Protein surface matching by combining local and global geometric information

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Drug Discovery

Background:

• Protein binding site comparison is crucial for predicting protein function.
• Accurate prediction of binding ligands from protein atomic structures is a persistent challenge in structural bioinformatics.
• Existing methods often struggle with the inherent plasticity of protein binding sites.

Purpose of the Study:

• To develop a novel algorithm for enhanced protein binding site comparison and ligand prediction.
• To address the limitations of rigid-body alignment in capturing the dynamic nature of protein binding sites.
• To improve the accuracy of predicting binding ligands based on protein structural data.

Main Methods:

• Designed TIPSA (Triangulation-based Iterative-closest-point for Protein Surface Alignment), an algorithm based on the iterative closest point (ICP) algorithm.
• Utilized 3D Delaunay triangulation to identify initial similar tetrahedra between binding sites.
• Incorporated the Hungarian algorithm for matching additional atoms and global geometric information (radius of gyration) for improved prediction.

Main Results:

• TIPSA effectively superposes atoms between protein binding sites within a given distance threshold.
• The algorithm demonstrated that rigid-body alignment alone is insufficient due to protein binding site plasticity.
• Incorporating global geometric features and nearest neighbor classification enhanced binding site prediction accuracy.
• Achieved performance comparable to state-of-the-art methods on benchmark datasets.

Conclusions:

• The developed TIPSA algorithm offers a robust approach for comparing protein binding sites.
• The method successfully identifies common atom sets and atom correspondences, aiding in ligand prediction.
• Integrating global geometric information significantly improves the prediction of binding ligands, advancing structural bioinformatics and drug discovery efforts.