FREAD

FREAD predicts protein loop backbone conformations by searching a database of known protein structures for candidates matched by local sequence similarity within loop anchor contexts.

Key Features:

• Database Search Approach: Uses a database search strategy to identify candidate loop conformations from known protein structures rather than relying on ab initio physics-based modeling.
• Environment-Specific Substitution Scores: Applies environment-specific substitution scores to quantify sequence similarity within structural contexts and refine candidate selection.
• Performance Against Ab Initio Methods: Outperforms ab initio methods such as MODELLER, PLOP, and RAPPER for a subset of loops, achieving predictions within 2 Å RMSD for two-thirds of shorter loops and for half of the longer loops tested.
• Length-Independent Predictive Ability: Maintains high accuracy across loop lengths within its effective subset, in contrast to other methods that can exceed 10 Å RMSD for loops of ~20 residues.
• Benchmarking Success: On CASP 7 and 8 model structures, FREAD made predictions for 127 of 212 tested loops and provided the best prediction in 61 cases.
• Independence from Whole Structure Quality: Reported loop prediction quality is independent of the overall quality of the protein model.

Scientific Applications:

• Structural biology: Improves loop conformation modeling to support structural interpretation of proteins.
• Protein engineering: Provides accurate loop backbones useful for redesigning surface-exposed or functional loop regions.
• Drug design: Refines loop conformations that can affect binding site geometry and ligand interactions.
• Protein function analysis: Aids understanding of protein function through more reliable structural models of variable surface loops.

Methodology:

Searches a database of known protein structures using local sequence similarity between loop anchors and scores candidate loops with environment-specific substitution matrices.