Abstract
Industrial fermentation continually improves biological process control for a wide range of microorganisms used in multi-billion-dollar industries including industrial enzymes, pharmaceuticals, foods, beverages, commodity chemicals, and bioenergy. In the case of recombinant protein production, batch and fed-batch phases of fermentation are usually followed by an induction phase, where chemical or thermal induction initiates the expression of a target protein. Fed-batch processes are usually automated, whereas "out-of-the-box" distributed control systems (DCS) are often unable to define the threshold for induction and respond accordingly. The present study demonstrates the integration of optical density (OD) process analytical technology (PAT) and Lucullus®, a process information management system (PIMS), to enable end-to-end automated fermentation at bench and pilot scale. Data aggregated from tens of fermenter runs and hundreds of offline training measurements enabled the development of an accurate multivariate model to predict OD in real-time. This eliminated the requirement to generate offline correlation models for each OD probe, allowed for model transfer, and incorporated additional predictor terms such as antifoam usage. Automating the induction phase enabled end-to-end fermentation, reducing labor and operational costs while increasing yield through higher reactor utilization within the same time period.