PROJECT SUMMARY Given that anorexia nervosa (AN) has the highest mortality rate of any psychiatric illness and current treatments show limited efficacy, there is a crucial need to better understand the brain mechanisms driving the pathophysiology of this disorder. This proposal combines an experimental medicine approach focused on gastrointestinal (GI) interoception with computational modeling to probe neural circuits of interoception and appetite-related gastric processing in AN. The goal is to identify perceptual and neural markers for AN at the individual patient level and apply machine learning methods to test clinical outcomes prediction longitudinally. Supported by our preliminary data, this proposal is based on the premise that the pathophysiology of AN includes a computational dysfunction manifested by cognitive suppression of the expected precision of afferent interoceptive signals associated with hunger, which reduces their motivational influence and facilitates maladaptive and avoidant eating behaviors. We propose a case-control study with 65 AN and 65 healthy comparisons who will undergo extensive baseline testing using a novel GI interoception probe during measurement of symptoms, behavior, circuits, and physiology. Sensory stimulation will occur during the premeal period, anchoring responses to an anticipatory context with high relevance to the disorder. These individuals will be followed for 180 days to examine clinical outcomes. A computational approach will examine the basic hypothesis that AN individuals have lower sensory precision for GI interoception and that the degree of sensory imprecision is related to clinical characteristics. Moreover, we will examine the relationship of this imprecision to circuit and physiological measures. We will then apply machine learning approaches to these neurophysiological and perceptual measures to longitudinally test the prediction of clinical outcomes. Achieving the aims of this project will provide unique insights into the pathophysiology of AN by arbitrating whether AN is a consequence of ?top-down? or ?bottom-up? dysregulation in the nervous system, which could transform our understanding of how intrinsic interoceptive disturbances lead to AN. Pragmatically, it will result in new technologies for identifying interoceptive dysfunction at the individual level, allowing psychiatry to develop diagnostic and predictive biomarkers of AN. Thus the neurocomputational assay of gastrointestinal interoception in AN could be used to develop low-cost, scalable, and objective tools for identifying dysfunction in individual patients, to facilitate neurobiologically-based definitions of recovery, and to predict the risk of relapse following treatment. Finally, this proposal lays the groundwork for the future development of precision psychiatric interventions such as perceptual retraining therapies to target (and recalibrate) abnormal brain- body interactions.