PROJECT SUMMARY/ABSTRACT Approximately 700,000 pancreatic cysts are incidentally detected by imaging each year. Up to 70% of these radiologically-detected cysts are intraductal papillary mucinous neoplasms (IPMNs), bona fide precursor lesions to pancreatic ductal adenocarcinoma (PDAC), a malignancy with a 5-year relative survival rate of only 12%. The goal of our parent grant is to fulfill the unmet need to discover a noninvasive biomarker and imaging approach that has greater accuracy in predicting IPMN pathology than conventional radiologic and clinical features, thereby enhancing clinical decision-making and promoting more good than harm for patients at- risk to harbor or develop early PDAC. Our central hypothesis is that radiomic features extracted from preoperative CT scans will more accurately predict IPMN pathology than conventional radiologic features, both individually and in combination with a plasma-based miRNA genomic classifier (MGC). We further hypothesize that the most promising radiomic features may serve as noninvasive surrogates for underlying biological processes (which are miRNA-mediated and/or linked to mucin expression) that drive IPMN development and progression to invasion. In the two-year extension period, we plan to continue to address this goal and hypothesis by applying new artificial intelligence (AI)-based approaches and incorporating additional classes of biomarkers (blood-based and behavioral). We aim to: evaluate the value of artificial intelligence (AI)-driven CT deep learning radiomic features in predicting malignant versus benign IPMN pathology in retrospective and prospective cohorts (Aim 1), evaluate telomere length and telomerase activity in the blood as candidate molecular markers of high-grade IPMNs or early-stage PDAC (Aim 2), and use behavioral AI to predict malignant transformation among patients with a high risk to develop PDAC (Aim 3). This line of translational research has potential to foster clinically actionable information that could be used to rapidly and cost-effectively personalize care for individuals with IPMNs and ultimately reduce the burden of PDAC as a major health problem, a goal in line with the parent award and with NCI’s mission to lead, conduct, and support cancer research across the nation to advance scientific knowledge and help all people live longer, healthier lives.

PROJECT NARRATIVE Public Health Relevance: Intraductal papillary mucinous neoplasms (IPMNs) are cystic precursor lesions to pancreatic ductal adenocarcinoma (PDAC) incidentally-detected by imaging in more than 500,00 Americans each year. There is an unmet need to discover noninvasive approaches to differentiate ‘low- risk/benign’ IPMNs that can be monitored from ‘high-risk/malignant’ IPMNs that warrant surgery and its associated risks of morbidity and mortality. In our parent award, we are evaluating the hypothesis that radiomic features extracted from preoperative computed tomography (CT) scans will more accurately predict IPMN pathology than conventional radiologic features, both individually and in combination with a plasma- based ‘miRNA genomic classifier’ (MGC). In this extension, we plan to 1) evaluate the value of artificial intelligence (AI)-driven CT deep learning radiomic features in predicting malignant versus benign IPMN pathology in retrospective and prospective cohorts, 2) evaluate telomere length and telomerase activity in the blood as candidate molecular markers of high-grade IPMNs or early-stage PDAC, and 3) use behavioral AI to predict malignant transformation among patients with a high risk to develop PDAC.