SurpResi

SurpResi identifies functionally important residues in proteins by probabilistic analysis of global radial distributions of atoms and information-theoretic evaluation of unexpected atomic locations to reveal functional sites such as ligand-binding pockets and protein-protein interfaces.

Key Features:

• Probabilistic Analysis: Uses continuous probability density functions to model preferred central distances between individual atoms and capture spatial preferences determined by amino acid composition, size, charge, and hydrophobicity.
• Mixture Models: Employs mixture models of radial density functions tailored to the specific amino acid makeup of proteins to estimate relative preferred burials of atoms.
• Information-Theoretic Evaluation: Assesses the unexpectedness of atomic locations using information-theoretic principles to identify residues that deviate from typical spatial distributions associated with functional roles.
• Validation and Performance: Demonstrated success rates comparable to several geometric methods for locating ligand-interacting binding sites and ranks protein docking predictions by identifying residues in hydrophobically unfavorable environments.

Scientific Applications:

• Binding-site identification: Detects candidate binding sites for ligands by locating residues with unexpected atomic distributions.
• Docking prediction ranking: Ranks protein-protein docking predictions by identifying residues in atypical or hydrophobically unfavorable environments.
• Protein-protein interaction detection: Identifies residues likely involved in protein-protein interfaces through deviations from expected radial distributions.
• Automated protein-function recognition: Complements existing techniques to assist automated recognition of protein functions based on probabilistic atomic distribution signatures.

Methodology:

Parametrization of continuous probability density functions describing preferred central distances of atoms from a set of selected globular proteins, use of mixture models of radial density functions tailored by amino acid composition, and information-theoretic assessment of unexpected atomic locations by analyzing those distributions to infer idealized spatial preferences and detect atypical residues.