Computational analyses of logic and mechanism in regulatory networks

Gene regulation is a key part of the information-processing capacity of the cell, and genomic transcription looms large as the initial, often limiting, step leading to many condition-specific phenotypes. My work attempts to address several questions on the origins of this phenomenon in the logical and combinatorial structure of gene regulatory networks.;The first question is how combinations of transcription factors (TFs) compose the 'logical' functions controlling genes and gene modules. A discrete probabilistic method was designed to infer condition-specific logical substructures within physical networks of TFs and genes, based on large-scale measurements of expression. This computational approach facilitates examination of how different types of biological function relate to the combinatorial organization of the Saccharomyces cerevisiae regulatory network.;A variety of effects were found to influence the logical structure of the yeast regulatory program. For example, classes of functions such as AND and OR involving yeast TFs show non-random patterns of preference and avoidance in different combinations, patterns which also correlate with spatial distributions of TF binding sites in many gene promoters. Inferred logical patterns also tend to correspond to conventional distinctions of cooperativity and additivity among multiple TFs. Further, the large-scale distribution of logical functions in the yeast network closely matches empirical distributions of similar functions in stable random Boolean networks, suggesting how the larger network is assembled from smaller functional components under the constraint of stability.;Combinatorial organization appears to explain only part of the condition-specific structure of the yeast regulatory network, however, which is also shaped by the content and affinity of DNA sequences to which TFs bind. Here, a relationship between cis binding site content and TF expression often appears to obey a simple scheme in which genes having cis sites with relatively low TF affinity are modulated at relatively higher levels of corresponding TF. This partitioning of different functional gene repertoires, based on site affinity and TF level is interestingly most dominant in cases where genes themselves aren't regulated by combinations of TFs, which suggests the complementary roles that combinatorial activity and binding specificity may play in the structure of regulatory networks.