Fig. 2: Overview of the experimental flow set-up and the iterative design of training data to train a kinetic model.
From: Iterative design of training data to control intricate enzymatic reaction networks
a Schematic of the experimental workflow. Enzymes are immobilized on gel beads and placed in CSTR with 6 inlets containing different substrates The output is measured offline on an Ion paired HPLC, 8 species (N = 1, indicated by the arrows, from left to right: uracil, UMP, GMP, adenine, ADP, GTP, UTP, and ATP) can be observed over time. b Computational workflow to design an information dense dataset and train a kinetic model. In step one the OED algorithm evolves control inputs (i.e. inflow rates of the 6 inlets) to be maximally informative. In step two this data is added to a training dataset which is subsequently used to fit a model in step three, resulting in a range of possible parameter values for each parameter (color). In step four we use the previous iteration of the model to predict the outcome of the latest experiment, utilizing this round as test data.
