Introduction: Polycystic ovary syndrome (PCOS) is one of the most common and heterogeneous endocrine disorders, affecting up to 10% of reproductive-age women worldwide. Its variability in clinical presentation contributes to delayed diagnosis, inconsistent subtyping, and trial-and-error treatment, while current approaches still lack robust biomarkers to guide personalized therapy. Advances in high-throughput multi-omics and artificial intelligence (AI) now provide a unique opportunity to unravel molecular heterogeneity and link biological signatures with clinical outcomes. Building on these developments, the present study aims to develop and validate an AI-driven multi-omics fusion framework for accurate diagnosis, molecular subtyping, and personalized management of PCOS.
Methods: Genomic, proteomic, metabolomic, clinical, and wearable sensor data will be integrated into a unified AI model. Preprocessing will include normalization and dimensionality reduction, followed by multi-modal fusion. Machine learning algorithms (deep learning, ensemble methods) will identify biomarker signatures and classify PCOS subtypes. Interpretability techniques will uncover key biological pathways. Dynamic cycle-tracking data from wearables will complement static omics profiles, enhancing prediction accuracy. Pilot testing will leverage existing PCOS cohorts and public data repositories to ensure real-world applicability.
Results: The proposed framework is expected to identify novel biomarker panels and delineate molecularly distinct subtypes of polycystic ovary syndrome (PCOS), such as metabolic-dominant and reproductive-dominant forms. It will also be designed to predict individual treatment responses, thereby enhancing the selection of appropriate lifestyle, metabolic, or fertility interventions. By doing so, the framework aims to reduce diagnostic ambiguity and streamline clinical decision-making processes. Beyond PCOS, it is intended to provide a scalable template that can be applied to other endocrine disorders through AI-driven multi-omics fusion.
Conclusion: This project proposes a transformative, systems-level approach to precision endocrinology. By fusing multi-omics and real-time health data through AI, it addresses clinical heterogeneity, improves diagnostic accuracy, and enables cost-effective personalized strategies in PCOS. Beyond PCOS, this paradigm holds global potential for advancing care in obesity, thyroid disease, and other endocrine conditions.