ZEPPI: Proteome-scale sequence-based evaluation of protein-protein interaction models

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Evaluating structural models of protein complexes is crucial for understanding biological functions.
• Existing methods often focus on model quality rather than direct interaction evidence.
• Incorporating evolutionary information can improve the accuracy of protein-protein interaction (PPI) predictions.

Purpose of the Study:

• To introduce ZEPPI (Z-score Evaluation of Protein-Protein Interfaces), a novel framework for evaluating structural models of protein complexes.
• To leverage sequence coevolution and conservation data from interfacial residues to assess PPIs.
• To enhance existing PPI prediction databases like PrePPI with evolutionary insights.

Main Methods:

• ZEPPI calculates a score by comparing interface metrics from structural models to those of random residues.
• It utilizes species-paired multiple sequence alignments, focusing on interfacial residues to enable use of shallow alignments.
• The framework is applied proteome-wide to millions of dimeric complex models in the E. coli and human interactomes from the PrePPI database.

Main Results:

• ZEPPI successfully incorporates evolutionary information into the PrePPI scoring function.
• Performance evaluation on CASP-CAPRI decoys and experimental complexes demonstrates ZEPPI's utility.
• ZEPPI identifies weak PPI signals missed by standard CAPRI scores and low-confidence PPIs from PrePPI.

Conclusions:

• ZEPPI provides a robust method for evaluating structural models of protein complexes using evolutionary data.
• The integration of ZEPPI with PrePPI offers a powerful tool for generating functional hypotheses about protein interactions.
• ZEPPI's ability to detect subtle evolutionary signals enhances the prediction and validation of protein-protein interactions.