Multi-omics Mendelian randomization and machine learning identify candidate therapeutic targets for Alzheimer's and Parkinson's diseases.

Abstract

Neurodegenerative diseases (NDDs), including Alzheimer's disease (AD) and Parkinson's disease (PD), are major public health challenges lacking effective therapies. To identify potential drug targets, we integrated large-scale genome-wide association studies with expression, methylation, protein, and splicing QTL datasets using Mendelian Randomization (MR) and summary-data-based MR (SMR). Colocalization analysis and machine learning were applied to prioritize candidate genes, followed by in silico druggability evaluation through molecular docking and molecular dynamics (MD) simulations. In animal models, candidate genes identified by transcriptomic analysis were further validated using integrative molecular and functional experiments. We identified several genes with potential causal links to AD (e.g., IQCE, HDHD2, ALPP) and PD (e.g., IL15, STK3, CHRNB1). Transcriptomic analyses indicated a consistent downregulation of IL-15 in PD model mice, corroborated by subsequent Western blot and immunohistochemical validation. Among predicted compounds, Prednisolone (ALPP), Sirolimus (IL15), and CHEMBL379975 (STK3) showed favorable binding affinities and stable MD trajectories, suggesting promising therapeutic relevance. Collectively, these findings highlight 12 QTL-regulated genes as promising molecular targets for further investigation in the context of NDDs. While the computational results provide a useful basis for hypothesis generation, experimental validation will be essential to determine the biological relevance and therapeutic potential of these candidate genes and compounds.