Good modeling practice for industrial chromatography: Mechanistic modeling of ion exchange chromatography of a bispecific antibody. (2nd November 2019)

• Record Type: Journal Article
• Authors: Rischawy, Federico; Saleh, David; Hahn, Tobias; Oelmeier, Stefan; Spitz, Julia; Kluters, Simon
• Is Part Of: Computers & chemical engineering. Volume 130(2019)
• Publication Date: 2019-11-02
• Subjects: Good modeling practice -- Mechanistic chromatography model -- Bispecific antibody -- Antibody charge variants -- Sensitivity analysis -- In silico robustness analysis; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2019.106532
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: A mechanistic chromatography model for a bispecific antibody format was developed. Guidelines for good modeling practice were applied throughout. Charge variants were found to be essential to model the elution profile. Over-parameterization, parameter correlations and systematic errors were discussed. Critical process parameters and acceptable operating ranges were assessed in silico . Abstract: In the biopharmaceutical industry, development and characterization of chromatography processes is typically based on statistical models. Although these approaches are easy to apply, the resulting models may fail to predict non-linear behavior in preparative chromatography with complex protein feed streams. An alternative to empirical methods are mechanistic models. In chemical engineering, mechanistic modeling has been a standard method for decades. As mechanistic models continue their advance in the biopharmaceutical industry, this study underlines the need of a standardized methodology for mechanistic model calibration. A lumped rate model was applied to the polishing chromatography of a bispecific antibody. Following guidelines for good modeling practice, the model was thoroughly analyzed. Potential limitations such as over-parameterization, parameter correlations, imprecise parameter estimates or systematic errors were considered by evaluation of parameter confidence intervals, visual sensitivity analysis and model validation across different scales. Application of simulations for identification of critical process parameters will be discussed.