(312b) Classification and Prediction of Protein Binding in Hydroxyapatite Chromatographic Systems
Ceramic hydroxyapatite (CHA) chromatography is a multi-model separation method and offers unique selectivity for biological molecules such as polyclonal and monoclonal antibodies, nucleic acids and proteins. However, CHA has not been exploited to its potential mainly due to lack of fundamental understanding and difficulties in predicting chromatographic behavior. In this paper, a significantly large commercially available protein library which covers a wide range of protein isoelectric points, molecular weights, CHA interaction types, etc. was investigated to elucidate mechanisms of protein binding in hydroxyapatite chromatography. This large library was employed in a wide range of sodium chloride gradients in the presence of different phosphate concentrations. In addition, new custom made molecular descriptors were created based upon a priori knowledge of CHA interactions. These new classes of descriptors focus on the different interaction types available with CHA, specific distances between interaction groups, clusters and/or densities of interaction groups and possible synergistic binding or repulsion effects between the protein and CHA surfaces. Using this data set in concert with these new types of descriptors, quantitative structure property relationship (QSPR) classification models were generated to provide better understanding and further insights into protein selectivity in CHA systems. Finally, nonlinear SVM QSPR prediction models were constructed and evaluated for the different gradient experiments employed as a predictive tool for protein affinity in CHA with applications for industrial use. This study provides a deeper understanding into the mechanisms and selectivity of protein adsorption on CHA and will help to create predictive models which could be used for methods development in industrial bioprocesses.