(466e) Model-Predictive Control Strategies for Improved Bioprocess Performance

Glennon, B., University College Dublin
Whelan, J., University College Dublin

There is an increase in interest within the biopharmaceutical industry to move from a quality-by-inspection to a quality-by-design (QbD) approach towards process design, optimisation and operation. The implementation of this philosophy through the use of a PAT-enabled control strategy for a bioreactor is decribed in this work.  In this study, a number of elements were developed and integrated to implement advanced feedback control of substrate concentrations in a fed-batch CHO cell bioprocess. It is desirable to control substrate and by-product concentrations to provide better process performance in terms of cell density, culture longevity and protein quantity and quality.  To implement such control, a Raman spectroscopy method for the simultaneous, real-time measurement of cell density, glucose, glutamine, glutamate, lactate and ammonia was developed.  Secondly, a first-principle engineering model of the fed-batch process was identified which facilitated process control simulations.  The simulations identified model predictive control (MPC) as a promising form of process control for the inherently complex and highly variable nature of bioprocesses due to its ability to reject measurement noise, handle long sample intervals, cope with non-linear processes and operate as a multiple-input-multiple-output (MIMO) control strategy.  Finally, MPC control of glucose and glutamine concentrations was successfully implemented on 3 L and 15 L bioreactors.  In transitioning from a bolus to continuous feeding strategy, a 30 % increase in cell density was achieved.