On-Line Reoptimization of Mammalian Fed-Batch Culture Using a Nonlinear Model Predictive Controller

Fed-batch culture is widely used in biopharmaceutical production owing to its superior productivity; however, optimizing feeding trajectories remains a challenge. In this study, we investigated the feasibility and benefits of using a nonlinear model predictive controller (NLMPC) for on-line reoptimization in mammalian fed-batch culture to compensate for process-model mismatch (PMM). We simulated a monoclonal antibody production process using a standard kinetic model and deliberately introduced PMM via parameter errors. The NLMPC optimized feeding trajectories for a single-feed case, in which a mixture of glucose and glutamine is fed, and for a multiple-feed case, in which glucose and glutamine are fed independently. Our results demonstrate that on-line reoptimization successfully compensates for PMM, improving the final product mass compared to off-line optimization. This study highlights the potential of on-line reoptimization using NLMPCs in mammalian fed-batch culture, which can enhance product yield even in the presence of insufficient parameter estimation.

Automated summary

This study investigates the feasibility and benefits of using NLMPC for online reoptimization in biopharmaceutical production, compensating for process-model mismatch. The results demonstrate that online reoptimization can improve the final product quality.