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 structure prediction for many proteins.
• Predicting structures for difficult targets with limited or noisy sequence data and complex domain arrangements remains a challenge.

Purpose of the Study:

• To develop an integrative system, MULTICOM4, to enhance protein structure prediction accuracy for challenging single-chain proteins.
• To improve the generation and ranking of structural models from existing prediction tools like AlphaFold2 and AlphaFold3.

Main Methods:

• Employed diverse multiple sequence alignment (MSA) generation strategies.
• Utilized large-scale model sampling and an ensemble model quality assessment (QA) approach.
• Integrated complementary QA methods and model clustering for improved ranking reliability.

Main Results:

• MULTICOM4-based predictors ranked among the top performers in the 16th Critical Assessment of Techniques for Protein Structure Prediction (CASP16).
• Achieved high accuracy on CASP16 domains, with an average TM-score of 0.902.
• Demonstrated superior performance compared to a standard AlphaFold3 predictor, with all domains correctly folded in best-of-top-5 predictions.

Conclusions:

• MSA engineering and extensive model sampling are crucial for accurate protein structure modeling, especially for difficult cases.
• The MULTICOM4 system offers practical strategies for modeling challenging single-chain proteins.
• Advances in protein structure prediction have significant implications for structural biology and drug discovery.