| Summary: | Complex networks are an important tool for the study of biological data. There are two main aims in this data-driven work, which are explored in tandem. We study (1) the nature of schizophrenia and (2) utility in novel additions to traditional network based spectral clustering methods. More specifcally, we explore three facets of schizophrenia. First, we study functional brain data in animal models of relevance to the condition. Second, we examine the impact of antipsychotic medication on gene expression in humans, and third we assess whole blood for potential as a suitable alternative to brain tissue. With regard to spectral clustering, we employ the Singular Value Decomposition and the Generalized Singular Value Decomposition in a way that allows us to incorporate additional information into the clustering problem. This work is of interest in the life sciences due to the complex heterogeneous nature of schizophrenia, which has created desire for analysis of large amounts of data. In addition, development of network based approaches is a timely area of study in general given recent explosions in the amount of data produced across many subject areas. Our interdisciplinary work leads to four main conclusions: (a) network approaches for functional brain animal model studies can produce results that are biologically meaningful in humans, (b) a novel node-weighted version of the Laplacian is a flexible tool that allows multiple sources of network information to be combined, (c) antipsychotic medication, used routinely to treat schizophrenia, has a dominant effect on gene expression as compared to the control state, masking the underlying nature of the disease and (d) human whole blood is useful for the study of gene expression in schizophrenia.
|
|---|
