Abstract (eng)
Metabolic models serve a dual purpose in understanding biological systems: (1) decoding mechanistic relationships after experimental data acquisition; and (2) acting as predictive tools for experimental design. In my thesis, I showcase both aspects in two key applications demonstrating the versatility of metabolic modeling: (1) normalization of the finger sweat metabolome measurements to enable a quantitative analysis for clinical applications; (2) designing an optimal industrial production process for plasmid DNA production. Up to date, finger sweat normalization has been a challenge as the sweat rate of participants cannot be controlled for and is hard to measure directly. As a case study on caffeine was conducted by my experimental collaborators, I developed a metabolic model that included the absorption, conversion, and elimination of caffeine in the human body as well as a term representing the mechanism of sweating. By fitting the experimental data onto the developed model, we were able to estimate personalized kinetic constants and showed that they shift little over time. In a follow-up study, I further improved the goodness of normalization by adding a previously published statistical normalization method on top of the metabolic model. Simulations and case studies of the combined model showed promising results for the quantification of time series measurements of biomarkers in the finger sweat and other body fluids with size effects. In the plasmid DNA production project, I used metabolic models for medium design of an industrial production process. Counterintuitively, I found that the partial removal of an essential medium component, namely sulfate, can lead to improved productivity and specific yield. The optimal concentration of sulfate in the medium was predicted with dynamic simulations using a genome-scale metabolic model of Escherichia coli. Validation experiments conducted by experimental collaborators indeed confirmed the theoretical predictions. We hypothesize that this strategy has high future potential as it is predictions are easily convertible to other biomolecule production processes. In conclusion, my thesis demonstrates the multifaceted utility of (dynamic) metabolic models in elucidating and predicting biological phenomena. Spanning scientific disciplines, from analytical chemistry to biotechnology, these models offer invaluable insights and hold the key to transformative advancements.