregation or using tissue-specific pQTLs with large effect sizes are employed to maximize statistical power while maintaining the biological specificity of the cis design.