Advances in inline quantification of co-eluting proteins in chromatography: Process-data-based model calibration and application towards real-life separation issues.

Pooling decisions in preparative liquid chromatography for protein purification are usually based on univariate UV absorption measurements that are not able to differentiate between product and co-eluting contaminants. This can result in inconsistent pool purities or yields, if there is a batch-to-batch variability of the feedstock. To overcome this analytical bottleneck, a tool for selective inline quantification of co-eluting model proteins using mid-UV absorption spectra and Partial Least Squares Regression (PLS) was presented in a previous study and applied for real-time pooling decisions. In this paper, a process-data-based method for the PLS model calibration will be introduced that allows the application of the tool towards chromatography steps of real-life processes. The process-data-based calibration method uses recorded inline mid-UV absorption spectra that are correlated with offline fraction analytics to calibrate PLS models. In order to generate average spectra from the inline data, a Visual Basic for Application macro was successfully developed. The process-data-based model calibration was established using a ternary model protein system. Afterwards, it was successfully demonstrated in two case studies that the calibration method is applicable towards real-life separation issues. The calibrated PLS models allowed a successful quantification of the co-eluting species in a cation-exchange-based aggregate and fraction removal during the purification of monoclonal antibodies and of co-eluting serum proteins in an anion-exchange-based purification of Cohn supernatant I. Consequently, the presented process-data-based PLS model calibration in combination with the tool for selective inline quantification has a great potential for the monitoring of future chromatography steps and may contribute to manage batch-to-batch variability by real-time pooling decisions.