In a major advancement for downstream biopharmaceutical manufacturing, Thermo Fisher Scientific has introduced a real-time, in-line solution for accurately quantifying monoclonal antibody (mAb) concentration using Raman process analysis. The method offers high precision across a dynamic range (0–135 g/L) and has demonstrated excellent transferability between different mAbs with varying buffer formulations.
A Need for Precise and Real-Time Protein Quantification
Therapeutic proteins like insulin, Fc-fusion proteins, and monoclonal antibodies (mAbs) are highly sensitive to environmental changes during processing. Conformational shifts caused by temperature, pH, shear forces, or chemical modification can result in protein denaturation or aggregation, reducing efficacy and safety. Monitoring protein concentration precisely throughout downstream processing steps — including ultrafiltration/diafiltration (UF/DF) — is therefore critical.
Traditionally, UV-Vis spectrometry has served as the industry standard for in-line protein quantification. However, Raman spectroscopy is emerging as a powerful complementary tool, offering additional insights into buffer composition, excipients, and critical quality attributes (CQAs).
The Solution: MarqMetrix™ All-In-One Process Raman Analyzer
Thermo Fisher’s study utilized the MarqMetrix™ Process Raman Analyzer in a UF/DF environment (see Figure 1), using a flowcell probe for continuous real-time monitoring. Two Partial Least Squares (PLS) chemometric models were developed:
- Amide I PLS Model (1550–1850 cm⁻¹ region)
- Extended Region PLS Model (850–1850 cm⁻¹ region, including protein-specific and buffer-related Raman shifts)
Both models demonstrated strong performance, with prediction errors under 5% in real-time and under 3% when applied to a different mAb (Product B) — validating the model’s transferability across protein formulations.
Key Findings
✔ High Model Accuracy
- Amide I Model:
- RMSEC: 0.526 mg/mL
- RMSECV: 0.607 mg/mL
- R² (CV): ~1
- Extended Region Model:
- RMSEC: 0.237 mg/mL
- RMSECV: 0.319 mg/mL
- R² (CV): ~1
✔ Model Specificity and Transferability
The Amide I model, anchored on the β-sheet carbonyl Raman signature at ~1670 cm⁻¹, exhibited high specificity and transferability. The Extended Region model, incorporating broader molecular vibrations (CH deformation, phenylalanine, tyrosine), demonstrated greater accuracy but slightly less robustness across varying buffer matrices.
✔ Real-Time Tracking in Dynamic Processes
During UF/DF runs, Raman predictions aligned closely with in-line and offline UV-Vis readings. Minor discrepancies were attributed to acquisition timing differences (Raman every 18s vs. UV-Vis every 12s), which are easily corrected with appropriate synchronization.
✔ Cross-Product Application
The same models trained on Product A were applied to Product B, a different mAb in a different buffer matrix, maintaining <3% absolute prediction error — a strong indicator of model robustness.
Benefits Over Conventional Methods
| Feature | UV-Vis | Raman |
|---|---|---|
| Protein quantification | ✔ | ✔ |
| Buffer composition analysis | ✘ | ✔ |
| Secondary structure/CQAs | ✘ | ✔ |
| Transferability across products | Limited | High |
| Interference from excipients | High | Low |
| Real-time data integration | ✔ | ✔ |
Process Raman enables non-invasive, label-free monitoring while delivering comprehensive process understanding, aligning with FDA and EMA process analytical technology (PAT) initiatives.
Conclusion
This study confirms that process Raman is a highly accurate and transferable solution for in-line protein quantification in downstream bioprocessing. The method provides fast, real-time results, reduces process variability, and ensures product quality. With added capabilities to monitor buffer composition, excipients, CQAs, and structural integrity, process Raman is fast becoming an essential tool in biopharmaceutical PAT strategies.
Thermo Fisher Scientific’s Amide I and Extended Region PLS models provide users with adaptable solutions tailored to the specific needs of robustness vs. sensitivity, making them ideal for scale-up, batch release, and QbD-based manufacturing.
