Calibration Transfer Across Instrument Vendors for Bioprocess Raman Monitoring

Raman spectroscopy is a proven Process Analytical Technology (PAT) for monitoring mammalian cell culture processes in biopharmaceutical manufacturing. In-line Raman probes provide real-time chemical fingerprints of metabolites and cell culture health analyzed via chemometric modeling. However, Raman hardware (e.g. cables, detectors, optics, probes, and lasers) and software - performing calibration, noise reduction, and cosmic ray removal - impart vendor specific spectral signatures, rendering Raman chemometric models specific to the vendors. This vendor specificity complicates method validation and transfer between manufacturing sites deploying different vendor equipment and impedes upgrades, maintenance, and replacement of obsolete Raman equipment. In this work, we compared two calibration transfer methods to address vendor-to-vendor variation. Piecewise Direct Standardization (PDS) and Spectral Subspace Transformation (SST) methods successfully reduced spectral response variation between previous (Parent) and new (Child) Raman systems. We tested calibration transfer results with offline samples and an established validation approach utilizing paired spectra from Parent and Child Raman systems. Finally, we explored the influence of calibration transfer parameters including training set size, preprocessing position, and window size (number of components) for PDS and SST. Through this investigation we demonstrate the feasibility of this proposed vendor-to-vendor calibration transfer approach as a promising and effective chemometric model transfer between Raman vendors without significant method re-development or re-validation. This approach improves agility within the deployment strategy of chemometric models across supply chain networks and bolsters Raman spectroscopy as a versatile PAT tool for advanced manufacturing within the biopharmaceutical industry.