Mass Action Stoichiometric. Simulation for Cell Factory Design

For a long time microorganisms have been used to produce beer and bread, and in the last century also molecules such as penicillin and insulin. These same microorganisms
can potentially be used to produce a diverse range of other molecules and contribute to a more sustainable future by reducing our dependency on oil. Producing a given
molecule at a yield high enough to be commercially viable, however, usually requires cell metabolism to be modified extensively. Traditionally, these modifications have
been introduced through random mutagenesis and selection, which has been gradually complemented by more targeted genetic engineering approaches that rely more and
more also on computational models of cell metabolism for target selection. Two main types of models can be used here, stoichiometric models or kinetic models. The former
are easily built at genome-scale and assume the cell to be in a steady-state, giving information only about the reactions’ fluxes, while the latter take into account enzyme
dynamics which makes it possible to model substrate-level enzyme regulation and get information about metabolite concentrations and reaction fluxes over time, although
at the cost of introducing more parameters. Kinetic models have been plagued by the lack of kinetic data.
The focus of this thesis are kinetic models of cell metabolism. In this work we start by developing a software package to create a model ensemble for individual
enzymes in metabolism, where we decompose each reaction into elementary steps, using mass action kinetics to model each step. The resulting rate constants are then
fitted to kinetic data (k_cat, K_m, K_i, etc.). We then use the package as the basis to build a system-level kinetic model. To do so, we take two different approaches, and
in both we drop the assumption that χ_free ≈ χ_tot , i.e. that the total concentration of metabolite in the cell is approximately the same as the free concentration. In both
approaches preliminary results show that the fraction of bound metabolite in the cell is not negligible, with some metabolites having an enzyme-bound concentration
up to 40%. Next, we address the issue of kinetic data scarcity by using molecular dynamics simulations to estimate the difference in binding energies, ΔΔG, between substrate(s) and a given enzyme and product(s) and the same enzyme for a chosen reaction. Here, we show that these in silico determined ΔΔG significantly reduce the
amount of rate constants combinations allowed in each model ensemble. Finally, we combine a kinetic model of glycolysis in Saccharomyces cerevisiae with time-resolved
NMR experiments to study the cellular response to a glucose pulse, and show the model simulations to be in agreement with the experimental results.