Key Findings
The study demonstrates that a metabolism-informed model can successfully transfer knowledge from fed-batch to continuous operation, with fine-tuning outperforming other transfer learning variants. Analysis reveals that the growth rate in fed-batch, rather than the operating mode itself, is the dominant axis of metabolic variation. This suggests that metabolic structure remains invariant across different modes of operation, providing a path for digital models to remain adaptive during regime shifts.
Technical Details
The model integrates reactor- and intracellular-level mass balances directly within a learning architecture. To maintain flexibility, the design deliberately avoids explicit steady-state assumptions and predefined kinetic rate laws. The researchers benchmarked the system using two metabolic network sizes and three metabolic optimisation strategies, finding that:
- Concentration predictions remained indistinguishable regardless of network size.
- Exchange flux fidelity was agnostic to the choice of solver.
- Network size only impacted the level of intracellular interpretability.
The model functions as a soft sensor, providing one-step ahead forecasting of extracellular concentrations and inferring unmeasured intracellular fluxes to support real-time process monitoring and control.
Background & Context
Chinese Hamster Ovary (CHO) cells are the primary platform for mAb manufacturing. While the industry is shifting from fed-batch toward continuous operation to achieve sustained high cell densities and volumetric productivities, this transition is hindered by limited intracellular observability and sparse extracellular measurements. This gap between mechanistic understanding and predictive deployment has historically complicated the real-time monitoring and control of continuous cultures.
Significance & Outlook
By utilizing conservation relationships that remain invariant across operating modes, this formulation provides a principled structural prior for cross-regime generalisation. The results identify potential regime-shift signals that can guide experimental design for the characterisation of continuous cultures, moving toward digital models grounded in first-principles knowledge.
Source: https://doi.org/10.64898/2026.10.28.762476
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime
