RESOLVING THE CHALLENGES OF BIG-DATA IMAGING GENETICS ANALYSIS TO UNDERSTAND GENETIC AND ENVIRONMENTAL RISK FACTORS IN PSYCHIATRIC DISORDERS

Worldwide efforts have led to large and inclusive imaging genetics datasets enabling examination of the contribution of genetic and environmental factors to development, clinical course and treatment effectiveness in psychiatric disorders. These datasets combine high-resolution neuroimaging and genetic data in large and inclusive samples. Classical genetic analyses can help to parse the variance in disorder-related brain patterns into additive genetic, specific SNP, household and environmental causes. Performing these inquiries at full imaging and genetic resolution is a formidable computational task where the computational complexity of classical genetic analyses rises as a square or cube of the sample size. We describe fast, non-iterative simplifications to accelerate classical variance component (VC) methods including heritability, genetic correlation, and genome-wide association in dense and complex empirical pedigrees derived in samples such as UKBB, HCP and ABCD. These approaches linearize computational effort while maintaining approximation fidelity (r∼0.95) with VC results and take advantage of parallel computing provided by central and graphics processing units (CPU and GPU). We show that the new approaches can help to tract the nature vs. nurture interaction in the development of major depressive disorder and psychosis in the longitudinal datasets such as ABCD. We also show how specific genetic risk factors for Alzheimer disease can interact with environment leading to development of brain patterns that are predictive of the risk of development of dementia.