Analysis of combinatorial gene regulation with thermodynamic models

Transcriptional control is a key regulatory mechanism for cells to direct their destinies. A large number of transcription factors (TFs) could simultaneously bind to a regulatory sequence. With the constellation of TFs bound, the expression level of a target gene is usually determined by the combinatorial control of a number of TFs. The interactions among regulatory proteins and their regulatory sequences collectively form a regulatory network. A major challenge in the study of gene regulation is to identify the interaction relationships within a regulatory network and further to reconstruct gene regulatory networks. In this thesis, we developed an analytical method, Interaction-Identifier, to identify a thermodynamic model that best describes the form of TF-TF interaction among a set of TFs for every target gene. Applying this approach to time-course microarray data in mouse embryonic stem cells, we have inferred five interaction patterns among three regulators: Oct4, Sox2 and Nanog on ten target genes. We further proposed a computational framework, Network-Identifier, utilizing Interaction-Identifier, to reconstruct gene regulatory networks. Applied to five datasets of differentiating embryonic stem cells, Network-Identifier identified a gene regulatory network among 87 transcription regulator genes. This network suggests that Oct4, Sox2 and Klf4 indirectly repress lineage specific differentiation genes by activating transcriptional repressors of Ctbp2, Rest and Mtf2