Sequential active learning for medium optimization in mAb production.

Area of Science:

• Biotechnology
• Biopharmaceutical Manufacturing
• Cell Culture Technology

Background:

• Monoclonal antibodies (mAbs) are crucial therapeutics for cancer and autoimmune diseases.
• Effective mAb production relies heavily on optimized cell culture media.
• Current optimization methods struggle to integrate experimental biological insights.

Purpose of the Study:

• To develop an active learning strategy for rational cell culture medium design.
• To enhance immunoglobulin G (IgG) titer in Chinese hamster ovary (CHO) cell cultures.
• To integrate machine learning predictions with biologically meaningful experimental observations.

Main Methods:

• Utilized a combination of Design of Experiments (DOE) and two machine learning models.
• Implemented an active learning strategy for iterative medium component adjustment.
• Optimized 44 components in a serum-free medium for CHO cell culture.

Main Results:

• Achieved significant improvement in IgG monoclonal antibody production.
• Successfully incorporated biological insights like osmolality control and amino acid composition.
• Demonstrated the strategy's effectiveness even with limited experimental resources.

Conclusions:

• The proposed active learning strategy offers a practical and effective approach to cell culture medium optimization.
• This data-driven method facilitates rational medium design in biopharmaceutical manufacturing.
• The strategy enables the integration of biological insights into computational optimization processes.