Early detection of ovarian cancer: An accurate high-throughput extracellular vesicle test.

5582 Background: The high mortality of Ovarian cancer (OC) has been attributed to late-stage diagnosis and the lack of an effective early detection strategy, particularly for asymptomatic women. In this study, we developed and validated a high-throughput OC detection test based on plasma extracellular vesicle (EV)-associated biomarkers. Methods: A case-control study was conducted to evaluate blood-borne EV-associated ovarian cancer biomarkers, including miRNAs, proteins, lncRNAs, miscRNAs, MtrRNAs, MttRNAs, rRNAs, scaRNAs, snRNAs, and tRNAs. Protein and RNA biomarkers were identified by mass spectrometry and RNA sequencing, respectively. Training (n=453) and independent test (n=471) sample sets were used to develop and validate a multivariate index assay (MIA). The MIA was further validated using a high-throughput, pathology laboratory compatible, EV isolation platform (EXO-NET) and two independent sample cohorts (n=97 and n=532). The classification accuracy, sensitivity and specificity of the MIA was compared to that of CA125 levels. Results: Discovery and Training phases - more than 100,000 EV-associated biomarkers were identified from 453 EV samples. The classification performance of these biomarkers was assessed using machine learning algorithms. EV-associated protein and miRNA biomarkers delivered the highest performing classifiers and, therefore, were used in subsequent MIA development and training. During the training phase, multivariate classification algorithms were validated using a 10-fold cross-validation method. The highest performing classifiers for EV-associated protein and miRNA, at specificity of 98%, achieved sensitivities of 90% and 82%, respectively. Validation phase: Locked classification algorithms ( i.e. MIAs) were validated using two independent sample cohorts and reported classification accuracies of 92-98%, significantly outperforming CA-125 (CE = 62%, p<0.001). Automated high-throughput MIA – All stages OC: the best performing automated high-throughput MIA demonstrated an overall sensitivity of 92% (95% CI, 75–96%) and specificity of 93% (95% CI, 86–96%) for all stages of OC, Positive Predictive Value of 95% (CI, 93-96%) and Negative Predictive Value of 80% (CI, 76-89%) at 98% specificity (n=532). Stage I OC: Importantly, the MIA displayed a sensitivity of 90% (95% CI, 76–100%) and specificity of 96% (95% CI, 40%–99%) for stage I OC. While CA125 have an overall sensitivity for all stages of OC of 61% (95% CI, 53–69%), with a sensitivity of 44% for stage I (95% CI, 28–62%). Conclusions: In this study we report the development and validation of an accurate, automated high-throughput EV-based test for early detection of ovarian cancer. The test delivers significant improvements in sensitivity and specificity compared to CA-125, especially in detecting early-stage OC.