Enhancing RNA 3D Structure Prediction in CASP16: Integrating Physics-Based Modeling With Machine Learning for...

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Accurate RNA structure prediction is crucial for understanding biological function.
• The Critical Assessment of Structure Prediction (CASP) benchmarks RNA structure prediction methods.
• The Vfold method integrates diverse computational approaches for RNA structure modeling.

Purpose of the Study:

• To present the Vfold approach for RNA structure prediction in the CASP16 competition.
• To evaluate the performance of Vfold in predicting RNA monomer and multimer structures.
• To demonstrate the benefits of integrating physics-based models with machine learning for RNA structure prediction.

Main Methods:

• Vfold employs a hierarchical and hybrid strategy for RNA structure prediction.
• It utilizes physics-based models (Vfold2D, VfoldMCPX) for 2D structure prediction.
• 3D structure prediction incorporates template-based and molecular dynamics simulations (Vfold-Pipeline, IsRNA, RNAJP), enhanced by AlphaFold3 and template knowledge.

Main Results:

• The Vfold approach was applied to RNA monomer and RNA multimer categories in CASP16.
• Integration of AlphaFold3 and template knowledge with physics-based models improved prediction accuracy.
• The study highlights the effectiveness of hybrid methods in RNA structure prediction.

Conclusions:

• The Vfold method demonstrates a successful integration of traditional and machine learning techniques for RNA structure prediction.
• Hybrid approaches significantly enhance the accuracy of predicting complex RNA structures.
• Vfold's performance in CASP16 validates the synergy between physics-based and AI-driven modeling.