Blind tests of RNA-protein binding affinity prediction

Area of Science:

• Computational Biology
• Biophysics
• Molecular Biology

Background:

• RNA-protein interactions are crucial for essential biological processes like translation and gene expression regulation.
• Accurate modeling of these interactions is vital for biological understanding and system repurposing.
• Existing computational methods face challenges due to the complexities of RNA-protein binding.

Purpose of the Study:

• To develop and validate a robust computational framework for predicting RNA-protein binding affinities.
• To integrate advances in free energy functions, mutation modeling, and RNA state calculations.
• To establish a new standard for computational modeling in RNA-protein interaction studies.

Main Methods:

• Developed the Rosetta-Vienna RNP-ΔΔG method, integrating a unified free energy function for bound states.
• Incorporated automated Rosetta modeling for mutations and secondary structure-based calculations for unbound RNA states.
• Validated the method against high-throughput experimental data and independent test sets.

Main Results:

• Achieved high accuracy with root-mean-squared errors (RMSEs) of 1.3 kcal/mol on MS2 coat protein-RNA data.
• Demonstrated strong performance on independent test sets (RMSE of 1.5 kcal/mol) including SRP, U1A, PUM1, and FOX-1.
• Showcased exceptional accuracy (RMSE of 1.4 kcal/mol) in blind predictions of human PUM2-RNA binding affinities.
• Outperformed previous structure-based approaches in predicting RNA-protein binding energies.

Conclusions:

• The Rosetta-Vienna RNP-ΔΔG method provides a significant improvement in modeling RNA-protein binding affinities.
• This framework offers a foundation for future advancements in the field, testable with high-throughput experiments.
• The study highlights the potential of computational approaches to unravel complex molecular interactions in biology.