PAIRpred: partner-specific prediction of interacting residues from sequence and structure

Area of Science:

• Computational biology
• Structural bioinformatics
• Machine learning in bioinformatics

Background:

• Predicting protein-protein interaction (PPI) sites is crucial for understanding biological processes.
• Existing methods often focus on single proteins, limiting their accuracy for complex interfaces.

Purpose of the Study:

• To develop a novel partner-specific method for predicting protein-protein interaction sites.
• To improve the accuracy of identifying inter-protein residue contacts and binding sites.

Main Methods:

• Developed PAIRpred, a machine learning method using pairwise kernels and support vector machines.
• Incorporated sequence and structure information of residue pairs from both interacting proteins.
• Trained and evaluated the method on benchmark datasets (Docking Benchmark 4.0) and CAPRI targets.

Main Results:

• PAIRpred achieves state-of-the-art accuracy in predicting protein-level binding sites and inter-protein residue contacts.
• Demonstrated robustness to binding-associated conformational changes compared to existing methods.
• Case study showed utility in analyzing specific protein interactions, like ISG15-NS1.

Conclusions:

• PAIRpred offers a more detailed and accurate model for protein binding interactions.
• The partner-specific approach enhances prediction of interaction interfaces.
• The method has practical applications in analyzing protein complex specificity.