PROJECT SUMMARY Schizophrenia (SCZ) and bipolar disorder (BD) are highly heritable, severe and complex brain disorders characterized by substantial clinical and biological heterogeneity. Despite this, case-control studies often ignore such heterogeneity through their focus on the average patient, which may be the core reason for a lack of robust biomarkers indicative of an individual’s treatment response and outcome. Although they are classified as independent diagnostic entities, SCZ and BD are highly genetically correlated, exhibit high relative risks among relatives of both BD & SCZ patients, and have partially overlapping symptomatology and treatment. In this project we will use tissue and cell-type specific imputed transcriptomes for individuals with SCZ or BD in our VA discovery cohort comprising the Million Veteran Program (MVP) and Cooperative Studies Program 572 (CSP #572, “The Genetics of Functional Disability in Schizophrenia and Bipolar Illness”), as an intermediate molecular phenotype, to identify, characterize and target subphenotypes of these disorders. Findings from the VA discovery cohort will be validated in the PsycheMERGE and BioMe cohorts. First, we will impute tissue and cell-type specific transcriptomes for all individuals with schizophrenia (SCZ) or bipolar disorder (BD) in the VA discovery cohort. To achieve this, we will train tissue (brain and peripheral tissues) and cell-type (glutamatergic & GABAergic neurons, astrocytes, oligodendrocytes, and microglia from DLPFC) specific EpiXcan transcriptomic imputation models at the gene and isoform level. Secondly, we will use the imputed transcriptomes as an intermediate molecular phenotype to identify genetically-regulated gene expression (GReX) based subpopulations and within them the key molecular drivers using deep neural networks (DNNs). Lastly, we will identify key non-genetic biomarkers and effective treatments for each validated subphenotype. Non-genetic biomarkers will be based on pre-mined features available from the electronic health records (EHR) and features extracted from the EHR via natural language processing (NLP). The subphenotypes will be validated in the civilian cohorts PsycheMERGE and BioMe. This project will take place at the Icahn School of Medicine, one of the leading centers of data science, genomics and precision medicine. The mentoring committee comprises experts in the fields of computational and functional genomics, integrative analysis, machine learning (including DNNs and NLP), and EHR mining. Dr. Voloudakis will develop the skills necessary to launch an independent academic career in genetically based EHR-informed precision psychiatry.