PROJECT SUMMARY / ABSTRACT The opioid crisis is a major public health problem in the United States. Over the past two decades, opioid use and abuse have increased dramatically, with over 5 million people in the United States using prescription analgesics without medical need or prescription. This has resulted in a significant increase in opioid-related deaths and addiction rates, with the crisis having a profound impact on individuals, families, and communities. The proposal aims to develop machine learning-based predictive models for opioid use disorder (OUD) leveraging genomic, social, and clinical factors. The project will utilize the diverse and equitable AllOfUs database to identify novel genomic markers associated with OUD in patients with and without co-existing pain conditions. A significant advantage of the AllOfUs database is the diversity of the patient population and clinical samples – over 50% of the population is considered underrepresented. This will be achieved through genome- wide association analysis to identify novel single nucleotide variants, copy number variants, and/or structural variants. The project will also use machine learning techniques to develop predictive models that classify the risk of OUD, integrating various data types such as clinical factors, social factors, and genomic data. The project aims to identify key features that aid in the development of improved models for predicting the risk of OUD. The first specific aim of the proposal is to identify associations between genomic profiles and OUD. The project will focus on patients with or without co-existing pain conditions and identify novel genetic markers associated with OUD in each of these unique patient populations. The second specific aim is to develop predictive models using machine learning techniques to classify the risk of OUD. The models will integrate social, clinical, and genomic data to provide clinicians with a tool to risk stratify their patients. The project aims to develop robust machine learning-based models predicting OUD and visualize the individual features' impacts on model performance to provide understanding of which factors are most impactful to predicting the outcome.