An improved protein decoy set for testing energy functions for protein structure prediction

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Modeling

Background:

• Protein structure prediction is crucial for understanding biological function.
• Existing decoy sets require improvement for rigorous scoring function evaluation.
• Rosetta's centroid/backbone decoy set is a foundational resource.

Purpose of the Study:

• To develop an improved protein model decoy set for evaluating scoring functions.
• To assess the performance of various all-atom energy functions in identifying native and near-native protein structures.
• To compare the accuracy of NMR and X-ray crystal structures in the context of computational modeling.

Main Methods:

• Generation of an enhanced decoy set with 1,400 models for 78 protein targets, focusing on near-native structures and clash minimization.
• Evaluation of diverse all-atom energy functions, including implicit solvent models and Lennard-Jones terms.
• Analysis of energy gaps between native and non-native conformations for X-ray and NMR structures.

Main Results:

• A solvent-accessible surface area-based solvation model demonstrated superior enrichment and discrimination of near-native decoys.
• Combining this solvation model with Lennard-Jones terms and Rosetta energy yielded the best performance.
• A significant energy gap was observed for X-ray structures, but not for NMR structures, indicating differences in their native-state representation.

Conclusions:

• The enhanced decoy set provides a robust benchmark for protein structure prediction scoring functions.
• Implicit solvent models, particularly surface area-based solvation, are critical for accurate scoring.
• Discrepancies in energy gaps between X-ray and NMR structures highlight the importance of experimental data quality in computational studies.