More than 2% of the US general population harbors an asymptomatic pancreatic cyst, and the incidence is rising due to increasing use of abdominal imaging. Pancreatic cystic lesions include mucin-secreting cysts - specifically, Intraductal Papillary Mucinous Neoplasms (IPMNs) and Mucinous Cystic Neoplasms (MCNs), which are bona fide precursor lesions of pancreatic adenocarcinoma, and others, such as serous cystadenomas (SCAs) that have minimal potential for malignancy. Even within the microcosm of IPMNs and MCNs, the majority of cysts are indolent, and current consensus guidelines for therapeutic stratification lead to over-treatment in many cases. Nonetheless, the long-term survival of patients with non-invasive mucinous cysts drops from 90-100% to nearly half once invasion occurs, underscoring the continuing importance of screening cyst populations. Distinguishing pancreatic cysts that harbor greatest potential for progression to malignancy from those that are unequivocally indolent will allow rational screening and management of this burgeoning epidemic. We hypothesize that genetic drivers of pancreatic cyst formation interact with the physical microenvironment to fuel malignant progression, and the latter is reflected in measureable imaging features and humoral immune responses. To address this hypothesis, we will create an "Imaging and Molecular Characterization Laboratory" (IMCL) at MD Anderson, and partner with four high volume pancreatic centers (UCSF, UCSD, Indiana University and University of Utah), for access to >1,000 surgically resected retrospective cyst samples, and an expected accrual of ~1,300 prospective cyst patients over the period of this proposal. In Aim 1, we will correlate quantitative measurements of the physical microenvironment of pancreatic cysts (obtained from standard-of-care diagnostic CT scans) with underlying histopathology and genomic profiles. In Aim 2, we will identify the host immune response to malignant progression in pancreatic cysts by measuring autoantibody responses in the serum to over three million genome-wide epitopes using a Human Peptide Array (Roche-NimbleGen). In both aims, we will use existing diagnostic imaging and banked serum samples from patients with surgically resected and histopathology validated pancreatic cysts as a training set, followed by application in a test set from prospectively accrued samples. Finally, in Aim 3, we will merge the data from both prior aims to develop an integrated algorithm for predicting neoplastic progression in pancreatic cysts. Multi-center coordination of retrospective and prospective biospecimen collection and all aspects of data management will be supported by a Statistical and Centralized Data Management Core (SCDMC) within the IMCL. By integrating the amplitude and targets of the host immune response with quantitative physical features obtained using standard-of-care diagnostic imaging studies, we hope to develop reliable algorithms that can predict the risk of progression in pancreatic cysts using relatively non-invasive and portable approaches, and thereby address an unmet need of highest significance to public health.