Reproducible interactome

Reproducible interactome identifies and removes explicit and implicit redundancies in aggregated protein–protein interaction (PPI) meta-databases to produce a reproducible interactome containing higher-confidence interactions.

Key Features:

• Redundancy Detection: Distinguishes explicit redundancies identified by aggregation processes and implicit redundancies arising when the same interaction is curated across different primary databases or within entries of a single database without being recognized as redundant.
• Semantic Web Technologies: Utilizes the Molecular Interaction ontology and Semantic Web approaches to detect redundancies that traditional methods may overlook.
• Reproducible Interactome Construction: Constructs a reproducible interactome that includes only interactions verified by multiple detection methods or multiple publications.
• Impact on Dataset Size: Removal of redundancies reduces dataset size substantially, reported as 59% reduction for yeast and 56% reduction for human datasets.
• Quantification of Implicit Redundancy in APID: Identifies that approximately 15% of entries in the Agile Protein Interactomes DataServer (APID) meta-database are implicitly redundant, with over 90% of these redundancies resulting from aggregations across distinct primary databases.

Scientific Applications:

• Enhanced Data Reliability: Produces PPI datasets with fewer redundant entries, reducing bias in downstream analyses.
• Confidence Metrics Improvement: Improves the interpretability of confidence-related metrics by ensuring they reflect reproducibility across methods or publications.
• Cross-Species Analysis: Applicable to multiple organisms, demonstrated on yeast and human PPI datasets.

Methodology:

Applies Semantic Web technologies, leveraging the Molecular Interaction ontology, to the Agile Protein Interactomes DataServer (APID) meta-database to detect implicit redundancies; this analysis found ~15% implicit redundancy in APID and that >90% of these cases arise from aggregation across distinct primary databases.