Bayesian Optimization for Favourable Suspension Culture of Orbitally Shaken Bioreactors with a Hollow Cylinder Wall

Abstract

Orbitally shaken bioreactors are extensively applied to the mass production of mammalian cells in suspension culture. Commonly used vessels are cylindrical tanks and their main disadvantage is the accumulation of cells at the bottom center. To eliminate this accumulation issue, an annular cylindrical vessel is proposed for orbitally shaken bioreactors using numerical simulation. This approach eliminates the center of the vessel physically and thus removes the possibility of such cell accumulation. The absence of a central section in the vessel significantly improves the dissolved oxygen concentration and cell suspension ratio in the culture medium without increasing the shear stress. It also efficiently removes the possibility of cell aggregation at the bottom. An advanced machine learning technology - Bayesian optimization algorithm is also employed in the hollow orbitally shaken bioreactor to optimize the three main control parameters of the reactor, namely, diameter ratio (di/do), shaking velocity (x), and shaking radius (R). The obtained data related to mass transfer ability, suspension ratio, and shear stress are analyzed. The numerical simulation results show that the “optimal” bioreactor is preferable for cell cultivation compared with the “initial state”, owing to its high capability for suspending cells. It can be concluded that the methodology described in this work is a feasible and reliable tool for performance prediction and process optimization in biotechnology.