Comparison of SVMR and PLSR for ATR-IR data treatment: Application to AQC of mAbs in clinical solutions

Abstract

Attenuated Total Reflectance Infrared spectroscopy (ATR-IR) enables rapid, preparation-free and cost-effective analysis of many clinically relevant samples. For instance, it can substitute traditional techniques like high performance liquid chromatography (HPLC) for Analytical Quality Control (AQC) of therapeutic solutions of anticancer drugs like monoclonal antibodies (mAbs). More reliable quantitative interpretation of the ATR-IR spectra is usually achieved with multivariate data mining protocols. In the present study, the analytical performance of ATR-IR was evaluated for four clinical solutions of mAbs: Bevacizumab (Avastin®), Trastuzumab (Ontruzant®), Cetuximab (Erbitux®) and Rituximab (Truxima®). As expected, the ATR-IR spectra were a combination of the bands of the antibody itself and those of the excipients. For quantitative analysis of the solutions, two types of regression methods have been compared namely, Partial Least Squares Regression (PLSR) and Support Vector Machine Regression (SVMR). While both approaches were relevant, the latter gave better performances in terms of accuracy, linearity and the percent relative error that was below 15%. The results are thus promising for future applications of ATR-IR Spectroscopy coupled to the SVMR algorithm as a clinical tool for AQC of mAbs.