Molecular Detection Of Pancreatic Neuroendocrine Tumors Using Methylated Dna Markers: Discovery And Tissue Validation.

e16705 Background: The prevalence of pancreatic neuroendocrine tumors (PNETs) has increased in the last decade. Despite being clinically asymptomatic PNETs can be biologically aggressive. There is currently no reliable non-invasive diagnostic biomarker for PNETs. In this study we aimed to identify and validate methylated DNA marker (MDM) candidates that differentiate PNET from normal pancreas. Methods: For discovery, reduced representation bisulfite sequencing (RRBS) was performed on DNA extracted from frozen normal pancreas (n = 13) and PNET (n = 51) tissues. Area under the receiver operator characteristic curve (AUC), fold-change, and p-value criteria selected candidates MDMs for blinded validation in independent FFPE tissues from primary PNET (n = 67; solid = 50, cystic = 17), normal pancreas controls (n = 24), and normal buffy coat (n = 36) using methylation specific PCR. MDM distributions in primary PNETs were compared to primary lung (n = 36) and small bowel (n = 36) NETs and metastatic PNET tissue (n = 25). The discrimination accuracy of candidate markers was summarized as the AUC with corresponding 95% confidence intervals (CI). Results: From the RRBS discovery, 31 candidate MDMs were selected for validation. Four MDMs ( SRRM3, HCN2, SPTBN4 and TMC6) achieved individual AUCs ≥0.95 in the validation set (Table). These MDMs were similarly discriminant in metastatic PNET tissue and in primary lung and small bowel NETs. Three out of these 4 MDMs perfectly differentiated PNET tissue from buffy coat with AUC of 1 and may be ideally suited for further development of a blood-based assay. Conclusions: We identified and validated novel MDMs in tissue that discriminate PNETs from controls with normal pancreas and buffy coat with high accuracy. These MDMs also differentiated metastatic PNETs from normal pancreas tissue. Further exploration of these candidate tissue MDMs in plasma can potentially guide diagnosis and management of PNETs. Funding: P30DK084567. [Table: see text]