Use of whole genome expression analysis of pancreatic adenocarcinoma to predict prognosis after OXi4503 surgery

Abstract

4503 Background: Surgery is the first line treatment for patients with resectable pancreatic ductal adenocarcinoma (PDA). However pancreaticoduodenectomy is challenging and some patients recur quickly, negating the benefit of aggressive resection. To assess for prognostic information in resected tissues, we compared gene expression profiles associated with long and short-survival derived from microdissected FFPE pancreatic cancer specimens. Methods: We selected pathologically confirmed PDAfrom archival FFPE specimens from a single center. Among 29 specimens, 19 were from long (> 300d) and 10 from short (< 300d) survivors. Samples from pancreaticoduodenectomy for non-neoplastic diagnoses (n=8) were handled similarly. Multiple slide sections taken from each paraffin block were microdissected to isolate tumor cells. Following isolation, RNA was amplified, labeled, and hybridized to Affymetrix U133 Plus 2.0 GeneChips. Differentially expressed gene lists were determined between long and short survivors by T-Test and hierarchical clustering (method 1) as well as adaptive linear splines (method 2). Leave one out cross validation (LOOCV) with weighted voting was performed for validation of the data. Results: Between the groups, adenocarcinoma samples were clearly distinguishable from normal specimens using unsupervised analysis. In the microdissected samples, gene clusters created with supervised analyses based on outcome identified ∼300 differentially expressed genes between the long and the short survivor groups (p=0.005). Adaptive linear splines identified multiple genes, the expression levels of which were strongly correlated with survival duration (top gene p-value = 2e-8, FDR = 0.03%). LOOCV analysis confirmed both methods. Conclusion: Gene profiling of microdissected PDA FFPE tissue is feasible and yields data indicative of prognosis after surgery, suggesting that the disparity in outcomes seen after intended curative resection for PDA is largely tumor-intrinsic (rather than due to host factors or surgical technique). Further, these results demonstrate the potential to derive preoperative recurrence risk profiles with ramifications for patient management in this challenging disease. Author Disclosure Employment or Leadership Consultant or Advisory Role Stock Ownership Honoraria Research Expert Testimony Other Remuneration Response Genetics Inc. Response Genetics Inc. Response Genetics Inc. Response Genetics Inc. Response Genetics Inc.