Calibration of multiplexed fiber-optic spectroscopy.

Area of Science:

• Analytical Chemistry
• Biotechnology
• Process Analytical Technology (PAT)

Background:

• Biopharmaceutical manufacturing often uses multiple bioreactors in parallel for large-scale production.
• In situ monitoring via multiplexed fiber-optic spectroscopy generates complex spectral data.
• Variations in optical probe properties introduce significant noise, hindering accurate analysis with standard methods.

Purpose of the Study:

• To develop a novel calibration model for analyzing spectral data from multiplexed probes in bioprocesses.
• To mitigate the impact of probe-specific optical variations on spectral data accuracy.
• To enhance the reliability of in situ monitoring in large-scale biomanufacturing.

Main Methods:

• Proposed a new calibration model explicitly incorporating multiplicative parameters for each optical probe.
• Implemented a "dual calibration" strategy to counteract probe-induced spectral variations.
• Validated the model on MIR ternary mixture data and NIR bioprocess data sets.

Main Results:

• The proposed multiplex calibration model effectively mitigated detrimental effects from probe differences.
• Achieved significantly more accurate predictions compared to standard multivariate linear calibration models (e.g., PLS).
• Demonstrated superior performance over empirical preprocessing methods like OSC, SNV, and MSC.

Conclusions:

• The novel calibration model offers a robust solution for analyzing complex spectral data from multiplexed probes.
• This approach improves the accuracy and reliability of in situ bioprocess monitoring.
• Enables more efficient and precise control of large-scale biopharmaceutical manufacturing processes.