DomainParser

DomainParser decomposes multi-domain protein structures into individual structural domains to support accurate domain identification for structural biology and domain database curation.

Key Features:

• Graph-Theoretic Representation: Each residue is represented as a node and each residue-residue contact as an edge with a capacity reflecting the type of interaction.
• Network Flow Formulation: Domain decomposition is formulated as a network flow problem that exploits the principle that residue-residue contacts are denser within domains than between them.
• Two-Domain Decomposition: For two-domain proteins, DomainParser identifies a bottleneck or minimum cut using the Ford-Fulkerson algorithm to minimize total cross-edge capacity.
• Multi-Domain Decomposition: For proteins with more than two domains, the method iteratively solves a series of two-domain problems to obtain a multi-domain decomposition.

Scientific Applications:

• Domain database curation: Supports updating domain databases and curating entries as structural repositories such as PDB expand.
• Structural biology and bioinformatics: Provides precise domain boundaries needed for structural analyses and annotation.
• Protein function and evolution studies: Enables analyses of functional units and evolutionary relationships at the domain level.
• Interaction network analysis: Facilitates mapping of protein interaction interfaces and domain-level interaction networks.

Methodology:

Residues are modeled as nodes and residue-residue contacts as capacity-weighted edges; the problem is posed as a network flow/minimum-cut problem, solved with the Ford-Fulkerson algorithm for two-domain cases and iteratively for multi-domain cases; the algorithm was validated on a standard test set of 55 proteins with 78.2% agreement overall and 96.7% residue-assignment accuracy for two-domain proteins.