DCON

DCON predicts disulfide connectivity in cysteine-rich proteins to improve three-dimensional (3D) structure prediction by identifying which cysteine pairs form disulfide bridges.

Key Features:

• Graph Matching Approach: Formulates disulfide connectivity prediction as a graph matching problem with vertices representing cysteine residues and edges weighted by contact potentials.
• Residue Contact Potentials: Employs and tests various residue contact potential models, including comparisons to general mean force contact potentials, to score likelihoods of cysteine pairing.
• Algorithmic Framework: Uses the Edmonds-Gabow algorithm combined with Monte-Carlo simulated annealing to find maximum-weight graph matches and improve prediction accuracy.

Scientific Applications:

• Protein Structure Prediction: Narrows conformational search space by providing probable disulfide bond configurations to assist 3D structure prediction from sequence.
• Research on Cysteine-Rich Proteins: Enables analysis of proteins with multiple disulfide bonds and, for proteins with four disulfide bridges, achieves accuracy reported as 17 times higher than random predictors.

Methodology:

DCON constructs a graph with cysteine residues as vertices and edges weighted by residue contact potentials, develops and tests different contact potential models, and optimizes connectivity by maximizing match weight using the Edmonds-Gabow algorithm and Monte-Carlo simulated annealing.