Performance of pyDock in 8th CAPRI: Energy-Based Scoring Applied to Docking and AlphaFold Models.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Artificial Intelligence in Biochemistry

Background:

• The Critical Assessment of PRedicted Interactions (CAPRI) experiment benchmarks protein structure prediction methods.
• Protein-protein complex structure prediction remains a significant challenge in structural biology.

Purpose of the Study:

• To evaluate the impact of emerging AI methodologies, such as AlphaFold, on protein complex structure prediction within the 8th CAPRI edition.
• To assess the performance of a hybrid modeling approach combining traditional docking protocols with AI-driven predictions.

Main Methods:

• Utilized pyDock for initial modeling and scoring, transitioning to AlphaFold2 for subunit modeling in later targets.
• Incorporated AlphaFold-Multimer for complex modeling and employed a scoring system combining pyDock energy and AlphaFold confidence scores.
• Applied standard restraints and filters for scoring provided models.

Main Results:

• Achieved success in 45% of targets as predictors (ranking 4th) and 64% as scorers (ranking 3rd).
• Performance was consistent with previous CAPRI editions despite the introduction of novel AI tools.
• Identified challenging targets that proved difficult for all participants, indicating areas for future research.

Conclusions:

• The integration of AI tools like AlphaFold has significantly advanced protein complex structure prediction capabilities.
• Energy-based scoring and other established methods remain valuable when combined with AI predictions.
• Protein-protein docking is an evolving field, with continued challenges and opportunities for improvement.