Dissecting polycystic metabolic ovarian syndrome heterogeneity (PMOS): multi-omics, AI subtyping, and precision management.

Polycystic metabolic ovarian syndrome (PMOS) is the most common endocrine disorder, affecting 8-12% of women of reproductive age and increasing the risk of infertility, type 2 diabetes, and heart disease. However, traditional symptom-based diagnostic criteria do not fully capture the underlying biological diversity of PMOS. Recent research using multi-omics and artificial intelligence (AI) has revealed that PMOS is best understood as a syndrome with multiple subtypes. Data-driven clustering of clinical variables consistently identifies four major subtypes: HA-PCOS, OB-PCOS, SHBG-PCOS, and LH-PCOS. Each subtype shows unique hormonal, metabolic, and genetic features, including distinct patterns in lipid, amino acid, and energy metabolism. Immune system disruptions and impaired blood clot breakdown in the ovarian environment are also notable features. Novel findings suggest that altered lactate metabolism and changes in histone lactylation may contribute to granulosa cell dysfunction, although further studies are needed to confirm these mechanisms. Machine learning models that integrate clinical, metabolic, and genetic data have achieved high diagnostic accuracy and can provide interpretable results. Potential targeted therapies, such as GLP-1 receptor agonists, SGLT2 inhibitors, N-acetylcysteine, and NK3R antagonists, have been proposed for specific subtypes, but have yet to be validated in clinical trials. Remaining challenges include small and homogeneous study cohorts, limited long-term data, and barriers to care in low-resource settings. Overall, recognizing PMOS as a syndrome with distinct subtypes highlights the promise of AI-powered, personalized approaches for more effective and equitable management.