Abstract This project will use PET imaging and machine learning to relate Mu and Kappa receptor levels in Alcohol Use Disorder (AUD) to clinical outcomes during a quit attempt. Over 14 million adults in the US suffer from AUD and there are few effective pharmacotherapies. Rates of lapse and relapse are remarkably high and this remains the biggest hurdle to successful recovery. Development of new treatments must be based on an understanding of the relationship between neurobiology and behaviors involved in recovery from AUD. Existing evidence suggests that the opioid system, through a balance between euphoria (Mu-Opioid Receptors; MOR) and dysphoria (Kappa-Opioid Receptors; KOR), regulates key clinical outcomes (e.g., alcohol craving, anhedonia, and withdrawal) which play critical roles in alcohol lapse/relapse behaviors in AUD. The balance between these OR receptors in healthy animals is disrupted by the consumption of alcohol and the progression of addiction. There is limited data in humans to characterize this disruption. Developing an understanding of the imbalance between MOR and KOR in individuals with AUD, and the associations of this imbalance with craving, mood, withdrawal, and time to lapse (first drink) during a quit attempt, will facilitate the development of tailored therapies to target the imbalance and improve clinical outcomes. PET studies of people with AUD have examined MOR and KOR availability, separately. Previous studies measured MOR in ventral striatum (VS) with an MOR-selective tracer and found higher MOR in VS of people with AUD compared to healthy subjects (HS). We used a KOR-selective PET tracer and observed that people with AUD had significantly lower KOR availability in amygdala and pallidum vs. HS. KOR availability in key regions (e.g., whole striatum) also predicted a reduction in drinking in participants treated with the opioid antagonist naltrexone. It seems that upregulation of MOR and downregulation of KOR occur in the course of AUD and that both are related to clinical outcomes. No one has ever imaged both targets in the same people. We will use PET to image both MOR and KOR availability in HS and AUD. We will quantify the relationships between MOR and KOR of AUD patients, separately and jointly, to key clinical outcomes during their subsequent quit attempt. We will use linear models to relate regional values of binding of each tracer to clinical outcomes. We will also use advanced clustering techniques to identify distinct groups of participants according to their imaging data. Our goal is to understand the interactions of the two major opioid receptor systems and how they encode the behaviors of AUD sufferers during a quit attempt. The results could guide development of new targeted therapies. Machine learning (Spectral Clustering) will also help us identify features in the images that may be useful in the future for prediction of clinical outcomes.