Bioprocess optimization under uncertainty using ensemble modeling.

Area of Science:

• Biotechnology
• Process Engineering
• Mathematical Modeling

Background:

• Model-based bioprocess optimization relies on accurate mathematical models.
• Bioprocess models often suffer from significant uncertainty due to parameter non-identifiability.
• Optimization using a single best-fit model can lead to poor real-world performance.

Purpose of the Study:

• To develop and demonstrate an ensemble modeling strategy for robust bioprocess optimization under model uncertainty.
• To improve the reliability of bioprocess optimization by accounting for parameter uncertainty.

Main Methods:

• Employed a Bayesian approach to define the posterior distribution of model parameters.
• Generated an ensemble of model parameters by sampling the parameter confidence region.
• Utilized a mean-standard deviation utility function to maximize the lower confidence bound of the bioprocess objective.

Main Results:

• The proposed ensemble modeling strategy effectively accounts for model uncertainty.
• Optimization based on the ensemble model demonstrated improved robustness.
• Successfully applied the strategy to optimize monoclonal antibody production in mammalian cell culture.

Conclusions:

• Ensemble modeling provides a robust framework for bioprocess optimization in the presence of model uncertainty.
• The Bayesian approach and ensemble generation are effective for improving optimization performance.
• This strategy enhances the reliability of bioprocess optimization for industrial applications.