Boosting AlphaFold Protein Tertiary Structure Prediction through MSA Engineering and Extensive Model Sampling and...

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• AlphaFold2 and AlphaFold3 significantly advanced protein tertiary structure prediction.
• Predicting structures for proteins with shallow multiple sequence alignments (MSAs) or complex domain architectures remains a challenge.

Purpose of the Study:

• To develop an integrative system, MULTICOM4, to enhance protein structure prediction accuracy for difficult targets.
• To improve the generation and ranking of protein structure models from existing tools like AlphaFold2 and AlphaFold3.

Main Methods:

• MULTICOM4 utilizes diverse MSA generation techniques and large-scale model sampling.
• An ensemble model quality assessment (QA) strategy combines multiple QA methods for improved model selection.
• The system incorporates MSA engineering and domain segmentation for better input data.

Main Results:

• MULTICOM4-based predictors were top performers in the 16th Critical Assessment of Techniques for Protein Structure Prediction (CASP16), ranking among the best out of 120 predictors.
• The best predictor, MULTCOM, achieved an average TM-score of 0.902 for 84 CASP16 domains.
• High accuracy (TM-score > 0.9) was reached for 73.8% of domains, and correct folds (TM-score > 0.5) were predicted for 97.6% in top-1 predictions.

Conclusions:

• MSA engineering, extensive model sampling, and ensemble QA are crucial for accurate protein structure prediction, especially for challenging targets.
• MULTICOM4 demonstrates a robust approach to improving protein structure prediction beyond standard methods.
• The findings highlight the importance of integrative strategies for advancing computational structural biology.