Protein function annotation by local binding site surface similarity.

Area of Science:

• Structural bioinformatics
• Computational biology
• Protein science

Background:

• Many protein crystal structures lack determined functions based on sequence similarity alone.
• Novel methods are needed to infer function for these uncharacterized proteins.

Purpose of the Study:

• To evaluate the enhanced Surflex-PSIM algorithm for automated protein binding site detection and functional inference.
• To assess the algorithm's accuracy in identifying known and predicting potential protein functions.

Main Methods:

• The Surflex-PSIM algorithm was employed for fully automated binding site detection.
• Methodology validated on apo/holo protein pairs, achieving high accuracy in ligand binding site identification (91% holo, 88% apo).
• Screened uncharacterized proteins against a database of ~60,000 PDB binding sites.

Main Results:

• Surflex-PSIM correctly identified functional matches for 5 out of 8 proteins with previously unknown but later annotated functions.
• The algorithm demonstrated high similarity between apo and holo binding sites (87%).
• Screening of 12 unannotated proteins yielded significant binding site matches, suggesting potential functions.

Conclusions:

• Automated Surflex-PSIM is a robust tool for hypothesizing protein function based on binding site similarity.
• The method accelerates functional annotation for proteins with determined structures but unknown roles.
• Surflex-PSIM facilitates the discovery of potential functions for poorly characterized proteins.