Multi-cell-line learning for the data-driven construction of mechanistic metabolic models

Area of Science:

• Biotechnology
• Metabolic Engineering
• Systems Biology

Background:

• Mammalian cell cultures are crucial for biopharmaceutical production.
• Mechanistic metabolic models capture cellular complexity but often overlook enzyme dynamics.
• Temporal changes in enzyme abundance and activity significantly impact metabolic modeling.

Purpose of the Study:

• To develop a framework for mechanistic metabolic models integrating growth-signaling control of enzyme activity and transcript dynamics.
• To apply this framework to Chinese hamster ovary (CHO) cell lines for biomanufacturing.
• To improve the accuracy and predictive power of cell line models.

Main Methods:

• Developed a framework for mechanistic metabolic models incorporating enzyme activity and transcript dynamics.
• Applied the framework to three Chinese hamster ovary (CHO) cell lines using fed-batch culture data and time-series transcript profiles.
• Implemented a multi-cell-line (MCL) learning approach for parameter estimation, combining data from different cell lines.

Main Results:

• Demonstrated the significant role of growth signaling and transcript variability in metabolic models.
• Showcased the effectiveness of the MCL approach for constructing accurate cell-line models with limited data.
• Achieved high accuracy in predicting distinct metabolic behaviors across different CHO cell lines.

Conclusions:

• The developed framework accurately models cell metabolism by integrating enzyme dynamics and transcript profiles.
• The MCL learning approach enhances model accuracy and is beneficial when data is limited.
• These models can accelerate bioprocess and cell-line development for protein therapeutics manufacturing.