Systems biology and chemoinformatics methods for biomining and systems metabolic engineering applications

Abstract

In the first chapter, this thesis aims to demonstrate the great potential of Constraint-Based Reconstruction and Analysis (COBRA) methods for studying and predicting specific phenotypes in the bacterium Acidithiobacillus ferrooxidans. A genome-scale metabolic reconstruction of Acidithiobacillus ferrooxidans ATCC 23270 (iMC507) is presented and characterized. iMC507 is validated for aerobic chemolithoautotrophic conditions by fixating carbon dioxide and using three different electron donors: ferrous ion, tetrathionate and thiosulfate. Furthermore, the model is utilized for (i) quantitatively studying and analyzing key reactions and pathways involved in the electron transfer metabolism, (ii) describing the central carbon metabolism and (iii) for evaluating the potential to couple the production of extracellular polymeric substances through knock-outs. The second chapter work outlines the effort towards advancing the field of systems metabolic engineering by using COBRA methods in conjunction with chemoinformatic approaches to metabolically engineer the bacterium Escherichia coli. A complete strain design workflow integrating synthetic pathway prediction with growth-coupled designs for the production of non-native compounds in a target organism of interest is outlined. The generated enabling technology is a computational pipeline including chemoinformatics, bioinformatics, constraint-based modeling, and GEMs to aid in the process of metabolic engineering of microbes for industrial bioprocessing purposes. A retrosynthetic based pathway predictor algorithm containing a novel integration with GEMs and reaction promiscuity analysis is developed and demonstrated. Specifically, the production potential of 20 industrially-relevant chemicals in E. coli and feasible designs for production strains generation is outlined. A comprehensive mapping from E. coli s native metabolome to commodity chemicals that are 4 reactions or less away from a natural metabolite is performed. Sets of metabolic interventions, specifically knock-outs and knock-ins that coupled the target chemical production to growth rate were determined. In the third chapter, in order to aid the field of cancer metabolism, potential biomarkers were determined through gain of function oncometabolites predictions. Based on a chemoinformatic approach in conjunction with the global human metabolic network Recon 2, a workflow for predicting potential oncometabolites is constructed. Starting from a list of mutated enzymes genes, described as GoF mutations, a range of promiscuous catalytic activities are inferred. In total 24 chemical substructures of oncometabolites resulting from the GoF analysis are predicted.