Predicting protein retention in ion-exchange chromatography using an open source QSPR workflow

Area of Science:

• Biopharmaceutical Process Development
• Computational Chemistry
• Protein Science

Background:

• High purity of protein-based biopharmaceuticals is crucial for safety, but process development is time-consuming.
• Computational approaches can significantly reduce development efforts by preselecting optimal process conditions.
• Quantitative Structure Property Relationship (QSPR) models are valuable for predicting properties exploited during purification.

Purpose of the Study:

• To present a novel, open-source Python tool for extracting protein features from 3D models.
• To enable transparent, local calculations for QSPR model development in bioprocess optimization.
• To demonstrate the application of QSPR for predicting chromatographic behavior and accelerating process design.

Main Methods:

• Development of an open-source Python tool for extracting protein surface features via grid representations.
• Training linear regression models using extracted features to predict chromatographic retention times/volumes.
• Validation of QSPR models using experimental data for anion and cation exchange chromatography.

Main Results:

• The Python tool allows transparent, local extraction of protein features for QSPR analysis.
• Trained QSPR models achieved high accuracy in predicting retention times for anion (R²=0.87) and cation (R²=0.95) exchange chromatography.
• The models demonstrate the potential of QSPR to significantly reduce experimental screening in bioprocess development.

Conclusions:

• The developed open-source tool facilitates the application of QSPR in bioprocess development.
• QSPR modeling effectively predicts protein chromatographic behavior, accelerating purification process design.
• This approach offers a cost-effective and widely applicable solution for optimizing biopharmaceutical manufacturing.