AF3Complex Yields Improved Structural Predictions of Protein Complexes.

Area of Science:

• Structural Biology
• Computational Biology
• Biochemistry

Background:

• Accurate protein complex structures are crucial for elucidating biological pathways.
• Previous advancements led to AF2Complex, an improved AlphaFold 2 model for protein complexes.
• The need for enhanced structural prediction models persists.

Purpose of the Study:

• Introduce AF3Complex, a novel model based on AlphaFold 3 for protein complex structure prediction.
• Incorporate ligand exclusion capabilities into the model.
• Evaluate AF3Complex's performance against existing state-of-the-art models.

Main Methods:

• Developed AF3Complex by enhancing AlphaFold 3 with improvements from AF2Complex.
• Implemented a novel ligand exclusion method.
• Benchmarked AF3Complex against AlphaFold 3 and other models on diverse complex datasets.

Main Results:

• AF3Complex significantly outperforms AlphaFold 3 in predicting protein complex structures.
• High-fidelity structures were generated for challenging protein-peptide and antibody-antigen complexes.
• AF3Complex's predictions would have ranked among the top models in the CASP16 competition.

Conclusions:

• AF3Complex represents a significant advancement in protein complex structure prediction.
• The model demonstrates superior accuracy and versatility across various complex types.
• AF3Complex is freely available, facilitating further research in structural biology.