DConStruct

DConStruct reconstructs protein three-dimensional structures from hybrid distance- and contact-based inter-residue interaction maps to improve ab initio folding accuracy.

Key Features:

• Hybrid Interaction Maps: Uses distance and contact information, including binary contacts and tri-level thresholded distance-based hybrid maps, to represent inter-residue interactions.
• Hierarchical Modeling with Iterative Self-Correction: Applies hierarchical modeling with iterative self-correction to progressively refine folding models.
• Scalability and Accuracy: Achieves higher reconstruction accuracy than CNS and scales performance with increased detail in input interaction maps, approaching near-optimal accuracy for complex proteins.
• Application to Soluble and Membrane Proteins: Capable of folding both soluble and membrane proteins with improved accuracy relative to CNS-based approaches.
• Performance in CASP Experiments: Outperforms several CNS- and fragment-based methods in CASP benchmarks, including comparisons against approaches using finer-grained distance maps.

Scientific Applications:

• Protein Structure Prediction from Sequences: Predicts three-dimensional protein structures from amino acid sequences using inter-residue interaction maps.
• Structural Biology and Folding Studies: Provides higher-accuracy models to support analysis of protein folding mechanisms and structure–function relationships.
• Method Benchmarking and Evaluation: Serves in large-scale benchmarking and CASP evaluations to compare and validate folding approaches.

Methodology:

Employs hybrid interaction maps that combine distance and contact data (including tri-level thresholded distance-based maps) and a hierarchical modeling approach with iterative self-correction to refine structural predictions, with accuracy scaling as input map detail increases.