Gastro Hep Advances (Jan 2022)
Discovery and Validation of Methylated DNA Markers From Pancreatic Neuroendocrine Tumors
Abstract
Background and Aims: Methylated DNA markers (MDMs) accurately identify several different cancer types, but there are limited data for pancreatic neuroendocrine tumors (pNETs). We aimed to identify MDM candidates in tissue that differentiate pNETs from normal pancreas. Methods: wUsing DNA from frozen normal pancreas (13) and pNET (51) tissues, we performed reduced representation bisulfite sequencing for MDM discovery. Validation in independent formalin fixed paraffin embedded tissues used pNET cases (67; solid = 50, cystic = 17), normal pancreas (24), and buffy coat (36) controls. Primary pNET MDM distributions were compared with lung (36), small bowel (36) NETs, and metastatic pNET (25) tissues. The discrimination accuracy was summarized as the area under the receiver operator characteristic curve (AUC) with corresponding 95% confidence intervals (CIs). Fisher’s linear discriminant analysis was performed to estimate a linear discriminate score (LDS) differentiating normal from pNET tissue and applied to all patient groups; discrimination accuracy of the LDS was summarized as the bootstrap cross-validated AUC. Results: Median AUC for distinguishing normal pancreas from pNET tissue was 0.91 (interquartile range: 0.80–0.93). The cross-validated AUC for the LDS discriminating normal pancreatic tissue from primary and metastatic pNETs was 0.957 (95% CI 0.858–1.0, P < .0001) and 0.963 (95% CI 0.865–1.0, P < .0001), respectively. The LDS for the MDM panel was significantly higher for primary pNET, metastatic pNET, lung NET, and small bowel NET, each compared with normal pancreas tissue (P < .0001). There was no statistical difference between primary pNET and metastatic pNET (P = .1947). Conclusion: In independent tissue validation, MDMs accurately discriminate pNETs from normal pancreas. These results provide scientific rationale for exploration of these tissue MDMs in a plasma-based assay for clinical application.