Accurate automated protein NMR structure determination using unassigned NOESY data

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Conventional NMR protein structure determination is labor-intensive, relying on fully assigned NOESY peak lists.
• Existing automated methods (Aria, CYANA, AutoStructure) are sensitive to spectral quality and can yield inaccurate models.
• High signal-to-noise ratios in peak lists are crucial for reliable automated structure generation.

Purpose of the Study:

• To develop a more accurate and reliable method for protein structure determination using NMR data.
• To improve upon existing computational approaches by integrating different data types and refinement strategies.
• To reduce the manual effort and time required for protein structure determination.

Main Methods:

• A hybrid approach combining CS-Rosetta (Rosetta with chemical shifts) and NOESY data was developed.
• Initial models from CS-Rosetta were filtered using the DP-score for goodness-of-fit with unassigned NOESY peak lists.
• High-resolution refinement was performed using the Rosetta rebuild-and-refine protocol on selected models.

Main Results:

• The hybrid method produced highly accurate and reliable protein models, outperforming AutoStructure/CYANA and CS-Rosetta alone.
• Accuracy was particularly improved when using raw, unedited NOESY peak lists.
• Rosetta refinement consistently enhanced the accuracy of models generated by CYANA and AutoStructure when high-quality NOESY data was available.

Conclusions:

• Combining backbone chemical shifts and unassigned NOESY data with Rosetta offers a robust strategy for accurate protein structure determination.
• This hybrid approach mitigates limitations of purely automated methods and improves model quality, especially with lower-quality spectral data.
• The method provides a significant advancement for structural biology, enabling more efficient and reliable structural insights.