bSiteFinder, an improved protein-binding sites prediction server based on structural alignment: more accurate and...

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Protein-binding site prediction is crucial for protein functional annotation and structure-based drug design.
• Existing methods often underutilize available 3D protein-ligand complex structures.
• Improvements are needed in template alignment, selection, and clustering for enhanced prediction.

Purpose of the Study:

• To develop an improved protein-binding site prediction server.
• To leverage a comprehensive database of high-quality bound protein templates.
• To enhance the accuracy and efficiency of binding site prediction algorithms.

Main Methods:

• Construction of the largest database of bound protein templates with stringent quality control.
• Development of the bSiteFinder prediction server incorporating novel algorithms.
• Integration of Homology Indexing, Chain Length Indexing, Stability of Complex, and Optimized Multiple-Templates Clustering.

Main Results:

• The bSiteFinder server demonstrated significantly improved efficiency.
• Accuracy was approximately 2-10% higher than existing algorithms on both bound and unbound datasets.
• Achieved high accuracies of 94.8% (MCC 0.95) for a bound dataset and 85.4% (MCC 0.72) for an unbound dataset.

Conclusions:

• The bSiteFinder server is now freely available online, supporting protein functional annotation and drug design.
• The developed methods offer a more accurate and time-efficient approach to binding site prediction.
• The server's performance is expected to improve further with the increasing availability of 3D protein structures.