PEPPI

PEPPI predicts proteome-wide protein-protein interactions by integrating structural similarity, sequence similarity, and functional association data to improve interaction identification across proteomes.

Key Features:

• Integration of Diverse Data Sources: Combines structural similarity, sequence similarity, and functional association data to create a multi-faceted basis for interaction prediction.
• Machine Learning-Based Classification: Uses a naïve Bayesian classifier to process and integrate diverse datasets for refined interaction scoring.
• Benchmarking Performance: Evaluated on 798 ground truth interactions and 798 non-interactions, achieving a 4.5% higher AUROC than comparator methods.

Scientific Applications:

• Host–Pathogen Interaction Mapping: Applied to SARS-CoV-2–human interaction mapping, identifying 403 high-confidence interactions that covered 73% of a PSICQUIC gold standard and showed complementarity with recent high-throughput experimental data.

Methodology:

Assess structural and sequence similarities alongside functional association data, process integrated features with a naïve Bayesian classifier, and validate predictions through benchmarking against established interaction datasets.