Multi-objective global optimization (MOGO): Algorithm and case study in gradient elution chromatography.

Area of Science:

• Biotechnology and Biochemical Engineering
• Computational Chemistry and Modeling
• Process Optimization

Background:

• Biotechnological separation processes require iterative optimization using experiments and models.
• Conflicting objectives such as productivity and yield complicate process optimization.
• Current methods often involve serial or parallel experiments and modeling for optimization.

Purpose of the Study:

• To introduce a novel algorithm for multi-objective optimization in biotechnological separation processes.
• To address the challenge of conflicting design targets like purity, yield, and processing time.
• To demonstrate the algorithm's effectiveness in optimizing chromatographic separation.

Main Methods:

• Utilizing statistical regression models, specifically Gaussian Process Regression Models (GPM), for predicting process performance.
• Applying a Multi-Objective Global Optimization (MOGO) algorithm that combines GPM with Expected Hypervolume Improvement (EHVI).
• Iteratively approximating the Pareto front using Markov Chain Monte Carlo (MCMC) sampling to plan experiments.

Main Results:

• The MOGO algorithm successfully optimized elution gradient and pooling strategy for a three-component system.
• Demonstrated simultaneous optimization of purity, yield, and processing time.
• A Monte-Carlo study confirmed the algorithm's efficiency, effectiveness, and robustness.

Conclusions:

• The proposed MOGO algorithm provides an efficient approach for multi-objective optimization in bioseparations.
• It enables finding optimal compromises between conflicting objectives with reduced experimental effort.
• The algorithm is robust and effective for complex separation challenges.