Exploiting sequence and structure homologs to identify protein-protein binding sites

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• The growing number of 3D protein structures aids in understanding and predicting protein-protein interactions.
• Identifying binding sites is crucial for understanding protein complex formation.

Purpose of the Study:

• To develop and evaluate a novel method for predicting protein-protein binding sites.
• To improve the accuracy of protein-protein interaction site prediction by integrating structural and sequence information.

Main Methods:

• Utilized multiple structure alignments to identify structurally conserved residues.
• Weighted conservation scores by crystallographic B-factor to account for structural flexibility.
• Combined sequence profile and accessible surface area data with conservation scores.
• Employed a Support Vector Machine (SVM) for binding site prediction.

Main Results:

• The integration of the conservation score significantly enhanced SVM performance.
• Precisely predicted over 52% of binding sites ( >70% residue identification).
• Correctly predicted 77% of binding sites ( >50% residue identification).
• Achieved partial prediction for 21% of binding sites.

Conclusions:

• The developed method effectively predicts protein-protein binding sites.
• Structurally conserved residues, weighted by flexibility, are key predictors.
• Results support the role of rigid residues in minimizing entropic costs during complex formation.