An endometrial tissue-based predictive model for polycystic ovary syndrome constructed from immuno-metabolic dysregulation features mediated by ACO1

Polycystic ovary syndrome (PCOS) is a multifactorial endocrine disorder characterized by reproductive and metabolic abnormalities. This study aimed to identify key immunometabolic regulators in endometrial tissue and construct a predictive model for PCOS using machine learning approaches. Three endometrial transcriptomic datasets (GSE277906, GSE193123, GSE199225) were integrated and analyzed for differentially expressed genes (DEGs), immune cell infiltration, and metabolic pathway enrichment. Core genes were identified via protein–protein interaction networks and functional annotation. A predictive model was developed using SVM-RFE, XGBoost, and random forest algorithms and validated through qRT-PCR on granulosa cell samples. Five core metabolism-related genes were identified, among which ACO1 was consistently downregulated and negatively correlated with CD8⁺ T cell infiltration. High ACO1 expression was enriched in oxidative phosphorylation and mTOR signaling, while low expression was associated with immune activation. The random forest model incorporating ACO1, CHPF, and STOML1 achieved strong predictive performance (AUC = 0.800). ACO1 may function as an immunometabolic modulator by linking iron metabolism, oxidative stress, and T cell activity. Its downregulation may contribute to local immune suppression and endometrial dysfunction in PCOS. The tissue-level model demonstrated good diagnostic value and biological interpretability across cohorts. This study highlights ACO1 as a key biomarker of immunometabolic dysregulation in PCOS and presents a robust predictive model for early diagnosis. The findings offer new insights into the molecular mechanisms underlying PCOS and suggest potential targets for precision treatment.