Negatome

Negatome compiles protein and protein-domain pairs that are unlikely to form direct physical interactions, providing a curated set of negative examples for evaluation of experimental protein-protein interaction data and for training interaction-prediction algorithms.

Key Features:

• Data Sources: Integrates experimentally supported non-interacting protein pairs from manual curation of the scientific literature and from analysis of protein complexes with known three-dimensional structures.
• Stringent Filtering: Employs criteria to exclude interactions detected by high-throughput methods, increasing confidence in the non-interaction status of listed pairs.
• Domain-Specific Data: Includes non-interacting protein domains derived from manual curation and structural analysis datasets.

Scientific Applications:

• Experimental Evaluation: Enables assessment of new experimental findings on protein and domain interactions by comparison against the curated list of known non-interactions.
• Algorithm Training: Provides robust negative examples for training and benchmarking computational models that predict protein-protein interactions, offering a set less biased than negatives generated by random selection based on cellular location.

Methodology:

Manual curation of the literature, analysis of protein complexes with established three-dimensional structures to identify spatially separated non-interacting pairs, and exclusion of data from high-throughput interaction-detection methods.