645 Background: Pancreas cancer remains a leading cause of cancer-related death. Improved detection of early relapse or early failure of chemotherapy also has the potential to further improve outcomes. Exploring circulating tumor DNA (ctDNA) in this setting is an area of active investigation. Methods: We previously developed an approach, CAPP-Seq, combining high-depth sequencing with several strategies of error-suppression to identify minute amounts of circulating tumor DNA. We then trained and validated a new capture panel for pancreas cancer from 640 tumors from three data sources (TCGA, ICGC, UTSW), targeting 265 kb of the genome. We enrolled two cohorts of patients with pancreatic cancers at Stanford Cancer Center: (1) patients with localized tumors undergoing resection with curative intent, and (2) patients with unresectable or metastatic disease undergoing systemic therapy. Results: As of August 2019, we recruited 131 patients with at least one blood collection, with 63% having resectable disease and 27% having advanced disease; 59 patients had 2 or more blood collections. Stage distribution included 34% stage I, 33% stage II, 18% III, 16% IV disease. Approximately 15% had normal CA19-9 levels. Deep sequencing (4,000x unique depth) of an initial set of resected pancreatic tumors and matched germline specimens identified 1-6 non-synonymous coding mutations per case (median=3, n=14), with the most frequently mutated genes involving KRAS (79%), TP53 (50%), SMAD4 (29%). Among newly diagnosed treatment-naïve patients with resectable adenocarcinoma (n=9), we detected ctDNA in 4 patients (44%) prior to surgery including with AFs ranging from 0.27% - 0.88%. Subsequent sequencing will compare patients with and without neoadjuvant therapy prior to resection, selection of unresectable patients across a larger range of tumor burden and across multiple timepoints, and integration of large-scale copy number variant detection using low-pass whole-genome sequencing. Conclusions: Circulating tumor DNA monitoring with CAPP-Seq shows promise for improved detection of PDAC. Two key applications include early detection of minimal residual disease after resection and early assessment of response to chemotherapy.