PPIDomainMiner

PPIDomainMiner infers domain-domain interactions (DDIs) from multiple protein-protein interaction (PPI) sources to predict and prioritize interacting protein domains.

Key Features:

• Tripartite Graph Modeling: Extends the CODAC method to perform inference in a tripartite graph setting for predicting DDIs.
• Vector Similarity Techniques: Applies vector similarity methods for link prediction within the graph model.
• Multiple PPI Sources Integration: Aggregates data from seven PPI resources and combines individual source scores using an optimized weighted average.
• Statistical Significance Assessment: Assigns p-values to inferred DDIs and classifies them into Gold, Silver, or Bronze confidence categories.
• Extensive DDI Dataset: Produces a dataset of 84,552 non-redundant DDIs.
• Validation against Databases: Compares and validates predictions against 3did, IMEx-curated datasets, and STRING non-curated interactions.

Scientific Applications:

• Protein Interaction Analysis: Provides predicted DDIs to support analysis of protein functions and molecular interaction mechanisms.
• Hypothesis Generation: Generates hypotheses about potential protein partners and domain-mediated interactions for experimental testing.
• Integration with High-Throughput Methods: Supports integration with cross-linking mass spectrometry to identify plausible protein partners.

Methodology:

Performs tripartite graph modeling extending CODAC, uses vector similarity-based link prediction, aggregates seven PPI sources via an optimized weighted average, and assigns p-values to inferred DDIs for confidence classification.