A multisensor approach for improved protein A load phase monitoring by conductivity‐based background subtraction of UV spectra

Abstract

Real‐time monitoring and control of protein A capture steps by process analytical technologies (PATs) promises significant economic benefits due to the improved usage of the column's binding capacity, by eliminating time‐consuming off‐line analytics and costly resin lifetime studies, and enabling continuous production. The PAT method proposed in this study relies on ultraviolet (UV) spectroscopy with a dynamic background subtraction based on the leveling out of the conductivity signal. This point in time can be used to collect a reference spectrum for removing the majority of spectral contributions by process‐related contaminants. The removal of the background spectrum facilitates chemometric model building and model accuracy. To demonstrate the benefits of this method, five different feedstocks from our industry partner were used to mix the load material for a case study. To our knowledge, such a large design space, which covers possible variations in upstream condition besides the product concentration, has not been disclosed yet. By applying the conductivity‐based background subtraction, the root mean square error of prediction (RMSEP) of the partial least squares (PLS) model improved from 0.2080 to 0.0131 g L−1. Finally, the potential of the background subtraction method was further evaluated for single wavelength‐based predictions to facilitate implementation in production processes. An RMSEP of 0.0890 g L−1 with univariate linear regression was achieved, showing that by subtraction of the background better prediction accuracy is achieved then without subtraction and a PLS model. In summary, the developed background subtraction method is versatile, enables accurate prediction results, and is easily implemented into existing chromatography setups with typically already integrated sensors.

Comparison of UV‐ and Raman‐based monitoring of the Protein A load phase and evaluation of data fusion by PLS models and CNNs

“…This paper focuses on a comparison of UV‐ and Raman‐based monitoring of the Protein A breakthrough as well as the evaluation of data fusion techniques for both sensor signals. UV data were preprocessed as described by Rolinger et al (2020b), which leads to a significantly improved prediction as it suppresses absorption from interfering co‐eluting species. For an analysis of the UV spectra during the load phase, a comparison to elution spectra, and a detailed discussion on the effects of the preprocessing, we refer to Rolinger et al In the following, the focus is set towards an analysis of the Raman spectra and the comparison of the prediction quality based on UV‐ and Raman‐based models.…”

“…In biopharmaceutical downstream processing of monoclonal antibodies (mAbs), a focus of Process Analytical Technology (PAT) research has been on the monitoring of the Protein A load phase (Feidl, Garbellini, Luna, et al, 2019; Feidl, Garbellini, Vogg, et al, 2019; Thakur et al, 2020; Rüdt et al, 2017) as this application promises the most economic benefits due to the high costs of Protein A resin (Rolinger et al, 2020b). Economic improvements may be achieved due to multiple aspects.…”

“…Different spectroscopic sensors, like, ultraviolet (UV) (Rolinger et al, 2020b; Rüdt et al, 2017), Near‐Infrared (NIR; Thakur et al, 2020), and Raman (Feidl, Garbellini, Luna, et al, 2019; Feidl, Garbellini, Vogg, et al, 2019), have been investigated for the purpose of quantifying the mAb concentration in the column effluent with varying success. On the basis of the literature data, UV spectroscopy and Raman spectroscopy seem to be the most promising techniques for the breakthrough monitoring of the Protein A load.…”

“…In comparison to Raman‐based techniques, UV spectroscopy offers a higher measurement speed and a better signal‐to‐noise ratio for quantification of proteins in aqueous solutions with the drawback of less selectivity for different protein features (Rolinger et al, 2020a). To compensate the lower selectivity and thereby improve the prediction of the UV‐based PAT methods, dynamic background subtraction methods have been investigated to remove the influence of process‐related impurities on the UV spectra (Rolinger et al, 2020b; Rüdt et al, 2017). Another drawback of the UV spectroscopy in comparison to Raman spectroscopy is the detector saturation at high protein concentrations.…”

“…As loss function Mean Square Error (MSE) was used.For the combined Raman and UV-based CNN model, only a hyperparameter optimization of an additional dense layer on top of the individual dense layers was done to combine both models.Bayesian optimization was used again with a range between 12 and 64 neurons in the dense layer and the same conditions on the learning rate as for single sensor models3 | RESULTS AND DISCUSSIONThis paper focuses on a comparison of UV-and Raman-based monitoring of the Protein A breakthrough as well as the evaluation of data fusion techniques for both sensor signals. UV data were preprocessed as described byRolinger et al (2020b), which leads to a significantly improved prediction as it suppresses absorption from interfering co-eluting species. For an analysis of the UV spectra during the load phase, a comparison to elution spectra, and a detailed discussion on the effects of the preprocessing, we refer to Rolinger et al In the following, the focus is set towards an analysis of the Raman spectra and the comparison of the prediction quality based on UV-and Raman-based models.…”