Single compound data supplementation to enhance transferability of fermentation specific Raman spectroscopy models.

Area of Science:

• Biotechnology
• Analytical Chemistry
• Process Engineering

Background:

• Raman spectroscopy offers real-time analyte quantification in fermentation using chemometric models.
• Current models calibrated on multiple fermentations lack specificity and transferability for process variations or different modes.
• Poor model transferability necessitates labor-intensive recalibration for related processes.

Purpose of the Study:

• To enhance the transferability and specificity of Raman spectroscopy chemometric models for fermentation processes.
• To improve real-time monitoring accuracy across different operation modes (batch vs. fed-batch) without extensive recalibration.
• To validate a strategy of single compound data supplementation for model enhancement.

Main Methods:

• Partial least-squares (PLS) models were calibrated for glucose, ethanol, and biomass using Saccharomyces cerevisiae batch fermentation data.
• Models were transferred to a fed-batch operation to assess transferability.
• Single compound spectral data was supplemented to enhance model performance without additional process runs.

Main Results:

• Supplemented models demonstrated increased target analyte specificity.
• Sufficient prediction accuracy was achieved for the fed-batch process (RMSEP: 3.06 mM glucose, 8.65 mM ethanol, 0.99 g/L biomass).
• High prediction accuracy was maintained for the batch process (RMSEP: 1.71 mM glucose, 4.20 mM ethanol, 0.17 g/L biomass).

Conclusions:

• Process data combined with single compound spectra offers a fast and efficient strategy for Raman spectroscopy application.
• This approach enables real-time process monitoring across related fermentation processes with improved model transferability.
• The method reduces the need for extensive recalibration, making Raman spectroscopy more versatile for industrial applications.