Abstract 478: Extracellular vesicle based multi-omics prediction model for the early detection of pancreatic cancer

Abstract

Abstract Introduction: Pancreas cancer (PC) is projected to become the third leading cause of cancer deaths in the US, however there is no screening test in routine clinical use. A minimally invasive test for early detection of PC is a critical unmet clinical need. Extracellular vesicles (EVs) shed into the circulation include a diverse array of surface and cargo biomolecules that can be leveraged as biomarkers for the early detection of PC. We hypothesize that systemic molecular characterization of plasma EVs, secreted by latent yet progressing precursor lesions (PL) of the pancreas, may lead to the development of high accuracy classification algorithms for stratification of patients for the early detection of PC. Herein, we developed a liquid biopsy-based PACE assay (Pancreatic Cancer Exosomics) comprised of a multi-omics (proteomics, metabolomics, and lipidomics) biomarker panel. Methods: To develop a functional EV-based multi-omics biomarker panel, we performed metabolomics, lipidomics and proteomics analysis of EVs from high quality pre-surgery, fasting plasma samples obtained from patients with early stage pancreatic cancer (IA, IB and IIA) (N= 60), normal control subjects without cancer (N = 50), patients diagnosed with pancreatitis (benign) (N= 39), and early stage colorectal cancer (N=30) as related disease cohorts. For selection of candidate biomarkers, we randomly divided 3/5th of the sample set as discovery and the remainder 2/5th as the test cohort. To consider the effect of potential correlation for predictors, we applied the elastic net penalty to generalized linear model (ELNET). A 100-fold cross-validation approach was used to determine tuning parameters and calibrate the prediction model in the discovery set. Results: These analyses allowed us to develop a 12 analyte multi-omics panel that could discriminate early stage PC from normal controls (NC) with 98.2% accuracy (Specificity (SP) = 85.2% and Sensitivity (SN) = 71.5%) and 95.4% accuracy (SP = 92.8% and SN = 59.3%) for discovery and validation sets, respectively. A unique aspect of the study design was the inclusion of control cohorts to eliminate non-specific markers. Furthermore, extensive clinical outcome data allowed us to develop multi-analyte predictors of overall survival and progression free survival with high accuracy. Validation of these findings using orthogonal methods is ongoing. Conclusions: In summary, these preliminary data provide strong support for the development of a liquid biopsy-based EV biomarker panel for the early detection of PC. We believe the positive predictive value (PPV) of this panel can be further improved in conjunction with serum CA 19-9 that is in routine clinical use. Citation Format: Yaoxiang Li, Keith Unger, Charles Hinzman, Meth Jayatilake, Anton Iliuk, Shivani Bansal, Michael Girgis, Partha Banerjee, John B. Tyburski, Todd Bauer, Amrita K. Cheema. Extracellular vesicle based multi-omics prediction model for the early detection of pancreatic cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 478.