From complete cross-docking to partners identification and binding sites predictions

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein interactions are crucial for cellular functions, medicine, and bioengineering.
• Existing experimental data on protein interactions are often heterogeneous, noisy, and contradictory.
• Ab initio methods offer a way to reconstruct protein-protein interaction networks without prior experimental data.

Purpose of the Study:

• To develop and evaluate a molecular cross-docking approach for identifying protein interaction partners.
• To investigate the impact of interface characteristics and scoring functions on prediction accuracy.
• To assess the method's ability to handle multi-use protein surfaces and compare it with deep learning techniques.

Main Methods:

• Utilized a coarse-grained, rigid-body molecular cross-docking algorithm.
• Applied the method to hundreds of proteins in their unbound conformations.
• Systematically analyzed factors like interface size, quality, and scoring function performance.

Main Results:

• Achieved significant improvements in protein-protein interaction prediction compared to previous methods.
• Demonstrated high discriminative power for specific functional classes of proteins.
• Quantified the contributions of shape, physico-chemical complementarity, and interface matching to prediction accuracy.

Conclusions:

• The developed cross-docking approach effectively reconstructs protein-protein interaction networks.
• The method shows promise for discovering novel protein partners and characterizing their interactions.
• This work facilitates the use of abundant protein structural models for biological discovery.