Integration of Bayesian optimization and solution thermodynamics to optimize media design for mammalian...

Area of Science:

• Biotechnology
• Bioprocess Engineering
• Machine Learning Applications

Background:

• Optimizing cell culture media is critical for cost-effective biomanufacturing of therapeutics, materials, and food.
• Current optimization methods are complex, time-consuming, and expensive, particularly for mammalian systems like Chinese Hamster Ovary (CHO) cells.
• Long development timelines for therapeutic production contribute significantly to high drug costs.

Purpose of the Study:

• To develop a data-efficient optimization strategy for cell culture media using Bayesian optimization (BO).
• To integrate thermodynamic constraints for ensuring feasible and soluble medium formulations.
• To improve product titers and reduce development timelines in biomanufacturing.

Main Methods:

• Adapted Bayesian optimization (BO) for bioprocess applications, enabling parallel experiments.
• Incorporated thermodynamics-based constraints on media solubility.
• Validated the approach in-silico and in experimental bioreactor settings.

Main Results:

• Achieved improved product titers compared to traditional Design of Experiments (DOE) methods.
• Demonstrated a more data-efficient process optimization strategy by combining machine learning and physical modeling.
• Validated the feasibility of the optimized medium formulations.

Conclusions:

• The developed Bayesian optimization approach offers a significant improvement for bioprocess optimization.
• Integration with robotics-assisted bioreactors can lead to automated bioprocess optimization.
• This strategy promises faster development and lower costs for biotherapeutics.