Integrated metabolomics and transcriptomics to reveal biomarkers and mitochondrial metabolic dysregulation of premature ovarian insufficiency.

Background: The metabolic characteristics of premature ovarian insufficiency (POI), a reproductive endocrine disease characterized by abnormal sex hormone metabolism and follicle depletion, remain unclear. Metabolomics is a powerful tool for exploring disease phenotypes and biomarkers. This study aims to identify metabolic markers and construct diagnostic models, and elucidate the underlying pathological mechanisms for POI. Methods: Non-targeted metabolomics was utilized to characterize the plasma metabolic profile of 40 patients. The metabolic markers were identified through bioinformatics and machine learning, and constructed an optimal diagnostic model by classified multi-model analysis. Enzyme-linked immunosorbent assay (ELISA) was used to verify antioxidant indexes, mitochondrial enzyme complexes, and ATP levels. Finally, integrated transcriptomics and metabolomics were used to reveal the dysregulated pathways and molecular regulatory mechanisms of POI. Results: The study identified eight metabolic markers significantly correlated with ovarian reserve function. The XGBoost diagnostic model was developed based on six machine learning models, demonstrating its robust diagnostic performance and clinical applicability through the evaluation of receiver operating characteristic (ROC) curve, decision curve analysis (DCA), calibration curve, and precise recall (PR) curve. Multi-omics analysis showed that mitochondrial respiratory chain electron carrier (CoQ10) and enzyme complex subunits were down-regulated in POI. ELISA validation revealed an elevation in oxidative stress markers and a reduction in the activities of antioxidant enzymes, CoQ10, and mitochondrial enzyme complexes in POI. Conclusion: Our findings highlight that mitochondrial dysfunction and energy metabolism disorders are closely related to the pathogenesis of POI. The identification of metabolic markers and predictive models holds significant implications for the diagnosis, treatment, and monitoring of POI.