Integrative Systems Biology: Elucidating Complex Traits

Risk-phenotypes and diseases are oen caused by perturbed cellular networks, as biological processes depend on an overwhelming number of heavily intertwined components. e impact of a genetically altered gene may ripple through its molecular neighborhood instead of being confined to the gene product itself. My doctoral studies have been focused on the development of integrative approaches to identify systemic risk-modifying and disease-causing patterns. ey have been rooted in the hypothesis that data integration of complementary data sets may yield additional etiologic insights compared to analyses conducted within a single type of data. e first line of research presented here outlines two integrative methodologies designed to identify etiological pathways and susceptibility genes. In Paper I, my coworkers and I present an integrative approach that interrogates protein complexes for enrichment in incident coronary heart disease (CHD) associations from genome-wide association (GWA) data. We show that integration of a moderately powered GWA data with protein-protein interaction (PPI) data successfully identifies candidate susceptibility genes for incident CHD. In Paper II, we present an integrative method that combines heterogeneous data from GWA studies, PPI screens, disease similarities, linkage studies, and gene expression experiments into a multi-layered evidence network, which can be used to prioritize the protein-coding part of the genome according to a particular indication. We applied the method to bipolar disorder and type  diabetes, and validated it by replicating a single-nucleotide polymorphism (SNP) within a novel bipolar disorder susceptibility gene. Next, I present the avenue of my research that has been focused on the analysis of genetic variation in obesity. In section ., I outline results from our bioinformaticsbased analysis of the FTO locus. Genetic variation within the FTO locus provides the hitherto strongest association between common SNPs and obesity, but the mechanisms leading to this association are still unknown. In Paper III, we demonstrate that body-mass index associated gene products coalesce onto distinct protein complexes, and show that these putative risk modules incriminate novel candidate obesitysusceptibility genes. e last overall line of research presented here, provides examples on how networks of human metabolism may serve as a data integration framework for differential gene expression data. In Paper IV, we present a method that can be used to identify metabolically-related sets of enzymes, which exhibit modest but concordant changes in gene expression. In Paper V, we used that approach to identify metabolites as biomarkers for weight maintenance upon dietary-induced weight loss. e approaches presented in this PhD esis provide integrative methodologies for the aggregation of multiple, functionally relevant data types. Together they represent a novel bioinformatics-based toolbox for analyses of genetic variation in human traits and disease. e esis is structured as follows. Chapter  presents a few introductory remarks to integrative systems biology, and Chapter  gives a brief description of human genetic variation and GWA analysis. Chapters - present the main topics in the esis (integrative methodologies for the analysis of GWA data, integrative analyses of genetic variation in obesity, and integrative analyses based on metabolic networks). Chapter  summarizes the esis with a few concluding remarks.