Protein structure prediction using sparse NOE and RDC restraints with Rosetta in CASP13

Area of Science:

• Structural biology
• Computational biophysics
• Biomedical research

Background:

• Accurate protein structure modeling from limited experimental data, such as nuclear magnetic resonance (NMR) spectroscopy, is crucial for biomedical research.
• The CASP13 NMR-assisted modeling challenge assessed current techniques for protein structure prediction using sparse, ambiguous, and error-prone NMR data.

Purpose of the Study:

• To evaluate the capabilities and limitations of computational modeling techniques in leveraging NMR data for protein structure prediction.
• To describe a novel two-stage protocol using the Rosetta software suite for de novo protein structure prediction assisted by NMR data.

Main Methods:

• A two-stage de novo prediction protocol using Rosetta: low-resolution model generation with non-NOE and RDC restraints, followed by refinement using comparative modeling with all NOE and RDC restraints.
• Iterative refinement of server-predicted models using Rosetta+NMR, incorporating non-NMR contacts and structural templates.

Main Results:

• Nine out of 16 Rosetta de novo models achieved the correct protein fold (GD_ST score > 45).
• High-resolution models (RMSD < 3.5 Å) were obtained in three instances.
• The meta-approach significantly improved model quality by integrating NMR refinement with server-predicted models.

Conclusions:

• The described Rosetta-based protocol effectively utilizes NMR data for accurate protein structure prediction.
• Integrating data-assisted refinement with advanced computational methods holds potential for determining structures of large proteins previously inaccessible to conventional NMR.
• This work advances the field of protein structure determination, with significant implications for future biomedical research.