Application of Machine Learning Models for Property Prediction in Targeted Protein Degraders

Introduction

Machine learning (ML) systems can model quantitative structure-property relationships (QSPR) using existing experimental data and predict properties for new molecules. With the rise of targeted protein degraders (TPDs), the applicability of ML models for TPD-centric projects is explored.

Key Points

• Performance on TPDs: ML models show comparable performance for TPD predictions in various properties.
• Predictive Errors: Different predictive errors observed for sub-modalities such as glues and heterobifunctionals.
• Transfer Learning: Strategies enhance predictions for heterobifunctionals.

Conclusion

ML-based QSPR models are effective in predicting ADME and physicochemical properties of TPD molecules, supporting their use in drug discovery for TPDs.