PROJECT SUMMARY Cannabis use disorders (CUD) are prevalent in the U.S., and highly comorbid with other substance use disorders (SUD) such as alcohol use disorder (AUD), as well as with other mental health problems. While the etiology of cannabis use/misuse have both environmental and genetic components, cannabis use and problematic use are found to be highly heritable. Thus, studies that identify the genetic risk factors for CUD in the general U.S. populations, and in the high-risk populations, are of high public health importance. However, the genetic factors identified in the human genome thus far by conventional methods are sparse and appear to have only captured a very small fraction of the overall heritability for the disorder. One key challenge in addiction genetics is how to identify genetic interactions and epistatic regulations that may play a more important role in determining risk for addictive behaviors than what gene variants do individually, and that may help explain a critical part of the missing link. Genetic interactions have rarely been systematically considered in studies of substance use, primarily due to lack of statistical power and shortage of computational methodology. To address the challenge, we propose a framework to systematically detect disease-relevant context specific genetic pathway interactions that underlie the risk for SUD. The framework will be applied to CUD and comorbid AUD to identify crucial genetic interactions and pleiotropic interactions, filling a critical gap in uncovering the genetic architectures of CUD. We will leverage genetic network and pathway topology and integrate multiple layers of omics including genomics, transcriptomic and epigenomic signals in drug abuse relevant tissues. By sharpening the focus on the functionally connected gene and regulation subsets through a priori analyses, we will be able to dramatically boost the statistical power to detect genetic interactions, arrive at highly biologically relevant and readily interpretable findings, and potentially provide clinically actionable insights. The proposed study will utilize outcomes from large GWAS studies for CUD and AUD, together with three high-risk population cohorts with elevated levels of severe cannabis and alcohol use disorders that have whole genome sequence data. We will complement the context specific pathway-level interaction analysis with high-dimensional variable screening machine-learning algorithms to identify both low and high order genetic interactions and regulatory epistatic effects associated with CUD. The findings that are carefully validated using independent study cohorts will be incorporated into a larger disease model of CUD for prediction and potential intervention, and will open up new avenues of research by allowing interrogation of the addiction genetics from a system’s level. The framework will be build in such a way that is readily transferable to other SUD and mental health studies and sets the stage...