Combining descriptive and predictive modeling to systematically design depth filtration-based harvest processes for...

Area of Science:

• Bioprocess Engineering
• Separation Science
• Computational Modeling

Background:

• Intensified upstream bioprocessing leads to increased cell debris, challenging depth filtration in biologics manufacturing.
• Current depth filtration development relies on resource-intensive screening, limited by feedstream variability and a lack of predictive tools.
• Existing semi-empirical fouling models are descriptive, hindering their application in guiding process development.

Purpose of the Study:

• To develop a predictive modeling approach for depth filtration performance in biologics harvest.
• To integrate mechanistic fouling insights with computational predictive capabilities.
• To establish a robust platform strategy for depth filtration development.

Main Methods:

• Developed a hybrid modeling approach combining mechanistic fouling models and computational models.
• Utilized historical bench-scale filtration data to build a partial least squares regression model.
• Predicted particle breakthrough based on filter and feedstream attributes to forecast filter performance.

Main Results:

• Successfully predicted depth filter performance a priori using the hybrid model.
• Demonstrated the model's ability to interpret fouling mechanisms and provide physical meaning.
• Established a robust platform strategy for depth filtration of Chinese hamster ovary cell cultures.

Conclusions:

• The hybrid modeling approach effectively predicts depth filtration performance.
• In silico tools, informed by continuous data, will be essential for accelerating harvest process development.
• This approach enables prospective experimental design and reduces overall experimental burden.