Predicting protein-peptide interaction sites using distant protein complexes as structural templates.

Area of Science:

• Molecular biology
• Biochemistry
• Computational biology

Background:

• Protein-peptide interactions are vital for cellular processes and implicated in human diseases.
• Experimental study of these interactions is challenging due to peptide flexibility and transient binding.
• Computational prediction of protein-peptide binding sites is necessary for detailed molecular understanding.

Purpose of the Study:

• To develop and present InterPep, a novel computational pipeline for predicting protein-peptide interaction sites.
• To provide a tool that aids in understanding and potentially regulating cellular functions and diseases linked to protein-peptide interactions.

Main Methods:

• InterPep utilizes a pipeline integrating structural template matching, sequence information, random forest machine learning, and hierarchical clustering.
• The pipeline predicts likely peptide binding regions on protein structures given a protein structure and peptide sequence.

Main Results:

• InterPep successfully identified 50.7% of binding sites at rank 1 and 69.3% within the top five predictions on experimentally determined structures.
• Performance was evaluated using structures with no significant sequence similarity as templates.
• The tool demonstrated 80% precision at 20% recall, indicating high accuracy in identifying true binding sites.

Conclusions:

• InterPep is a powerful and accurate tool for identifying peptide-binding sites on protein structures.
• Its high precision makes it a valuable starting point for experimental investigations and docking protocols.
• The source code is publicly available for broader scientific use.