Mechanistic modeling of ion exchange chromatography of a bispecific antibody
When bringing a new biopharmaceutical drug to the market, biopharma companies must ensure efficacy and safety of the drug, but also a stable and robust production process. The regulatory bodies thereby require the identification of critical process parameters and acceptable operating ranges of the process. But how can critical process parameters and acceptable operating ranges be accurately defined before running the final production process?
Introducing Good Modeling Practice for protein chromatography
The problem tackled by the team around Federico Rischawy and Simon Kluters from Boehringer Ingelheim (BI) is well known amongst biopharma companies that are about to bring new products to the market. Traditionally, most companies have to establish and qualify experimental scale-down models, so that the small-scale process reacts the same way to variations as the commercial process. Already at BioProcess International Europe 2019, BI showed a promising alternative: Using mechanistic models, they could predict the behavior of commercial processes more accurately than with the experimental scale-down models.
Applying GoSilico’s simulation software ChromX, BI investigated the process in silico. For a bispecific antibody purification process, BI created a predictive mechanistic process model from very few lab-scale experiments while following Good Modeling Practice (GMoP) methods. In total, seven validation runs were performed to prove the predictive power of the model and the representativeness for large scale. With the established model, BI was then able to determine the potentially critical process parameters and asses acceptable operating ranges for the at-scale process, solely by simulation with ChromX.
Development of a standardized workflow
As mechanistic models continue their advance in the biopharmaceutical industry, this study underlines the need of a standardized methodology for mechanistic model calibration. Following guidelines for good modeling practice allows to critically analyze the and its predictive power. Potential limitations such as over-parameterization, parameter correlations, imprecise parameter estimates or systematic errors should be considered by evaluation of parameter confidence intervals, visual sensitivity analysis and model validation across different scales.