Artificial intelligence-driven integration of multi-biofluid omics and clinical phenotype enables stratification of endometrial cancer

Endometrial cancer (EC) incidence is rising, yet current diagnostics lack precision and scalability. We develop an artificial intelligence (AI)-based platform integrating multi-biofluid omics and clinical data for EC stratification. Using two independent cohorts from different clinical centers (531 participants for model development, 204 for external validation), we collect 1,179 samples (plasma, cervical/uterine secretions) and the corresponding clinical data (age, ultrasound, etc.). Machine learning identifies EC-specific signatures, and the AI framework fuses omics features with clinical factors to enable multilevel risk stratification. On the external validation cohort, the platform achieves 95.65% sensitivity for minimally invasive EC screening and balanced performance with an area under the curve (AUC) value of 0.94 for EC confirmation. The model also shows potential for high-risk subtype detection. Biological plausibility is supported by identified omics signatures. A web tool is developed to support clinical translation. This platform demonstrates the potential of multi-omics and AI in precision oncology.