Enhancing Monoclonal Antibody Yield and Quality Through Automated Multi-Component Feedback Control Loops Using the MarqMetrix All-In-One Process Raman Analyzer

Significance of the topic

The application note documents an advanced Process Analytical Technology (PAT) approach using in-line Raman spectroscopy to enable real-time, multi-analyte feedback control in mammalian cell culture. Maintaining precise nutrient and metabolite levels (here glucose and lactate) is critical for consistent monoclonal antibody (mAb) yield and quality. Raman-based, non-destructive monitoring allows frequent (near real-time) quantitative measurements that support dynamic control strategies which reduce metabolic by-products, improve cell viability, and limit undesirable product modifications such as glycation.

Objectives and study overview

This study implemented and evaluated automated control strategies in 5 L fed-batch bioreactors seeded with CHO-K1 GS cells. The main objective was to demonstrate that simultaneous Raman-based monitoring of glucose and lactate can be used in closed-loop control to maintain a target total carbon concentration (glucose + lactate = 2 g/L) and thereby improve titer, product quality, and cell viability compared with conventional fed-batch bolus and single-analyte continuous glucose control.

Methodology

- Cell culture: CHO-K1 GS in Efficient‑Pro medium with insulin and anticlumping agent, inoculated at 0.75 million cells/mL; 14 day fed‑batch runs in duplicate. Standard platform feeds (Efficient‑Pro feed and Enhancer) were dosed either as bolus or continuous.
- Control strategies compared: (1) Automated continuous glucose control (target 8 g/L); (2) Automated fed‑batch bolus glucose control (bolus if glucose <3 g/L to 6 g/L); (3) Automated total carbon control (maintain glucose + lactate = 2 g/L) using multi‑analyte Raman feedback.
- Chemometrics: Partial Least Squares (PLS) regression models developed for glucose and lactate. Glucose used three spectral windows (1065–1232, 1595–1863, 2704–3078 cm⁻¹) with Savitzky–Golay 1st derivative (order 2; window 13), Standard Normal Variate (SNV), and mean centering. Lactate model used 800–1750 cm⁻¹ with Savitzky–Golay (1st derivative; window 11), an L1 norm normalization (area = 1 for 1540–1750 cm⁻¹) and mean centering. Leave‑one‑out cross‑validation (LOOCV) and RMSECV were used to set latent variable count and avoid overfitting.
- Data handling: Cosmic ray removal, averaging, timestamp alignment in an internal Python platform and in SOLO 9.3.1.
- Integration and control loop: MarqMetrix Raman analyzer produced real‑time glucose and lactate predictions (data cadence reported on the order of minutes). Predictions were exported to TruBio control software; TruBio relayed values to a DeltaV pump controller which calculated and dispensed glucose doses to maintain the chosen control policy.

Used instrumentation

- Thermo Scientific MarqMetrix All‑In‑One Process Raman Analyzer (785 nm excitation).
- MarqMetrix Performance BallProbe sampling optic and FlowCell sampling optic options.
- Raman acquisition settings reported: 785 nm laser, 450 mW nominal power, 3000 ms integration time, averages = 20.
- TruBio software for PAT data exchange and DeltaV distributed control system for pump actuation.

Main results and discussion

- Model performance: Chemometric models provided accurate, low‑bias predictions versus offline reference methods. Reported metrics across runs included correlation coefficients (R²) in the range ~0.80–0.96 and standard error of prediction (SEP) values generally below ~0.7 g/L depending on analyte and run; biases were small. These metrics supported robust real‑time control.
- Process performance: Implementing total carbon (glucose + lactate) control at 2 g/L produced superior outcomes relative to the two comparator strategies:

• Titer: >10% increase in final monoclonal antibody titer compared with standard strategies.
• Product quality: Marked reduction in glycation. Reported glycation levels were 2% for total carbon control versus 12% (continuous glucose) and 6% (fed‑batch bolus), corresponding to substantial percentage reductions (up to ~83% vs continuous control and ~66% vs fed‑batch control in the examples shown).
• Cell health: Viability improved (reported increases on the order of ~15% in some comparisons) and viable cell density profiles were comparable or improved vs other controls.

