Abstract

Kinetic models are useful tools for predicting the dynamic behavior of metabolic systems and to optimize the performance of bioprocesses. However, the difficulty of fitting their parameters to limited data hinders their widespread use. These models have various disparate parameters, are nonlinear, and display complex interactions. To address this challenge, this study introduces a computational framework for constructing bioreactor models integrating detailed enzyme kinetics under data-limited conditions. The framework is illustrated by the construction of a dynamic model that describes the growth kinetics of Saccharomyces cerevisiae and the production of beta-ionone, an apocarotenoid extensively utilized in the flavor and fragrance industries, in batch cultivations. The model was initially formulated and described using 78 free kinetic parameters. Through the systematic application of sensitivity and identifiability analyses and reparameterization, the model was reduced to 9-parameter candidate structures. Multi-criteria decision-making techniques were then employed to select a robust model structure. When validated against an independent experimental dataset with condition-specific cofactor adjustment, the final model achieved comparable overall predictive performance compared to the original model structure, with notable improvements for key pathway metabolites, including up to 38% reduction in the normalized mean absolute error for the target product beta-ionone. This framework successfully addresses the challenge of constructing predictive kinetic models from limited experiments while maintaining mechanistic interpretability, providing a quantitative tool for identifying metabolic bottlenecks and guiding metabolic engineering interventions.