Abstract 1925: Development of a multi-omics diagnostic approach for the early detection of ovarian cancer in asymptomatic women

Abstract Objectives: Ovarian cancer (OC) is the fifth leading cause of cancer-related deaths among women. Unfortunately, for most patients, detection of OC occurs at late stages (III/IV) when the five-year survival rate is <30%. To change this paradigm, screening methods must be developed that are minimally invasive, highly sensitive, and disease specific. We have determined that utilizing multi-omics, which include novel classes of lipid and protein biomarkers with machine-learning, enables the robust detection of ovarian cancer across stages and subtypes, while requiring <500µL of serum. Methods: We utilized a multi-omics approach to characterize a clinically annotated cohort comprising serum samples from individuals with ovarian cancer (N=235) and normal donors (N=82). The cohort was obtained from the University of Colorado Gynecologic Tissue and Fluid Bank (IRB #07-935 and 21-4787) and commercial vendors. UHPLC-MS data were generated from 20µL of extracted serum to profile lipids and fatty acids, while manual immunoassays were performed for a panel of protein biomarkers using unextracted serum. Machine learning (ML) model training was performed using the biomarker classes separately and in combination to identify top-performing models, using 20-fold cross validation. Results: We profiled a total of 611 features of the lipidome by UHPLC-MS and protein biomarkers by immunoassay. The main molecular drivers contributing to best performing initial OC-specific signatures included a combination of lipids, fatty acids, and proteins together: multi-omic model consistently exhibited highest AUC when compared to individual biomarker classes. The AUC for High Grade Serous OC (HGSOC) was 0.98. At 98.2% specificity, sensitivity was 93.1% for all stages and subtypes of OC (93.8% sensitivity for all subtypes of early-stage (I/II) OC). The sensitivity for early stage HGSOC was 92.6%. Conclusions: Early detection of OC is critical to improve patient outcomes, but current screening tools for OC lack early-stage sensitivity and specificity. The application of a multi-omics, machine learning approach for the screening of post-menopausal asymptomatic women offers significantly improved performance over CA125 for the detection of early-stage HGSOC. Future research will validate the performance of this approach in a large prospective cohort. Citation Format: Rachel Culp-Hill, Charles Nichols, Collin Hill, Brendan Giles, Robert A. Law, Enkhtuya Radnaa, Kian Behbakht, Benjamin G. Bitler, Maria Wong, Connor Hansen, Mattie Goldberg, Vuna Fa, Violeta Guthrie, James R. White, Anna Jeter, Abigail McElhinny. Development of a multi-omics diagnostic approach for the early detection of ovarian cancer in asymptomatic women [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1925.