- Mechanistic interpretation: Maintaining a fixed total carbon pool encouraged cells to consume lactate later in the run (reduced net lactate accumulation), aligning metabolic fluxes away from overflow glycolysis and minimizing formation of reactive sugar species that drive non‑enzymatic glycation of antibodies. Continuous multi‑analyte feedback enabled dynamic dosing to match metabolic demand rather than simple threshold or open‑loop feeding.

Benefits and practical applications

- Multi‑analyte, single‑scan Raman monitoring enables non‑destructive, frequent measurement of key CPPs (glucose, lactate, viable cell density proxy, titer predictors), supporting closed‑loop feed control without time‑consuming offline assays.
- Automation integration: Demonstrated connectivity into commercial control platforms (TruBio and DeltaV), making Raman PAT a practical solution for automated bioreactor operation and reduced operator intervention.
- Operational advantages: Reduced glycation and higher titer improve downstream processing yield and product quality consistency; frequent measurements reduce decision latency and batch‑to‑batch variability.

Future trends and potential applications

- Extension to additional analytes: Raman models can be expanded to monitor amino acids, other metabolites, and product‑related quality attributes enabling more comprehensive metabolic steering and feed formulations.
- Model transferability and scale‑up: Continued work on chemometric robustness, transfer learning, and calibration transfer will ease deployment across cell lines, media, and scales.
- Integration with AI and advanced control: Combining high‑frequency PAT data with model predictive control (MPC) or machine learning controllers could enable anticipatory feeding policies and further optimization of yield and quality.
- Continuous and perfusion processes: Real‑time multi‑analyte control is directly applicable to continuous bioprocessing, where tight setpoint maintenance is essential.
- Regulatory and data governance: Adoption will require demonstration of model validation, traceability, and alignment with regulatory expectations for real‑time release and PAT frameworks.

Conclusion

The MarqMetrix All‑In‑One Raman analyzer, coupled with chemometric PLS models and closed‑loop integration to TruBio/DeltaV, enabled robust multi‑analyte feedback control in CHO fed‑batch bioreactors. A total carbon control strategy (glucose + lactate = 2 g/L) reduced lactate accumulation, increased mAb titer by >10%, substantially lowered glycation, and improved cell viability compared with standard single‑analyte or bolus feeding approaches. The work demonstrates the practical value of in‑line Raman PAT for automated, quality‑focused biomanufacturing.

Reference

1. Abu‑Absi NR; Kenty BM; Cuellar ME; Borys MC; Sakhamuri S; Strachan DJ; Hausladen MC; Li ZJ. Real Time Monitoring of Multiple Parameters in Mammalian Cell Culture Bioreactors Using an In‑Line Raman Spectroscopy Probe. Biotechnology and Bioengineering. 2011;108(5):1215–1221. doi:10.1002/bit.23023.
2. Zhou M; Crawford Y; Ng D; Tung J; Pynn AFJ; Meier A; Yuk IH; Vijayasankaran N; Leach K; Joly J; Snedecor B; Shen A. Decreasing Lactate Level and Increasing Antibody Production in Chinese Hamster Ovary Cells (CHO) by Reducing the Expression of Lactate Dehydrogenase and Pyruvate Dehydrogenase Kinases. Journal of Biotechnology. 2011;153(1):27–34. doi:10.1016/j.jbiotec.2011.03.003.
3. Villa J.; Khadka N.; Keck K.; Zhang L.; Zustiak M. Process Raman as Platform Solution For Automated Glucose Feeding in Fed‑Batch Bioreactors. (Thermo Fisher Scientific application work).
4. Villa J.; Zustiak M.; Ramirez D.; Kruger J.; Kuntz D.; Zhang L.; Khadka N.; Broadbelt K.; Woods S. Demonstrating Chemometric Model Transferability for 5 Mammalian Cell Lines and 5 Media Types Using the Thermo Scientific MarqMetrix All‑In‑One Process Raman Analyzer to Monitor Upstream Bioprocesses. (Thermo Fisher Scientific application work).
5. Villa J.; Zustiak M.; Kuntz D.; Zhang L.; Khadka N.; Broadbelt K.; Woods S. Use of Lykos and TruBio Software Programs for Automated Feedback Control to Monitor and Maintain Glucose Concentrations in Real Time. (Thermo Fisher Scientific application work).