A gated graph transformer for protein complex structure quality assessment and its performance in CASP15

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein complexes are essential for biological functions.
• Predicting protein complex structures (quaternary structures) is vital for research.
• Accurately assessing the quality of predicted structures without native templates remains a significant challenge.

Purpose of the Study:

• To introduce DProQA, a novel method for predicting the quality of 3D protein complex structures.
• To develop a computational tool that can effectively rank predicted protein complex structures.

Main Methods:

• Developed a gated neighborhood-modulating graph transformer architecture.
• Incorporated node and edge gates to control information flow in graph message passing.
• Trained and evaluated DProQA on curated protein complex datasets prior to CASP15 and performed blind testing during CASP15.

Main Results:

• DProQA achieved 3rd place among single-model quality assessment methods in CASP15 for ranking protein complex structures.
• Demonstrated effectiveness in ranking predicted protein complex structures based on TM-score.
• Rigorous internal and external validation confirmed the method's performance.

Conclusions:

• DProQA is an effective tool for assessing the quality of predicted protein complex structures.
• The method facilitates the selection of high-quality predictions for downstream applications.
• Accurate quality estimation aids protein function analysis and drug discovery efforts.