Specific approaches and limitations in (multi)-omic Mendelian randomization

GWAS
causal inference
instrumental variables
mendelian randomization
multi-omics
pleiotropy
  • Purpose: This paper reviews critical methodological challenges in Mendelian Randomization (MR) when applied to complex exposures derived from multi-omics data.
  • Key Challenge: Pleiotropy: The authors advocate for using a biologically motivated strategy for selecting genetic instrumental variables (IVs) to avoid pleiotropy, such as explicitly excluding confounding genetic regions (e.g., the IL6R locus when using IL-6 levels).
  • Data Quality: The review highlights the importance of the measurement technique for the exposure (e.g., avoiding low-resolution 16S sequencing for microbiome or techniques that cannot separate lipid isomers) as data quality issues can invalidate MR assumptions and introduce pleiotropy.
Published

23 January 2026

PubMed: 39147365 DOI: 10.1016/j.jlr.2024.100619 Overview generated by: Gemini 2.5 Flash, 28/11/2025

Introduction to Mendelian Randomization in the Multi-Omics Era

Mendelian Randomization (MR) is a statistical technique that uses genetic variants as instrumental variables (IVs) to assess the causal effect of an exposure (X) on an outcome (Y). This approach minimizes confounding and reverse causation inherent in traditional observational studies. Given the explosion of multi-omics data (proteomics, metabolomics, transcriptomics), MR is increasingly applied, making it crucial to understand the specific limitations and necessary approaches to ensure valid causal inference.

Specific Approaches to Address Key MR Limitations

The core challenge in MR is validating the IV assumptions, particularly the assumption of no pleiotropy (the IV affects the outcome only through the exposure). The authors review strategies to enhance the reliability of MR studies, particularly in the multi-omics context.

1. Consideration of Biological Pathways (Addressing Pleiotropy)

Researchers must have a clear research question and deeply consider the biological relevance of their exposure variables and the selected genetic instruments.

  • Biologically Motivated IV Selection: A strategy focusing on variants in or near the gene encoding the exposure (e.g., a protein or a regulator of that protein) is generally preferred over a broad, genome-wide approach. This helps limit the potential for the IV to affect the outcome through separate pathways.
  • Example: IL-6 Signaling: The effect of inhibiting Interleukin-6 (IL-6) signaling on Cardiovascular Disease (CVD) can be studied using variants in the IL6R gene (encoding the receptor). However, these variants not only lower IL-6 signaling but also increase systemic IL-6 levels. When using a GWAS of IL-6 levels as the exposure, researchers must exclude the IL6R region to avoid spurious results caused by this type of vertical pleiotropy.

2. Consideration of Measurement Techniques (Addressing Data Quality)

The reliability of MR findings is tied directly to the quality and resolution of the data used for the exposure GWAS.

  • Microbiome Data: MR studies using 16S genomic data for the microbiome must be cautious, as this sequencing technique often lacks the depth to reliably quantify bacteria down to the species level, potentially invalidating basic MR assumptions.
  • Lipidomics/Metabolomics Data: GWAS data derived from lipidomic or metabolomic techniques that are unable to separate isomers can introduce biases and increase the risk of pleiotropy.

3. Addressing Weak Instrument Bias (The Relevance Assumption)

The relevance assumption states that the genetic IV must be robustly associated with the exposure. Violation of this assumption leads to weak instrument bias, which pulls the causal estimate toward the null.

  • Assessment: The strength of the instrument is appraised using the F-statistic and the proportion of variance explained (\(r^2\)).
  • Mitigation: To increase power and minimize bias, researchers often combine multiple independent genetic variants (Single Nucleotide Polymorphisms or SNPs) into genetic risk scores or use multi-variant meta-analysis approaches.

Conclusion

The review emphasizes that while MR remains a powerful tool for causal inference, its application in the multi-omics field requires rigorous attention to the biological and technical details of the data and the genetic instruments chosen. Adopting specific, biologically-motivated approaches to IV selection and critically appraising the limitations of measurement techniques are essential steps for generating reliable and interpretable results.