OPUS-Refine

OPUS-Refine refines protein backbone torsion angles Φ and Ψ and global protein conformations to improve the accuracy of protein structure predictions.

Key Features:

• Sampling-Based Methodology: Employs a sampling-based approach to refine predicted torsion angles and global structures and to integrate constraints from other prediction methods.
• Integration with Existing Predictions: Utilizes outputs from SPIDER3 and SPOT-1D as inputs or constraints to increase Φ/Ψ prediction accuracy.
• OPUS-TA Database: Incorporates the OPUS-TA neighbor-dependent statistical torsion angles database to boost sampling efficiency.
• Global Structural Constraints: Integrates contact maps predicted by RaptorX as global structural constraints to improve local and global accuracy measured by TM-score and RMSD.
• High Efficiency: Refines torsion angles in approximately 4 seconds and global structures in about 30 seconds for a 100-residue protein on a typical desktop computer.

Scientific Applications:

• Protein structure refinement: Improves predicted protein structures at both local (Φ/Ψ) and global levels for downstream structural analysis.
• Functional and interaction inference: Enhances structural accuracy used to interpret protein function and interaction networks.
• Structure-guided therapeutics: Provides refined models that support structure-based design and therapeutic intervention analyses.
• Integration-driven performance improvement: Serves as a post-processing step to boost accuracy of predictors such as SPIDER3 and SPOT-1D.

Methodology:

Sampling-based refinement of backbone torsion angles Φ and Ψ and global structures using the OPUS-TA neighbor-dependent statistical torsion database, integrating constraints from SPIDER3, SPOT-1D and RaptorX-predicted contact maps, with accuracy evaluated by TM-score and RMSD.