A structure-based model for the prediction of protein-RNA binding affinity.

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Protein-RNA recognition is crucial for cellular functions and is driven by binding affinity.
• Understanding the factors governing protein-RNA binding affinity is essential for deciphering molecular mechanisms and engineering interactions.

Purpose of the Study:

• To curate a comprehensive dataset of protein-RNA complexes with known binding affinities.
• To develop and validate a predictive model for protein-RNA binding affinity based on structural and physicochemical interface parameters.

Main Methods:

• Curated a dataset of 40 protein-RNA complexes from the docking benchmark with available unbound partners.
• Trained regression models using interface parameters such as relative hydrophobicity, conformational change, and hydration.
• Validated the best-fit model on a test dataset of mutated yeast aspartyl-tRNA synthetase structures with experimental binding free energy (ΔG) values.

Main Results:

• The dataset spans eight orders of magnitude in affinity and includes diverse RNA structural classes.
• Complexes with single-stranded RNA generally exhibit higher affinity than those with duplex RNA.
• The developed regression model, utilizing three key interface parameters, accurately predicts binding affinity, showing high correlation with experimental data.

Conclusions:

• The curated dataset and developed predictive model advance the understanding of protein-RNA binding affinity.
• The model provides a valuable tool for predicting binding affinity and engineering protein-RNA interfaces with tailored properties.