Prediction of protein assemblies by structure sampling followed by interface-focused scoring

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Proteins function in complexes, requiring 3D structure knowledge for understanding assembly function.
• Accurate prediction of individual protein structures by AlphaFold has advanced structural biology.
• Modeling protein complex structures remains a significant challenge in the field.

Purpose of the Study:

• To develop an improved protocol for predicting the three-dimensional structures of protein complexes.
• To enhance the accuracy of protein assembly modeling beyond current standard methods.
• To introduce a novel scoring method for evaluating protein complex models.

Main Methods:

• Developed a protocol involving extensive model sampling and interface-focused scoring.
• Diversified AlphaFold models via modified multiple sequence alignments and increased recycles.
• Employed monomer/subcomplex docking for complex cases and utilized VoroIF-jury for scoring.

Main Results:

• Achieved top results in CASP15, significantly outperforming the standard AlphaFold-Multimer pipeline.
• Demonstrated that structure sampling, rather than scoring, primarily limits assembly model accuracy.
• VoroIF-jury scoring is independent of AlphaFold scores and applicable to diverse prediction methods.

Conclusions:

• The developed protocol enhances protein complex structure prediction accuracy.
• Further improvements in sampling strategies, particularly for antibody-antigen complexes, are recommended.
• The protocol and VoroIF-jury scoring method offer valuable tools for modeling and scoring protein assemblies.