Bayesian cell therapy process optimization.

Area of Science:

• Biotechnology
• Process Engineering
• Cell Therapy Manufacturing

Background:

• Optimizing complex bioprocesses, especially in cell therapy, is challenging due to limited biological understanding and resource-intensive experimentation.
• Developing customized, closed, and automated processes is vital for industrial translation and cost-effective patient treatment.
• Bayesian optimization (BO) is an experimental design method, but its application in noisy, parallel bioprocess optimization requires further investigation.

Purpose of the Study:

• To evaluate the performance of noisy, parallel Bayesian optimization (BO) for bioprocess optimization.
• To assess the impact of increasing noise levels and parallel batch sizes on BO's effectiveness.
• To compare the proposed BO method against the current industry state-of-the-art.

Main Methods:

• In silico evaluation of noisy, parallel BO on two bioprocess models.
• Testing with varying noise levels and parallel batch sizes.
• Application to an in vitro monocyte purification unit operation.

Main Results:

• Noisy, parallel BO significantly outperformed the industry state-of-the-art in silico.
• The method required approximately 50% fewer experiments on average compared to existing approaches.
• Successful application to an in vitro monocyte purification process demonstrated practical utility.

Conclusions:

• Noisy, parallel Bayesian optimization is a powerful tool for optimizing complex bioprocesses.
• This approach offers substantial efficiency gains in cell therapy process development.
• The findings support the adoption of advanced BO techniques for industrial bioprocessing.