Development of a Digital-Twin-Based Controller for High-Intensity Fed-Batch Cultivation of Lactobacillus rhamnosus LRH30 Using Dairy-Based Feedstock for Anti-Inflammatory Biomass Production

An artificial neural network (ANN)-based digital twin feed controller, aiming to optimise bioprocess performance, was implemented in fed-batch bioprocesses of Lactobacillus rhamnosus LRH30 for the production of biomass. The online data from several fed-batch bioprocesses in dairy-based cultivations with L. rhamnosus were utilised to develop an ANN that predicts the biomass and growth rate of L. rhamnosus in the process. This growth rate estimator was first validated through unit testing before being applied in model predictive control to select the best control action to drive feed addition to the process. The controller was implemented in fed-batch bioprocesses and the growth rate was estimated in real time, leading to the highest cell dry weight biomass concentrations obtained for L. rhamnosus LRH30 of 36.23 g/L in this study, with growth-rate overshoot and increased feed addition also contributing to the biomass accumulation despite imperfect setpoint tracking. The dose-dependent effect of L. rhamnosus LRH30 samples on cytokine secretion from murine macrophages was further explored, and an anti-inflammatory effect was found at a dosage of 10 mg/mL of bacterial cell dry weight. These experiments demonstrate the first steps in the implementation phase that will ultimately lead to a better controlled bioprocess with higher yields of L. rhamnosus biomass. Future research will focus on the continuous development of the ANN through the incorporation of more datasets, while simultaneously exploring the functional potential of L. rhamnosus LRH30.