Reconstruction and medical applications of genome-scale metabolic networks

Abstract

Biological systems are inherently robust against various internal and external perturbations. This in-trinsic property of biological systems becomes very important in medical field, especially treating both infec-tious and cancer diseases as problematic microbial pathogens and cancer cells destroy homeostasis and ro-bustness of our human body, and they are robust against chemical and/or physical therapeutic perturbations. Studies conducted in this thesis started with this medical problem, and specifically aimed at elucidating meta-bolic phenotypes of disease-causing microbial pathogens and cancerous organ under various conditions for subsequent drug targeting. Fundamental technology that underlies all the studies in this thesis is the recon-struction and simulation of comprehensive genome-scale metabolic networks. In this regard, Chapter 1 intro-duces the concept of metabolic networks and their simulation methodology as a tool for studying cellular robustness. Reconstruction of genome-scale metabolic networks and their simulation with a method called constraints-based flux analysis reveal counterintuitive properties of in silico cell from a holistic perspective that cannot be seen at local level. Constraints-based flux analysis is an optimization-based method, and can only be used when metabolites that constitute the metabolic network are well mass-balanced and consistently connected throughout the network. Only under this completeness of the metabolic network, constraints-based flux analysis successfully predicts intracellular fluxes under conditions of interest. Using this technology, Chapter 2 describes and applies metabolite essentiality analysis to previously reported genome-scale metabolic networks of Escherichia coli, Helicobacter pylori, Mycobacterium tuberculosis and Staphylococcus aureus for drug targeting. Metabolite essentiality analysis is based on constraints-based flux analysis, and predicts essential metabolites whose absence causes cell death. Chapter 3 and 4 describe first ever reconstruction of genome-scale metabolic networks of microbial pathogens Acinetobacter baumannii AYE and Vibrio vulnificus CMCP6, respectively, to study their robustness against various genetic perturbations and predict drug targets using metabolite essentiality analysis. These network models went through rigorous validation process. In particular, predicted essential metabolites were further filtered to select solid drug targets using the criteria of organism specificity and maximal destructive impact on the cellular robustness. In Chapter 5, as a proof-of-concept, final essential metabolites predicted from V. vulnificus in Chapter 4 were used to select only structural analogs out of huge chemical compound library, and screen these analogs for possible identification of antibacterials. This led to the identification of a hit compound that effectively kills V. vulnificus without any structural modification. Chapter 6 shifts its research focus to human lung cancer. In this chapter, genome-scale human metabolic network was reconstructed by additionally considering intracellular compartments through extensive information search. It was, then, simulated with newly developed simulation framework that maximizes the number of reactions whose activity status is consistent with metabolome data. This work led to the large-scale comparative study of normal and cancerous lung at metabolic level. Chapter 7 discusses current status of other biological networks, including transcriptional and protein interaction networks in addition to metabolic network with respect to their role of data integration and analysis. Finally, Chapter 8 summarizes the studies conducted in this thesis, and suggests the expected future role of the techniques developed herein that could contribute to drug discovery pipeline.