Grammar-guided symbolic regression as tool for model development of enzymatic reactions

The transition of the chemical industry toward renewable biocatalysis requires robust kinetic models to scale and optimize enzymatic processes. However, deriving these models typically relies on manual, trial-and-error fitting of predefined equations, while data-driven machine learning approaches often produce accurate but physically meaningless “black-box” models. Here, we present an adaptive kinetic model evaluation framework that automates the discovery of enzymatic rate laws directly from batch progress curves. By employing Grammar-Guided Symbolic Regression (GGSR) the algorithm autonomously constructs rate equations using only physically permissible kinetic building blocks. The framework’s reliability was first validated against 13 synthetic kinetic mechanisms, confirming structural and parametric identifiability. The pipeline was then applied to a real experimental case study of glucose-fructose isomerization using Streptomyces murinus isomerase. The algorithm successfully recovered a reversible kinetic equation featuring an apparent uncompetitive-type interaction term providing specific activities and Michaelis constants that align well with literature. This work demonstrates that GGSR can bridge the gap between data-driven automation and mechanistically interpretable biochemical engineering.