Primary Submission Category: Applications in Health and Biology
Identifying causal proteomic targets for cardiovascular disease via robust Mendelian randomisation.
Authors: Christopher Aldous Oldnall, Sjoerd Viktor Beentjes, Ava Khamseh,
Presenting Author: Christopher Aldous Oldnall*
Mendelian randomisation (MR), a biological application of the instrumental variable (IV) framework, is a powerful tool for identifying causal relationships in biomedical research. By using genetic variants such as single nucleotide polymorphisms (SNPs), MR bridges genome-wide association studies (GWAS) and translational medicine, enabling the discovery of causal proteins in diseases like cardiovascular disease (CVD). However, pleiotropy—violations of the exclusion restriction criterion—remains a challenge. While traditional methods like Generalised Summary-data-based Mendelian Randomisation (GSMR) attempt to address pleiotropy by excluding invalid instruments, they rely on assumptions about instrument validity. Newer novel pleiotropically robust estimators leverage potentially invalid instruments against valid ones, offering a systematic, assumption-light approach to handling pleiotropy.
Using proteomic data from the UK Biobank, we identify causal proteins across multiple CVD risk traits, including hypertension and diabetes. We evaluate the overlap and differences between proteins identified by traditional MR methods and the pleiotropy robust estimators, critically assessing their consistency and robustness. This work highlights the importance of pleiotropically robust MR in advancing causal discovery, providing a reliable framework for translating genetic insights into clinically meaningful applications.