PrePPI

PrePPI assigns Bayesian-derived probabilities to predicted and experimentally determined protein-protein interactions for yeast and human, integrating structural, functional, evolutionary, and expression evidence.

Key Features:

• Integration of predicted and experimental data: Predicted PPIs are combined with experimentally validated interactions to provide a comprehensive interaction dataset for yeast and human.
• Bayesian probability assignment: Each interaction receives a probability score computed from likelihood ratios (LRs) derived from multiple data sources.
• Multimodal evidence: Structural, functional, evolutionary, and expression information are incorporated as features that contribute to the LR calculations, with structural information playing a pivotal role.
• Compilation of experimental interactions: Experimentally determined PPIs are aggregated from public databases and contribute LRs that influence final interaction probabilities.
• High-confidence interaction set: The database contains approximately 2 million PPIs with a subset of high-confidence interactions (probability > 0.5), including ~60,000 for yeast and ~370,000 for human.
• Structural models: Structural models are provided for many PPIs to represent molecular interfaces underlying predicted and experimental interactions.

Scientific Applications:

• Functional genomics: High-confidence PPIs support inference of functional relationships between genes and proteins.
• Drug discovery: Interaction networks and structural models aid identification of potential drug targets and mechanistic hypotheses for therapeutic compounds.
• Systems biology: Extensive PPI data enable construction and analysis of protein interaction networks for systems-level studies of cellular processes.

Methodology:

PrePPI applies a Bayesian framework that computes likelihood ratios from structural, functional, evolutionary, and expression features and integrates LRs from aggregated experimental PPI databases to assign a probability to each interaction.