Identifying novel druggable targets and repurposable drugs for premature ovarian insufficiency by integrated multiomics and causal inference analysis.

Premature ovarian insufficiency (POI) is a leading cause of female infertility. Its mechanisms are poorly understood, and effective therapies are lacking. In this study, we aimed to identify novel druggable targets and repurposable drugs for POI through an integrated multiomics and computational pharmacology approach. We integrated large-scale proteomic data from two independent cohorts (deCODE, N = 35,559; UK Biobank, N = 54,219) using Mendelian randomization, Bayesian colocalization, and single-cell RNA sequencing. Seven high-confidence targets were identified: EPHA4, FSTL3, NUCB2, OXT, SERPINA12, TNFRSF6B, and FABP1. Among these genes, EPHA4, FSTL3, and NUCB2 were significantly dysregulated in cisplatin-induced mouse and human granulosa cell models (P < 0.05 to P < 0.001) and exhibited high diagnostic accuracy (AUC = 0.92-0.96), supporting their potential as both biomarkers and therapeutic targets. Molecular docking revealed strong binding affinities, notably for cycloheximide binding to EPHA4 (-7.8 kcal/mol), with molecular dynamics confirming stable interactions (root mean square deviation, RMSD < 2.0 Å), providing a structural basis for drug repurposing or lead optimization. The functional enrichment results suggested that fibrosis, inflammation, and metabolic dysregulation are involved in POI pathogenesis. Collectively, our findings establish a multiomics-to-therapy pipeline that not only prioritizes causal targets for POI but also provides translational opportunities, from biomarker-guided diagnosis to computationally driven drug repositioning, paving the way for mechanism-based interventions in ovarian aging.