Causal analysis of Alzheimer’s disease molecular data

Abstract

Abstract Background Genome-wide studies have revealed many molecular associations with Alzheimer’s disease (AD), but association alone does not identify cause, mechanism, or treatable target. This review surveys methods used to infer causal relationships from AD molecular data and assesses their core findings. Main body We reviewed causal-inference work applied to AD genetics, transcriptomics, epigenomics, proteomics, and metabolomics, covering Mendelian randomization, probabilistic graphical and Bayesian network models, causal reasoning over knowledge graphs, time-resolved methods including Granger causality, and mediation analysis. Genetic-instrument and network methods have identified candidate causal molecules, mapped pathways, and supported some drug targets. However, several associations reported in observational studies have not been supported by genetic analyses, and apparent disagreements between studies often reflect differences in the estimand, the tissue from which genetic instruments are derived, or assumptions of the methods used. Conclusion Brain-relevant genetic instruments, longitudinal multi-omics, single-cell resolution, and triangulation across methods are needed to separate cause from consequence and to prioritize intervention targets for experimental validation.