A computational framework to elucidate regulatory networks: A case study with secondary metabolism in Streptomyces coelicolor

Gene regulatory networks determine the functional state of a cell. Understanding the topology and dynamics of these networks is essential to gain insight into the working of an organism as well as to harness its biochemical potential. Regulation of secondary metabolite biosynthesis in Streptomyces coelicolor entails an intricate network involving multiple overlapping circuits. The availability of its genome sequence and the feasibility of obtaining non-lethal mutations in practically all genes involved in secondary metabolism make S. coelicolor an ideal system to study regulation. A systems biology approach integrating genome-wide gene expression profiling, informatics and mathematical modeling is taken to study this system. The experimental component includes constructing deletion mutants and performing large-scale gene expression profiling on these mutants. The computational framework required to analyze the vast amount of experimental data is developed in this dissertation. It consists of three parts. First, a methodology based on dynamic time-warping algorithm is presented to analyze time-series data. This procedure aligns the gene-expressions from different cultures to eliminate inherent variations between them. The differentially expressed genes are then identified between different strains to build the first transcriptional regulatory map of S. coelicolor. Next, algorithms based on Boolean analysis are developed that reconstruct network topology from large-scale gene expression data. Compared to existing methods, the algorithms greatly reduce the number of false predictions and result in fewer plausible networks. Finally, the dynamics of regulation are elucidated using mechanistic models. A two-gene signaling system that controls the onset of antibiotic production in S. coelicolor is considered. The parameter range is explored to demonstrate that this system can function as a bistable switch under the control of a threshold concentration of the butyrolactone inducer. For certain other parameter values this system can also exhibit transient dynamics and oscillations in the expression of the two genes. The dissertation also presents a comprehensive model for real-time polymerase chain reaction, one of the most sensitive techniques for estimation of mRNA transcript levels. The model is used to study the effect of varying operating conditions on the efficiency of this process and identifies procedures that maximize this efficiency.