Exploring pathophysiology of neuropsychiatric disorders based on multi-omics analysis and deep learning

Abstract

Despite of its importance and severity, the pathophysiology of neuropsychiatric disorders remains in hidden. Recent advances of analytic technologies lead mass production of high-quality multi-omics data and data-driven research. In this dissertation, we developed data-driven methods to explain the pathophysiology by combining data analysis including deep learning and suggested new aspects of neuropsychiatric disorders. First, we developed novel network comparison method for gene coexpression network. We applied new method to comparative analysis between Huntington’s disease (HD) and brain aging and found that HD and aging have common aspects in gene expression related to neurodegeneration and immune response. Second, we developed step-wise deep learning with multi-precision data technique and constructed an interpretable network model between genotypes and schizophrenia. We found novel combinatorial SNP marker of schizophrenia and proposed relationship between neuronal growth and schizophrenia susceptibility by analyzing the model. I expect that data-driven approaches to neuropsychiatric disorders like this dissertation will expand our understanding on the diseases.