UC Irvine

Gene expression is a tightly controlled process in all cells and at all stages of life. Expressing the wrong gene at the wrong time in the wrong cell can be deadly to an organism and is one of the hallmarks of disease. The primary point of control of gene expression is transcriptional regulation, which is the process gating the transcription of a gene into mRNA. This process is largely controlled by protein-DNA interactions where specific proteins recognize a specific DNA sequence potentially in combination with other proteins in order to bind to that location and either recruit or repel the general transcriptional machinery. These protein-DNA interactions can be abstracted into connections on a gene regulatory network (GRN) for visualization. GRNs have been drawn for many cellular functions from the bottom-up, in which each interaction is exhaustively studied one-at-a-time, representing months or years of work. In this thesis, I present two works that build these networks from the top-down with self-organizing maps using (1) single cell gene expression and single cell chromatin accessibility drawn from a mouse pre-B cell differentiation system and (2) a large dataset of bulk functional genomics assays of mesendodermal development in Xenopus tropicalis. The resulting networks not only recapitulate known interactions, but they also introduce a large number of new potential regulatory connections for each system. Finally, I present a process for performing this analysis on a growing dataset with iterative releases without requiring a full re-classification. In all, the results of this work provide a novel way to study the regulation of gene expression using integrative analysis of large functional genomics datasets.