COSMIC-dFBA: A novel multi-scale hybrid framework for bioprocess modeling

Area of Science:

• Biotechnology
• Metabolic Engineering
• Computational Biology

Background:

• Cellular metabolism is crucial for biomanufacturing, driving cell growth and productivity.
• Dynamic metabolic shifts in bioreactors limit the predictive power of traditional mechanistic models.
• Accurate modeling is essential for optimizing bioprocess design and control.

Purpose of the Study:

• To introduce COSMIC-dFBA, a novel hybrid multi-scale modeling paradigm.
• To enhance the predictive capability of dynamic flux balance analysis (dFBA) for bioreactor processes.
• To accurately forecast cell density, antibody titer, and metabolite concentrations.

Main Methods:

• Developed COSMIC-dFBA, integrating machine learning with genome-scale metabolic models.
• Decomposed metabolite uptake/secretion rates into contributions from distinct cell states (growth, antibody production).
• Parameterized the model using readily available spent media metabolite concentrations.

Main Results:

• COSMIC-dFBA accurately predicted cell density and antibody titer within 10% of measured values.
• Demonstrated significant improvements over standard dFBA: 90% for cell density and 72% for antibody titer.
• Successfully modeled dynamic metabolic changes within the bioreactor environment.

Conclusions:

• The hybrid COSMIC-dFBA framework effectively captures dynamic cellular metabolism in bioreactors.
• This approach expands the applicability of dFBA to complex, real-world biomanufacturing conditions.
• COSMIC-dFBA offers a powerful tool for bioprocess optimization and predictive control.