Enabling rapid identification of biotherapeutic protein products using handheld Raman spectrometers and principal component analysis

• Published in: Journal of Raman spectroscopy Vol. 52; no. 7; pp. 1281 - 1293
• Main Authors: Soto, Robert J., Meriage, David S., Wagner, Robin L., Wang, Ting, Semin, David J.
• Format: Journal Article
• Language: English
• Published: Bognor Regis Wiley Subscription Services, Inc 01.07.2021
• Subjects: Algorithms; Analyzers; Biopharmaceuticals; Chemical compounds; identity; Methods; multivariate; Pharmacology; Preventive maintenance; Principal components analysis; Product differentiation; Proteins; Raman; Raman spectra; Raman spectroscopy; Spectrometers; Spectrum analysis; Wavelengths
• ISSN: 0377-0486; 1097-4555
• DOI: 10.1002/jrs.6130

Herein, the utility of Raman spectroscopy for rapid identity (ID) verification of biotherapeutic protein products in solution is demonstrated. The success of this approach hinges on the unique combination of Raman features associated with both the therapeutic agent and excipients as the basis for product differentiation. Product ID method development subjects Raman spectra of the target product(s) to dimension reduction using principal component analysis (PCA) to define product‐specific models which serve as the basis for the ID method decision algorithm. In the initial stages of this work, instrumental sources of variance such as intensity‐axis multiplicative scaling and wavenumber‐axis alignment lead to high rates of erroneous false‐negatives when the methods were executed on other instruments that were not used for model development. To overcome these issues, the PCA‐based spectral models were systematically studied and iteratively adjusted to elucidate key design requirements for transferable models. The use of specific spectral preprocessing algorithms was implemented to compensate for instrumental variance, which reduced false‐negative rates without sacrificing method specificity. Five different methods, along with the associated product samples, were evaluated on 15 Raman instruments that had not been used for method development. No false‐positive analyses occurred for any of the five methods. A small number of false negatives (~2%) were observed on analyzers that performed outside of the instrument manufacturer's specifications and thus would have been identified during preventive maintenance. Taken together, this paper lays the groundwork for developing and implementing Raman ID methods for solution‐based protein products in the biopharmaceutical industry. Multivariate Raman spectroscopic models have been implemented for identity (ID) testing of biotherapeutic protein products. The models are demonstrated to be highly selective to the target product and are readily transferrable across instruments of ranging age and functionality.