Mapping causal biology: Mendelian randomization in the era of big data

Mendelian Randomization (MR) is a method that utilizes genetic variants as tools to determine causal relationships between exposures and outcomes, thereby reducing the confounding bias and reverse causality inherent in observational studies.Rooted in Mendel's laws of inheritance, MR has undergone rapid development since Katan first introduced the concept in 1986, followed by the formal methodological establishment of Davey Smith and Ebrahim in 2003.This review comprehensively summarizes the theoretical foundations, methodological innovations, and expanding applications of MR. We discuss the three core assumptions underpinning MR—relevance, independence, and exclusion restriction—and examine advanced methodological extensions, including two-sample MR, multivariable MR,bidirectional MR, nonlinear MR, mediation MR, and multi-population MR. We further provide an overview of commonly used R packages, databases, and analytical workflows that facilitate the implementation of MR. Application domains spanning gene–environment interactions, public health, complex diseases, and drug target validation are highlighted through representative case studies demonstrating MR's translational potential. Lastly, we critically assess methodological limitations, such as weak instruments, horizontal pleiotropy, population stratification, and data heterogeneity, while proposing future directions for improving robustness and expanding the applicability of MR through integration with multi-omics and cross-ancestry analyses. Overall, MR serves as a cornerstone in modern causal inference research, bridging genetics, epidemiology, and precision medicine to advance our understanding of disease etiology and therapeutic innovation.