Proteomics-based, multivariate random forest method for prediction of protein separation behavior during...

Area of Science:

• Biotechnology
• Protein Purification
• Chromatography

Background:

• Downstream purification of recombinant proteins is costly due to unknown host cell protein (HCP) separation behavior.
• Molecular pharming is a growing method for producing recombinant pharmaceuticals.
• Efficient purification is critical for the economic viability of biopharmaceutical production.

Purpose of the Study:

• To model the separation behavior of complex protein mixtures in cation-exchange chromatography (CEX).
• To reduce the effort and cost associated with purification process development.
• To predict the elution patterns of host cell proteins (HCPs) from corn germ extract.

Main Methods:

• Proteins were characterized using isoelectric point, molecular weight, and surface hydrophobicity via aqueous two-phase system (ATPS) partitioning and two-dimensional electrophoresis (2DE).
• A multivariate random forest (MVRF) model was developed using these characterization variables.
• The MVRF model predicted the elution behavior of individual corn HCPs in CEX.

Main Results:

• The MVRF method successfully predicted the elution pattern of characterized proteins.
• An average root mean squared error (RMSE) of 0.0406 was achieved, indicating high prediction accuracy.
• The study demonstrated the effectiveness of the characterization and modeling approach for protein purification.

Conclusions:

• The developed MVRF method accurately predicts HCP elution in CEX, significantly aiding purification process development.
• This approach offers a cost-effective solution for optimizing downstream processing in recombinant protein production.
• The findings support the application of predictive modeling for enhancing biopharmaceutical manufacturing efficiency.