A model based approach for monitoring Bordetella pertussis fermentation with an inline spectro-fluorescence probe

Presently, the extraction of the antigen pertactin poses a challenge in the manufacturing of the whooping cough vaccine due to its low and variable yield [1]. In this work, a hybrid model that combines empirical and mechanistic parts and in-line fluorescence measurements is used to design an estimator for monitoring the manufacturing process in bioreactors. The empirical part of the hybrid model uses Partial Least Squares (PLS) regression to estimate biomass, carbon source, and pertactin productivity from fluorescence data. In view that significant correlations are observed between oxidative stress and productivity, the mechanistic part of the hybrid model is based on key oxidative reaction pathways. Estimation based on a hybrid model is shown to improve the prediction accuracy of antigen productivity as compared to purely empirical or purely mechanistic model-based estimators. The proposed estimator enables real-time monitoring of the manufacturing process and opens the possibility of future implementation of mid-point corrective actions. • An online estimator was designed for monitoring pertussis vaccine manufacturing process. • Significant correlations were observed between oxidative stress and Pertactin (Antigen) productivity. • The hybrid model developed is based on key oxidative stress pathways. • Hybrid model has better accuracy than purely mechanistic model and purely statistical model (PLSR).