Abstract 4883: A glycan gene signature that robustly predicts prognosis in patients with pancreatic ductal adenocarcinoma

Abstract

Abstract Purpose: Pancreatic ductal adenocarcinoma (PDAC) remains the fourth leading cause of cancer-related deaths in the United States, with an overall 5-year survival rates of ~8%. The currently used clinicopathological factors (e.g. tumor size and grade, lymph node etc.) for determining patient prognosis are suboptimal. The only FDA-approved molecular non-invasive prognostic biomarker for PDAC patients is CA-19-9, which suffers from inadequate sensitivity and specificity. Nonetheless, the clinical use of CA-19-9, along with the growing evidence that altered glycosylation plays an important role in PDAC pathogenesis, provided us with a rationale to explore the clinical significance of glycan-related genes as potential prognostic biomarkers in PDAC. Experimental design: We performed a comprehensive biomarker discovery (TCGA cohort, n=182), and validation (ICGC, n=80; GSE62452, n=65) to identify a glycan gene signature for predicting overall survival (OS) in PDAC patients. Subsequently, this gene signature was validated in two, independent, clinical cohorts (test cohort, n=103; validation cohort, n=227). The performance of this signature was further evaluated by univariate and multivariate CoxPH analyses. Lastly, using a logistic-regression model, we explored the feasibility of our glycan signature in identifying various molecular subtypes of PDAC. Results: A comprehensive analysis of 411 glycan genes using Cox-LASSO regression modelling led to the identification of a 12-glycan gene signature, which robustly predicted overall survival (OS) of PDAC patients in the discovery (AUC=0.78), and two validation cohorts (ICGC, AUC=0.72; and GSE62452, AUC=0.70). Subsequent qRT-PCR validation in two in-house clinical cohorts revealed that a 9-gene signature was a robust predictor of OS (Test Cohort, HR: 1.81, 95% CI, 1.22-2.69, p=0.003; Validation Cohort, HR: 2.72, 95% CI, 2.00-3.69, p<0.0001). In univariate analysis, in addition to the 9-gene signature, the nodal status and CA-19-9 levels were significant predictors; while in the multivariate analyses, the gene signature emerged as an independent predictor of OS. A risk-assessment model including the 9-gene signature and the two clinical factors further improved the OS prediction (Test Cohort, HR: 2.71, 95% CI, 1.85-3.99, p<0.0001; Validation Cohort, HR: 2.72, 95% CI, 2.03-3.63, p<0.0001). Intriguingly, our signature was also highly accurate in identifying PDAC subtype with poor survival (i.e. squamous subtype) in the TCGA (AUC=0.87, P<0.0001) and the ICGC cohorts (AUC=0.89, P<0.0001). Conclusion: Our systematic biomarker discovery and validation efforts led to the identification and establishment of a 9-gene glycan signature that robustly predicts survival in PDAC patients, and can accurately identify poor PDAC subtypes - highlighting its potential clinical significance for the personalized management of PDAC patients. Citation Format: Priyanka Sharma, Raju Kandimalla, Jasjit K. Banwait, Masayuki Sho, Yasuhiro Kodera, Ajay Goel. A glycan gene signature that robustly predicts prognosis in patients with pancreatic ductal adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 4883.