In 2008, Kelley et al. presented an approach to identify feasible separation conditions by measuring a partition coefficient (Kp) in the linear portion of the isotherm using high-throughput screening. The refined Kp screening approach of Genentech published by McDonald et al. in 2016 clearly showed that a smart use of automated batch binding experiments can speed up process development significantly.
Using Kp screening results and empirical correlations, McDonald et al. were able to predict product yield and purity, as well as pool volume of step elutions at high load conditions. Still, the approach was not adopted industry-wide, as many experiments would be needed to create correlations for other mAb libraries. In addition, the mechanistic reasons for the approach are neglected, so that the proof of correctness cannot be provided.
Model building from batch adsorption data
In this study, a rigorous mechanistic approach was applied to analyze Kp data of an industrial monoclonal antibody and a high molecular weight species. In order to obtain adsorption isotherm parameters for column modeling, a modified method for fitting batch adsorption data to mechanistic model equations is applied that relies only on the applied and measured supernatant concentrations.
For this project, an isotherm fitting toolbox was developed as technology demonstrator. An integrated module for model building fits batch adsorption data at various pH and ionic strength condition to isotherm equations. Apart from binding kinetics, all isotherm parameters could be identified from the Kp data. The parameters were then transferred to ChromX to simulate column runs and to obtain retention times under various conditions. For validation and to obtain the remaining kinetic parameter, three gradient elution runs at 5g/L feed concentration were performed.
In Silico Scale-Up to 12k fermenter scale
All tools of model-based process development can be applied to the obtained column model, including the identification of optimal step elution conditions as originally conducted by McDonald et al. As the mechanistic model is able to extrapolate from gradient to step elutions, an in silico optimization of step elution conditions was performed in ChromX. Depending on the specific goals, different combinations of yield and purity can be obtained. To demonstrate the predictive power, an in silico scale-up was performed that accurately described the result of a production run in 12k fermenter scale.
Please fill out this form to
download your files