Fig. 4
From: A machine learning approach to predict metabolic pathway dynamics from time-series multiomics data
Limonene pathway kinetic Michaelis–Menten model. This kinetic model was compiled from sources in the BRENDA database along with guidance from Weaver et al.93 This system is composed of ten nonlinear ordinary differential equations, which describe the concentration for each metabolite in the pathway (see Supplementary Material for details). The dynamics of this model are rich and complex enough to pose a significant challenge to be predicted through machine learning. This model is used in this work to: (1) compare its predictions with machine learning predictions, and (2) generate simulated data sets to check scaling dependencies with the amount of time series used for training of machine learning algorithms. The method presented in this paper focuses on substituting these Michaelis–Menten expressions by machine learning algorithms (see Supplementary Fig. S1). Kinetic constants were left as free parameters when fitting experimental data in Fig. 6
