How Feasible Is Docking of PROTACs to POI-E3L Complexes? Testing Physics-Based and ML-Based Docking Tools

Area of Science:

• Computational chemistry and drug discovery
• Structural biology and molecular modeling
• Pharmacology and chemical biology

Background:

• Targeted protein degradation (TPD) is a novel therapeutic strategy.
• Proteolysis-targeting chimeras (PROTACs) are a key TPD modality.
• Computational modeling of PROTACs presents significant challenges due to their complexity.

Purpose of the Study:

• To evaluate physics-based and machine learning (ML) docking tools for modeling PROTAC-E3L ternary complexes.
• To establish performance benchmarks for PROTAC docking strategies.
• To provide guidelines for computational modeling in PROTAC discovery.

Main Methods:

• Analysis of 43 experimentally resolved POI-PROTAC-E3L structures.
• Benchmarking GLIDE, MOE, rDock, DiffDock, and GeoDirDock.
• Incorporation of van der Waals scaling, hydrogen bond constraints, NMA for flexibility, and AlphaFold2 structures.

Main Results:

• rDock with high sampling outperformed other physics-based docking tools.
• ML-based tools showed competitive RMSD but limited generalizability and required postprocessing.
• Normal mode analysis (NMA) significantly improved docking accuracy by incorporating receptor flexibility.

Conclusions:

• rDock is a strong candidate for PROTAC docking, especially with high sampling.
• ML docking tools require careful validation and refinement for PROTAC applications.
• Integrating receptor flexibility via NMA is crucial for accurate PROTAC ternary complex modeling.