An integrated computational approach for the determination of metazoan cis regulatory codes and cis regulatory modules

Modulation of metazoan gene expression occurs largely through the combinatorial interactions between transcription factors and their target cis regulatory DNA sequences. These interactions have evolved to choreograph transcriptional levels in response to environmental stimuli and according to developmental stage and cell type, allowing complex and variable expression patterns. For a limited number of model genes, the mechanisms of gene regulation are well understood; the regulating transcription factors and their DNA binding sites are known, as are the resulting gene expression patterns. For the vast majority of metazoan genes, however, a functional understanding of the subsets of transcription factors responsible for generating a given spatio-temporal expression domain is lacking. In this thesis, I present an integrated approach for computationally mapping transcription factors to their target genes that can be applied in a metazoan setting. This is comprised of three distinct, yet related projects. First, I present a design for a universal protein binding microarray that can be used to comprehensively characterize the DNA binding preferences of transcription factors with unprecedented resolution and throughput. This design utilizes a mathematical object borrowed form algebraic combinatorics called a de Bruijn sequence that is able to represent a maximum number of sequence variants within a sequence of minimum length, allowing all candidate binding sites of a given length to be placed onto a single array. Second, I have developed a suite of computational tools that integrate information on the DNA binding specificities of transcription factors with information on gene expression and function in order to systematically map transcription factors to their target genes. This is achieved by quantifying the enrichment of motifs and combinations of motifs within independently defined gene sets, allowing an in silico screen for cis regulatory codes and cis regulatory modules. Finally, I have applied these tools to identify cis regulatory codes and cis regulatory modules in specific biological systems, including the development of the embryonic Drosophila mesoderm and the differentiation of human myoblasts into myotubes. In each of these systems, novel cis regulatory modules were computationally predicted and subsequently experimentally verified, yielding new biological insights.