SiteFerret: Beyond Simple Pocket Identification in Proteins.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Drug Discovery

Background:

• Accurate identification of binding pockets on protein surfaces is crucial for understanding molecular interactions and drug design.
• Existing methods may struggle with diverse pocket types, including shallow or peptide-binding sites.

Purpose of the Study:

• To introduce SiteFerret, a novel computational method for the automatic and precise detection of protein binding pockets.
• To demonstrate the versatility of SiteFerret across various binding site types, from small molecules to peptides.

Main Methods:

• Utilizes hierarchical clustering of virtual probe spheres derived from protein solvent-excluded surfaces.
• Employs the Isolation Forest unsupervised learning method for ranking putative pockets based on feature importance.
• Analyzes both geometric and amino acid composition features to characterize binding sites.

Main Results:

• SiteFerret accurately detects a wide range of protein binding sites, including challenging shallow pockets.
• The method provides precise localization of binding sites through pocket segmentation into subpockets.
• Feature importance analysis offers insights into key characteristics of effective binding sites.

Conclusions:

• SiteFerret offers a versatile and accurate solution for automated protein binding site prediction.
• Subpocket segmentation enhances the precision of binding site identification.
• The method aids in understanding structure-activity relationships and facilitates drug discovery efforts.