UC San Diego

A fundamental challenge in human genetics is translating genetic associations into knowledge of the biological mechanisms underlying complex traits. By linking genetic associations to genes, interactions between genes and proteins, and global cellular pathways and networks, network biology can provide a systems perspective on this genotype-phenotype translation. Such network-based approaches have demonstrated success in generating novel biological insights and have been applied to prioritize disease genes, identify functional modules, and interpret genome-wide association studies (GWAS). However, challenges remain in adapting network-based integration across different biological contexts and in understanding the role of network selection. In this dissertation, I address these challenges by optimizing a network approach for assessing the convergence of diverse genetic data and systematically assessing the completeness, quality, and performance of existing network resources. First, I show that cross-species translation of GWAS can be greatly improved using a network colocalization approach, demonstrating that disparate genetics in humans and rats converge on a conserved molecular network. Analysis of this conserved network reveals that neuronal and hormone signaling pathways are critical shared mechanisms underlying the regulation of body mass index in humans and rodents. Second, I examine 45 current human network resources to understand differences in network contents and structure, and the impact of these differences on the network-based analysis of genetic data. This study provides a benchmark of networks across different applications and clarifies the factors that influence network performance. Finally, I systematically analyze common and rare variant associations for 373 complex human phenotypes to understand whether their impacts manifest through the same effect genes and molecular mechanisms. While the common and rare variant studies implicate few shared genes, I show that they converge on shared molecular networks for more than 75% of traits. Taken together, these studies provide a foundation for utilizing networks to understand complex genotype-phenotype associations. This research enhances our ability to decipher the genetic underpinnings of complex traits and diseases, an important step towards personalized medicine and improved health outcomes.