Proteomics-based, multivariate random forest method for prediction of protein separation behavior during cation-exchange chromatography

The most significant cost of recombinant protein production lies in the optimization of the downstream purification methods, mainly due to a lack of knowledge of the separation behavior of the host cell proteins (HCP). To reduce the effort required for purification process development, this work was aimed at modeling the separation behavior of a complex mixture of proteins in cation-exchange chromatography (CEX). With the emergence of molecular pharming as a viable option for the production of recombinant pharmaceutical proteins, the HCP mixture chosen was an extract of corn germ. Aqueous two phase system (ATPS) partitioning followed by two-dimensional electrophoresis (2DE) provided data on isoelectric point, molecular weight and surface hydrophobicity of the extract and step-elution fractions. A multivariate random forest (MVRF) method was then developed using the three characterization variables to predict the elution pattern of individual corn HCP. The MVRF method achieved an average root mean squared error (RMSE) value of 0.0406 (fraction of protein eluted in each CEX elution step) for all the proteins that were characterized, providing evidence for the effectiveness of both the characterization method and the analysis approach for protein purification applications.

► Complex mixture of proteins characterized by 3 properties: pI, MW, hydrophobicity. ► Characterized proteins’ separation behavior during cation-exchange chromatography. ► Multivariate statistical method used to predict protein's separation behavior. ► Could be applied to other separation methods aiding purification process selection.