Applications of Genome-Scale Metabolic Network Reconstructions to Characterize Drug-Induced Toxicity

Metabolism is the process of deriving energy by breaking down compounds and using that same energy to build new compounds, which is necessary for other biological processes. One way we can model metabolism computationally is with GEnome-Scale metabolic Network REconstructions (GENREs), which use linear algebra to analyze an underdetermined system under steady-state assumptions. GENREs allow us to simulate how the growth of an organism is affected by its surrounding system, as well as how genetic mutations can inhibit normal function of an organism. In the context of human metabolism, GENREs can be used to evaluate how drugs or diseases can impact the function of healthy cells, and potentially lead to the discovery of novel drug targets. In this dissertation I describe what GENREs are and create a simplified GENRE, iSIM, to explain how to make various types of predictions. To make GENREs more accessible to biologists without a programming background, source code is provided on how to run these analyses. For drug-induced hepatotoxicity and nephrotoxicity, I developed a paired transcriptomics data and metabolomics data pipeline to use an existing GENRE, iRno, to make predictions using experimental data and create validation data for model predictions from the same biological sample. This approach identifies why metabolite production levels are changed and can lead to the identification of biomarkers of drug-induced toxicity. Additionally, I predict changes in metabolite level changes with iRno, and compared results to metabolomics data and existing literature. This dissertation provides new tools for biologists who want to learn to program and gives toxicologist and other scientists interested in drug discovery a tool and pipeline to identify potential impacts of new drugs.