Identification of Glucocorticoid-Related Hub Genes and Therapeutic Candidates in Polycystic Ovary Syndrome via Multiomics Integration and Machine Learning

Abstract

Polycystic ovary syndrome (PCOS) is a common endocrine and metabolic disorder with significant diagnostic and therapeutic challenges. Increasing evidence suggests that glucocorticoids (GCs) contribute to PCOS development, yet the underlying molecular mechanisms remain unclear. In this study, transcriptomic datasets of granulosa cells from the Gene Expression Omnibus were integrated and batch-corrected to identify differentially expressed genes (DEGs). Weighted gene co-expression network analysis was performed to define GC-related modules, and hub genes were screened using machine learning algorithms. Six hub genes (CRIP1, EPS8, PCSK9, PLEKHH2, RBP7, and SYTL5) were identified as potential biomarkers, with receiver operating characteristic curve analysis confirming their strong diagnostic performance (area under the curve > 0.8). A nomogram integrating these genes achieved an AUC of 0.9481, indicating superior diagnostic value. Candidate therapeutic compounds were predicted via the Drug–Gene Interaction Database and evaluated through molecular docking, which demonstrated high-affinity binding of DH376, LEI105, and DO34 to CRIP1 (binding energy < −5.8 kcal/mol). Regulatory network analysis further suggested the involvement of GC-related pathways in cholesterol metabolism and insulin resistance. These findings reveal novel molecular insights into GC-associated mechanisms in PCOS, propose a six-gene diagnostic signature, and highlight CRIP1-targeting compounds, particularly DO34, as computationally predicted candidates warranting experimental validation, thereby laying a foundation for precision diagnostics and targeted therapies in PCOS.