Multi-information source Bayesian optimization of culture media for cellular agriculture

Area of Science:

• Biotechnology
• Cell Culture Optimization
• Bioprocessing

Background:

• Optimizing culture media for industrial bioprocessing and cellular agriculture is hindered by complex design spaces and a lack of robust mathematical models.
• Current methods rely on inaccurate rapid growth assays or time-consuming cell counting, limiting experimental efficiency.

Purpose of the Study:

• To develop and validate a novel Bayesian optimization algorithm for efficient cell culture media optimization.
• To improve cell yield and experimental efficiency in bioprocessing and cellular agriculture.

Main Methods:

• Utilized a multi-information source Bayesian optimization algorithm with an uncertainty-weighted desirability function.
• Employed murine C2C12 cells and a combination of AlamarBlue, LIVE stain, and trypan blue exclusion assays for cell number determination.
• Compared the optimized medium against a commercial variant and a design-of-experiments method.

Main Results:

• Designed a cell culture medium yielding 181% more cells than a commercial alternative at a similar cost.
• Achieved this optimization in 38% fewer experiments compared to an efficient design-of-experiments approach.
• The optimized medium demonstrated robust performance across multiple cell passages.

Conclusions:

• The multi-information source Bayesian optimization algorithm significantly enhances cell culture media design.
• This approach offers a more efficient and robust method for optimizing media for bioprocessing and cellular agriculture compared to traditional assays.
• Improved media optimization can accelerate advancements in cellular agriculture and industrial bioprocessing.