GlycoKnow Ovarian: a Glycoprotein-based, Serum Diagnostic to Distinguish Ovarian Cancers from Benign Pelvic Masses

ObjectiveBlood-based biomarkers offer an unprecedented opportunity to realize the promise of precision medicine in improving diagnostic workflows. Previous peer-reviewed studies have established the association of the circulating glycoproteome with ovarian cancer. Here a glycoproteomic classifier was built, tested, and applied to both internal and external validation cohorts to distinguish malignant from benign pelvic masses. Study DesignSerum samples from healthy patients and patients with pelvic masses were collected from both retrospective biobanks and prospective trials. In total, 38 peptides and glycopeptides were quantified by InterVenns targeted mass spectrometry platform. A classifier to predict malignancy was built, locked, and evaluated in a hold-out test set. The locked diagnostic was then evaluated in an internal validation cohort as well as an external validation cohort from UT MD Anderson Cancer Center. ResultsLASSO-regularized logistic regression in the training cohort resulted in a locked classifier with 16 features that was evaluated on a hold-out test set, with strong performance in ovarian cancer and benign pelvic masses (AUC=0.909; sensitivity=86.7%; specificity=89.7%). Comparable performance was observed in further validation cohorts: one internal (stage I-II sensitivity=63.6%; stage III-IV sensitivity=85.7%; specificity=82.7%) and one external (stage I-II sensitivity=64.0%; specificity=86.2%), with varying per-stage prevalence. ConclusionsA novel, CA-125-independent glycoprotein panel was developed to help distinguish benign conditions from ovarian cancer. These circulating biomarkers have great potential to detect ovarian cancer while retaining high specificity and could open new avenues for an improved ovarian cancer diagnostic. Research HighlightsO_LIA glycoprotein-based liquid biopsy demonstrates strong performance in distinguishing ovarian cancer from benign masses C_LIO_LIThe locked and validated classifier exhibits comparable performance in multiple validation cohorts C_LIO_LIAs an underexplored layer of biology, protein glycosylation has the potential to enable novel precision medicine solutions C_LI