WebFEATURE

WebFEATURE identifies and scores three-dimensional physicochemical motifs in protein and nucleic acid structures to locate functional sites for structural annotation.

Key Features:

• Input and site scanning: Accepts an atomic structure or a Protein Data Bank (PDB) identifier plus a statistical model of a functional site, then scans the query structure to return rank-scored hits that locate regions with similar physicochemical property distributions to the model.
• Supervised learning model generation: The underlying FEATURE package employs a supervised learning algorithm to create three-dimensional physicochemical site models.
• Recognition by comparison: Query site properties are compared against distributions derived from known sites and nonsites to assess similarity.
• Log-odds scoring: A log-odds scoring function based on Bayes' Rule computes scores indicating the likelihood that a query region is a functional site.
• Multi-level microenvironment descriptors: Descriptions of protein microenvironments use physical and chemical properties at atomic, chemical group, residue, and secondary structural levels.

Scientific Applications:

• Structural genomics annotation: Rapid annotation of molecular function in newly determined protein and nucleic acid structures within structural genomics projects.
• Calcium binding site identification: Identification of calcium binding sites in proteins with reported high sensitivity and specificity.
• General motif detection: Detection of various functional motifs across a wide range of biomolecules including proteins and nucleic acids.

Methodology:

FEATURE generates detailed microenvironment descriptions at atomic, chemical group, residue, and secondary structural levels, uses supervised learning to build site models, compares query properties to distributions from known sites and nonsites, and computes log-odds scores based on Bayes' Rule.