Separation of Fab therapeutic charge variants by ion-exchange chromatography using iterative mathematical and artificial neutral network modeling approaches.

Linear pH, salt, or dual pH-salt gradient elution is the most common ion-exchange chromatography method for monoclonal and Fab antibody purification, but maintaining precise gradients during biomanufacturing is challenging. In the present study, using chromatographic data of linear salt gradient elution of Fab therapeutic performed at different conditions of pH and linear gradient lengths, a step gradient elution has been developed using iterative mathematical and artificial neutral networks modeling approaches. The proposed approaches utilizes classical Yamamoto method and Mollerup's thermodynamic approach offering satisfactory prediction of distribution coefficient of protein species (main Fab and two A1 and A2 acidic variants) as a function of salt concentration based on the stoichiometric displacement model for different fixed pH conditions. The standard deviation (σ) between mathematical and neural network approach was compared for the model optimized thermodynamic parameters and binding charge. The deviation was found to vary from small {± (0-0.3)} σ, to moderate {± (0.4-0.6)} σ range for main Fab and acidic (A1 and A2) variants. The optimized conditions of pH and salt concentration were successfully identified from the model predicted distribution coefficient curves and were utilized to design a three-step salt gradient elution at a fixed pH chromatography process without further optimization, giving final purified Fab with a purity of 92% and yield of 76%. The presented approach for conversion of linear gradient to step gradient can be highly useful for developing a robust and simplified commercial scale chromatography purification process for complex biomolecules.