Identification of somatic deletions in neuropsychiatric diseases using massively parallel sequence data

Abstract

Neuropsychiatric disease is a term that includes all diseases that profoundly alter the behavior and intellectual ability of patients. Despite of significant research in neuroscience, the etiologies and mechanisms of most neuropsychiatric diseases have remained out of reach. Many genomic studies showed the significant contribution of genetic influence in these diseases, however, clear disease relevant genes have not been identified due to their genetic complexity. In recent studies, somatic mutations in brain cells have been shown to affect the onset of several neuropsychiatric diseases. As the role of somatic mutations could not be studied with previous research based on genetic variations, neuropsychiatric disease studies based on somatic mutations are now getting interests for their high potential. However, most of current studies based on somatic mutations were only focused on single-nucleotide variations, and no studies with somatic copy-number alterations, which have larger potential of functional effects compared to single-nucleotide variations, have been reported in neuropsychiatric disease research due to the lack of algorithms to detect somatic copy-number alterations in brain tissue. In this thesis, we identified and validated the existence of somatic deletions, a specific type of somatic copy-number alterations, in neuropsychiatric disease patients, and proposed a computational method to detect somatic deletions in neuropsychiatric disease samples using massively parallel sequence data. First, we identified and validated somatic deletions in a patient with schizophrenia, a specific type of neuropsychiatric diseases. As only a minority of cells contains somatic mutations even in clearly diseased tissue of brain, we developed a novel method to detect somatic deletions by combining the methods for calling deletions based on read-depth and anomaly mapped reads. We successfully identified and validated somatic deletions in schizophrenic samples, and showed the possibility of functional impacts of somatic deletion by the disruption of exonic sequence. We also showed the existence of somatic deletions in normal samples with similar occurrence, but the result of functional analysis showed significant effects of somatic deletions only in schizophrenic cases. Second, we developed an algorithm to detect low-frequent somatic deletions from unmatched samples. To discover somatic deletions without matched control, we built a probabilistic model of somatic deletion distribution based on Gaussian mixture model and estimated parameters using EM algorithm. With a variety of simulation data sets, we successfully validated the performance of developed algorithm. In a comparison with previous algorithms for detecting somatic deletions in cancer samples, our method showed better performance even with the harsh conditions such as the absence of matched control and the case with multiple sub-clones. Finally, we successfully validated somatic deletion candidates called by our algorithm in schizophrenic cases.