Impact of organ-adjacent extracellular vesicle proteomics from uterine lavage on accurate detection and discrimination of ovarian cancer.

5581 Background: Survival in ovarian cancer (OvCA) exceeds 90% when disease is localized, yet most cases present at advanced stage and effective early detection remains elusive. Blood-based liquid biopsies may be limited by low tumor signal in early disease. We evaluated whether organ-adjacent sampling via uterine lavage extracellular vesicle (EV) proteomics combined with machine-learning (ML) classification could enable accurate detection of OvCA and clinically relevant discrimination from endometrial cancer (EndoCA). Methods: Uterine lavage samples were collected under IRB-approved protocols from women undergoing gynecologic evaluation for abnormal uterine bleeding and/or abnormal pelvic imaging. EVs were isolated using an affinity-based capture method and analyzed by liquid chromatography-tandem mass spectrometry. Protein features meeting predefined quality thresholds were analyzed using a novel ML pipeline incorporating entropy-based marker scoring and correlation filtering. Classifier performance was assessed using repeated random two-fold validation and receiver operating characteristic analysis. Results: Among 807 participants, diagnoses included OvCA (n=85), benign conditions (n=488), and EndoCA (n=234). An OvCA-versus-benign classifier derived from a 91-protein panel demonstrated strong discrimination (AUC >0.9). Across 100 validation splits, 83 of 85 OvCA cases were correctly classified, including all stage I cases (27/27). A separate 21-protein classifier distinguished OvCA from EndoCA with a sensitivity of 0.94 and specificity of 0.92. Conclusions: EV proteomic analysis of uterine lavage specimens enables accurate detection of ovarian cancer, including stage I disease, and reliably distinguishes OvCA from EndoCA in women undergoing gynecologic evaluation. These findings support further prospective clinical validation of organ-adjacent EV proteomics as a translational diagnostic strategy for earlier and more precise classification of gynecologic malignancies.