Rapid Simulation of Unprocessed DEER Decay Data for Protein Fold Prediction

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Monte Carlo methods struggle with de novo protein structure prediction for large or complex proteins.
• Existing methods using site-directed spin labeling and electron paramagnetic resonance (EPR) spectroscopy have computational limitations.
• Current approaches often rely on distance distributions with inherent uncertainties, requiring manual refinement.

Purpose of the Study:

• To develop a computational method for accurate de novo protein structure prediction using sparse experimental data.
• To overcome the resolution-accuracy trade-offs in existing spin labeling-based modeling techniques.
• To improve the efficiency and reliability of protein structure prediction for challenging protein targets.

Main Methods:

• Developed RosettaDEER, a scoring method within the Rosetta software suite.
• Simulated double electron-electron resonance (DEER) spectroscopy decay traces and distance distributions.
• Integrated DEER data simulation for de novo protein folding and refinement.

Main Results:

• RosettaDEER achieves accuracy in distance distributions comparable to or exceeding more computationally intensive methods.
• Generated decay traces accurately recapitulate intermolecular background coupling parameters, even with truncated data.
• RosettaDEER effectively discriminates between native-like and poorly folded models using DEER decay traces.
• Demonstrated improved protein fold prediction efficacy on two challenging test cases.

Conclusions:

• RosettaDEER enables efficient and accurate de novo protein structure prediction by leveraging sparse experimental data.
• The method overcomes limitations of previous computational implementations for spin labeling data analysis.
• RosettaDEER enhances the Rosetta software suite's capabilities for diverse protein modeling applications.