Development of rule-based models for regulatory mechanisms of gene expression and bacterial metabolism

Abstract

Systems modeling is a wide field in science and engineering aimed to understand how the components of a system evolve through time and space. For centuries, modeling has been used for the description of a variety of simple systems and with the advent of computation, we were able to model highly detailed systems, analyze them, and propose hypotheses for further experimental testing. In this doctoral thesis, the utilization of rule-based models was proposed for the rapid development of draft models describing the regulation of bacterial gene expression. The developed methodology is able to model transcription, translation, and the degradation of macromolecules as well the bacterial metabolism. The aforementioned processes are essential for cell viability, albeit modeling is widely adopted for metabolism. To circumvent the time-consuming development of models, the modeling proposition is accompanied by the development of computational tools to automatically model each process, a technique previously available only for metabolic models. Regulation of gene expression is essential for cell homeostasis and adaptation. This regulation relies on transcription factors and other proteins that bind specific DNA sequences and control genetic programs. However, the complexity of this regulatory network precludes efforts to model gene regulation at a genome-scale. In the first place, we propose a methodology build upon the Kappa Biobrick Framework, which was automated and extended to describe correctly the genome architecture, initiation of bacterial transcription, and incorporate metabolism. We developed Atlas, a software that is able to convert the many interactions and reactions encoded in biological networks in rule-based models. Rules are similar to chemical equations describing only the characteristics involved in a reaction. In doing that, a rule could represent thousands and even millions of individual reactions. The method was employed with known gene regulation data and metabolic reactions of the bacterium Escherichia coli. Later, we developed Pleione, a software that employs a genetic algorithm to calibrate rule-based models. The tool gives support to four stochastic simulators (compatible with two rule languages). Pleione distributes simulations and calculations of the goodness of fit in high-performance computing infrastructures, and more importantly, harness equivalence statistical tests to determine the pertinence of stochastic simulations to experimental data. We also developed tools to estimate parameter uncertainty of calibrations and to estimate sensitivity indexes of user-selected parameters. Modeling is viewed as a specialized task, inaccessible without proper knowledge of modeling frameworks. Atlas takes inspiration on available tools to reconstruct draft Genome-Scale Metabolic Models, and this thesis satisfactorily presents a software library to develop and analyze rule-based models for gene regulation and metabolism in bacteria. The developed tools allow the assess the impact of modifications like gene copy number, operon architecture, and other common genetic modifications to understand bacterial physiology, pathogenicity, and eventually, the engineering of bacterial cells for biotechnology and biomedicine applications. The tools presented in this doctoral thesis are open-source and freely available for download from the Python Package Index and from github.com/networkbiolab/pleione and github.com/networkbiolab/PythonCyc.