reFOLD

reFOLD refines user-supplied three-dimensional protein models using iterative global and local refinement guided by Accuracy Self Estimates (ASEs) to improve local and global model quality.

Key Features:

• Iterative refinement: Performs iterative refinement cycles on input 3D protein models.
• Global and local ASE integration: Integrates global and local Accuracy Self Estimates (ASEs) to assess and guide per-region reliability.
• Multiple alternative models: Generates multiple alternative refined models for each input to explore conformational variations.
• Residue-level comparison: Enables comparison of key residue locations across alternative models to assess local improvements.
• Local/domain-focused refinement: Targets local regions or individual domains for refinement when global chain refinement is challenging.
• High-throughput demonstration: Produced hundreds of alternative refined models in CASP12, contributing to performance in TBM and FM/TBM categories.

Scientific Applications:

• Tertiary structure prediction: Improves accuracy and reliability of predicted tertiary structures.
• Template Based Modeling (TBM) and Free Modeling (FM): Supported enhanced performance for TBM and FM/TBM targets in CASP12.
• Error identification and correction: Facilitates detection and correction of likely local errors in protein models.
• Structural biology and protein engineering: Aids interpretation of structural features relevant to protein function and interactions and guides engineering efforts.
• Model confidence estimation: Provides per-region reliability estimates to inform experimental design or downstream computational analyses.

Methodology:

Iterative refinement driven by global and local Accuracy Self Estimates (ASEs) to generate multiple alternative refined models and enable residue-level comparisons.