Differentiation of RNA-protein docking structures through molecular dynamics simulation and machine learning methods.

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Accurate prediction of RNA-protein complex structures is crucial but challenging.
• Existing methods often rely on docking simulations, requiring post-processing for accuracy validation.

Purpose of the Study:

• To develop an integrated computational method for accurate RNA-protein structure prediction.
• To enhance the identification of correct binding poses from docking simulations using machine learning.

Main Methods:

• Integration of specialized molecular dynamics (MD) simulations with machine learning (ML) techniques.
• Utilizing steered MD simulations to assess candidate structure stability.
• Employing ML models trained on simulation data for classification of correct/incorrect poses.
• Further refinement using thermodynamic simulations and ML for enhanced accuracy.

Main Results:

• Achieved 0.934 accuracy in classifying correct RNA-protein docking poses.
• Demonstrated high accuracy on challenging complexes predicted by AlphaFold3 (0.80-0.96).
• The method effectively distinguishes correct structures from numerous docking outputs.

Conclusions:

• The developed method offers a powerful approach for accurate RNA-protein complex structure prediction.
• Integration of MD simulations and ML significantly improves the reliability of structural predictions.
• This approach addresses a key challenge in structural biology and bioinformatics.