Reliable protein structure refinement using a physical energy function.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Structure Prediction

Background:

• Significant advancements in protein structure prediction have been achieved.
• Refining predicted models to experimental resolution for functional studies remains challenging.

Purpose of the Study:

• To couple a modular protein-solvent energy function with a global optimization method.
• To refine template models from Critical Assessment of Techniques for Protein Structure Prediction (CASP) blind prediction contests.

Main Methods:

• Utilized a protein-solvent energy function incorporating a potential of mean force for hydrophobic solvation.
• Employed the stochastic perturbation with soft constraints global optimization method.
• Applied a prediction protocol selecting structures with the lowest energy.

Main Results:

• Successfully refined all test proteins from CASP predictions.
• The energy function demonstrated robust performance without prediction degradation, even when sampling was exhausted.
• Achieved refinement to a resolution comparable with experimental results.

Conclusions:

• The coupled energy function and optimization method effectively refines protein structure models.
• The approach holds promise for improving the accuracy of protein structure prediction for functional analysis.