UCLA

Large-scale transcriptomic datasets provide valuable opportunities to better understand the regulation of gene expression and its role in human health. However, these studies can be confounded by issues such as cell type heterogeneity. Furthermore, these datasets are growing extremely large with complex study designs, such as gene expression measured across a multitude of tissues, that must be accurately and efficiently modeled. Finally, better understanding of the mechanisms that influence gene regulation are required to integrate novel associations with biological understanding. In this dissertation, we introduce methods that address these issues in the analysis of tissue-level gene expression data. We present a method to accurately estimate cell type composition from these data by integrating single-cell information, as well as a scalable approach to model multi-tissue expression datasets and identify expression quantitative trait loci. We also present analyses of bulk expression data that support a hypothesized mechanism of gene regulation that occurs in the general female population through non-random X chromosome inactivation. The work presented in this dissertation allows researchers to perform efficient and accurate analyses of gene expression data and provides additional insight into the mechanisms that underlie associations between genetics, transcriptomics, and complex phenotypes.