Raman Spectroscopy as a PAT Tool for Quantifying Monoclonal Antibody Aggregates

The production of monoclonal antibodies (mAbs) is a multi-step process that exposes mAbs to different stresses that may lead to the formation of aggregates. Protein aggregation is a critical quality attribute as it alters the final product quality, safety, and efficacy. Monitoring aggregate in near real-time allows for better process understanding, which will lead to a reduction in batch-to-batch variation and produce consistent, high quality drug products. Currently, size exclusion chromatography (SEC) is the standard method for quantifying aggregate. While SEC is highly effective, it does have limitations. Raman enables near-real-time monitoring of CQAs as it is a non-destructive analytical technique. Raman spectroscopy measures a biochemical fingerprint that enables characterization of protein structural changes specific to the secondary and tertiary structure. Specific Raman bands associated with protein aggregation have been previously published and include changes in Amide I and III regions, and changes in phenylalanine, tryptophan, and tyrosine.
In this study we explored the use of Raman spectroscopy to predict protein aggregation; we experimentally induced aggregation using a combination of pH and temperature to generate high (>30%) and low (<3%) aggregated samples. To experimentally generate a range of samples that varied in aggregate levels between 2-15%, mixing experiments of the low and high aggregate material were performed.
Samples were analyzed by SEC and Raman. Partial least squares regression was used to predict the percent high molecular weight (%HMW). The training model demonstrated good performance with an error of 1.1% and r2 of 0.88. Utilizing this training model, we were able to predict the test samples with an error of 2%. These results highlight the potential of Raman spectroscopy as a PAT tool for real-time aggregate detection, with future directions aimed at directly integrating the Raman instrument with the process.