Reproducible serum metabolomics drives biomarker discovery in ovarian cancer

Ovarian cancer (OC) remains one of the deadliest gynecologic malignancies, with a clear unmet need for novel biomarker strategies. Altered metabolism has emerged as a cancer hallmark, with serum reflecting rewired cellular pathways supporting tumor growth and survival. Advancements in mass spectrometry (MS) have improved precision and confidence in feature identification, allowing for impressive analytical depth in extremely low volumes (<10µL) of biofluid. Our previous work has focused on lipidomics, but the biological mechanisms revealed through metabolomics can offer novel and complementary insight into tumor metabolism. Metabolomics, therefore, has the potential to become a powerful tool for biomarker discovery through the characterization of systemic metabolic alterations in individuals diagnosed with OC.Two independent, clinically annotated cohorts of serum representing the population of women experiencing symptoms of OC were analyzed using untargeted metabolomics (UHPLC-HRMS). Cohort #1 (N=519), obtained from University of Colorado and commercial vendors comprised patients diagnosed with OC across stages and subtypes (N=219: 80 stage I/II, 139 stage III/IV), benign gynecological disorders (N=168), gastrointestinal disorders (N=50), and healthy donors (N=82). Cohort #2 (N=400) comprised patients diagnosed with OC (N=116: 50 stage I/II, 66 stage III/IV), benign adnexal masses (N=116), borderline tumors (N=19), other symptomatic individuals (N=114), and healthy donors (N=35). We observed alterations in several key metabolic pathways, many of which have been individually implicated in OC biology, including decreased amino acids and bile acids, with increased acyl-carnitines and fatty acids. To evaluate cohort reproducibility, we compared statistical trends (cancer v. non-cancer) from each cohort. Of the significantly altered features in both cohorts, 68% maintained directionality. 90% of those maintaining directionality are a part of the key pathways above. The consistency in metabolic alterations across two independent cohorts suggests these pathway intermediates transcend batch-to-batch variability and could be leveraged as diagnostic biomarkers . In fact, initial machine learning-based modeling shows that models combining metabolites, lipids, and proteins result in AUCs >90%. MS-based metabolomics represents a snapshot of an individual's disease state, offering insight into metabolic alterations indicative of early-stage cancer through a non-invasive sample. Here, we identified consistent metabolic alterations in OC serum across two independent, clinically annotated cohorts that align with previously reported pathway alterations. This demonstrates the analytical robustness of our approach and highlights the potential of MS-based metabolomics to move beyond discovery, toward development of clinically viable biomarkers for early-stage OC detection.