Predictive accuracy of the DEB model: forecasting Escherichia coli growth and its uncertainty following substrate shifts

Efficient control of industrial bioprocesses requires robust mathematical models capable of predicting microbial dynamics under different environmental conditions. This study evaluates the Dynamic Energy Budget (DEB) theory for predicting Escherichia coli metabolism during sudden transitions between different carbon sources. We focused on assessing the sensitivity and identifiability of model parameters to determine substrate-specific factors and evaluate predictive limits. The DEB model demonstrated high in-sample precision and accuracy. The observed decrease in out-of-sample prediction accuracy was most likely caused by an accumulating systematic error. This error arose from the emergence of unmodeled factors during the final stage of the experiment, which can be mitigated by either preventing these factors experimentally or explicitly incorporating them into the model. Consequently, leveraging its mechanistic foundations derived from mass and energy balances, the DEB model provides application-useful accuracy even with limited data, showing great potential to be utilized as a Soft Sensor within Digital Twin technologies for bioprocess monitoring.