Genome-scale metabolic models (GEMs) are mathematical network models that comprehensively encompass all cellular metabolic reactions, enzymes, and metabolites, allowing the in silico modeling and simulation of cell metabolism in a holistic manner using mathematical programming techniques. With the advent of higher-quality GEMs and effective modeling algorithms, GEM analysis has been repeatedly shown to generate accurate predictions and informative hypotheses for metabolism research. In this dissertation, I first describe the comprehensive evaluation and improvement of a GEM algorithm named metabolic transformation algorithm (MTA), which predicts the metabolic targets whose inhibition can facilitate the transformation between two arbitrary cellular metabolic states. The new MTA algorithm is shown to achieve significantly better prediction accuracy than the previous version using a large collection of validation datasets on known metabolic gene knockouts/knockdowns or metabolic protein inhibitions. Next I demonstrate the broad applications of MTA and the GEM analysis framework in both basic and translational research.

First, MTA was used to identify metabolic processes essential to the anti-tumor function of CD8 T cells in cancer immunotherapy. T cells play a central role in cancer immunosurveillance and current cancer immunotherapies. In order to improve the patient response rate to immunotherapy, it is pivotal to understand the factors regulating T cell function. It has been recognized that metabolism can greatly affect different aspects of T cell function, including differentiation, cytokine production, longevity and exhaustion. Via the MTA analysis of adoptive cell transfer (ACT) therapy, including CAR-T therapy datasets, the mitochondrial uncoupling protein UCP2 was predicted as a key determinant of ACT therapy response and CD8 T cell anti-tumor function. GEM analysis also points to the generation of reactive oxygen species (ROS) as a mechanism underlying the effect of UCP2. The role of UCP2 in T cells and the ROS-related mechanisms were thoroughly validated with genetic knockout/overexpression and pharmacologic inhibition experiments in a Pmel-1/B16 ACT mice model.

Second, MTA was used to predict anti-SARS-CoV-2 targets that act via counteracting the virus-induced metabolic alterations. The COVID-19 pandemic caused by the SARS-CoV-2 coronavirus has been ongoing for about a year to date, with still very few therapeutic options available. Viruses, including coronaviruses, are known to hijack the host metabolism to facilitate their own proliferation, making targeting host metabolism a promising antiviral approach. I performed an integrated GEM analysis of 12 published in vitro and human patient gene expression datasets on SARS-CoV-2 infection, revealing a wide-spread metabolic reprogramming induced by the virus. MTA successfully predicted metabolic anti-SARS-CoV-2 targets, with significant overlap with the experimentally validated targets identified from drug and genetic screens. Further, MTA was used to identify potential targets for combining with remdesivir (an approved anti-SARS-CoV-2 drug) to achieve higher efficacy.

In conclusion, I developed the improved MTA algorithm with an updated GEM workflow, which were applied to cancer immunotherapy and coronavirus infection. The integrated GEM analysis identified mitochondrial uncoupling as an essential metabolic process in anti-tumor CD8 T cells, and effectively predicted host metabolism-targeting anti-SARS-CoV-2 strategies. This demonstrates the value of GEM in a broad range of metabolism-related research problems.