Multiobjective Design of Growth Media with Genome-Scale Metabolic Models and Bayesian Optimization.

Area of Science:

• Biotechnology and Bioprocessing
• Metabolic Engineering
• Computational Biology

Background:

• Optimizing culture media is crucial for efficient and cost-effective cellular production systems.
• Traditional media design methods involving extensive experiments or statistical analyses are time-consuming and expensive.

Purpose of the Study:

• To present genome-scale Multiobjective Bayesian Optimization (gsMOBO) as a general and flexible computational tool for media design.
• To demonstrate gsMOBO's capability in efficiently exploring and optimizing nutrient combinations in high-dimensional spaces for improved biomanufacturing.

Main Methods:

• Integration of genome-scale metabolic models with a top-layer Bayesian optimization loop.
• Application of gsMOBO to identify optimal medium formulations balancing growth, production, and component costs.
• Validation using models of *Escherichia coli* for antibody fragment production and *Bacillus subtilis* for cyclic lipopeptide synthesis.

Main Results:

• gsMOBO successfully identified optimal medium formulations along a Pareto front, balancing key production parameters.
• The method accurately predicted media compositions and trade-offs consistent with existing experimental data.
• Demonstrated applicability in diverse microbial systems (*E. coli* and *B. subtilis*) for different bioproducts.

Conclusions:

• gsMOBO offers a broadly applicable computational strategy for designing cost-effective and productive culture media.
• This approach significantly accelerates the medium development process in biomanufacturing.
• gsMOBO provides a pathway to optimize nutrient combinations for enhanced cellular production.