Abstract Background Despite advances in treatment for cancer, the 5-year survival rate of pancreatic adenocarcinoma(PDAC) remains approximately 10% as of 2023. Early detection of PDAC is crucial for improving therapeutic outcomes. In order to detect PDAC at early stages, monitoring patients with clinical risk factors is important but relying solely on these risk factors is not sufficient. There is an urgent need for detection of biological markers such as microRNAs (miRNAs). In this study, we analyzed urinary miRNAs as a potential non-invasive biomarker and developed a machine learning classifier to detect pancreatic cancer from high-risk patients. Method A total of 165 participants with Stage0 to Ⅳ PDAC and 109 high-risk participants were analyzed. High-risk patients were defined as those who had at least one of the following diseases and/or risk factors: type 2 diabetes, chronic pancreatitis, intraductal papillary mucinous neoplasm, pancreatic cysts, dilation of main pancreatic duct and family history of PDAC. Urinary miRNA were sequenced by next generation sequencer. The data were split into 80% of training data and 20% of validation data for the development of a classifier. Machine learning based algorithm was deployed to construct an algorithm from the training data and its performance was evaluated by the remaining validation data. Results: Differential expression analysis revealed 12 upregulated and 7 downregulated miRNAs. Among downregulated miRNAs, PDAC related pathways such as growth signaling were significantly enriched. The performance of the classifier was AUC: 0.87, sensitivity: 0.80, and specificity 0.77. The sensitivity for early stage (Stage0-Ⅰ)PDAC was 0.80 and 0.88 especially when limited to StageⅠ. Conclusion: The classifier was able to detect PDAC at early stages with high sensitivity and specificity. This study demonstrated the potential for effective monitoring of high-risk patients for the early detection of PDAC, highlighting the feasibility and advantages of urine sampling due to its simplicity. To further evaluate the performance of the algorithm, a validation study is recommended. Citation Format: Tomoya Kawase, Hiroyasu Kusaka, Koji Yoshida, Yasutaka Kato, Junji Kita, Hiroshi Kurahara, Shunsuke Kondo, Atsushi Satomura, Yuji Yoshida, Hiroki Yamaguchi, Yoriko Ando, Motiki Mikami, Mika Mizunuma, Yuki Ichikawa. Development of a Urinary miRNA-Based Classifier for Detecting Pancreatic Cancer in High-Risk Groups [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research; 2024 Sep 15-18; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2024;84(17 Suppl_2):Abstract nr B026.