GeDex

GeDex extracts consensus gene-disease associations from biomedical literature to identify and support curation of gene-disease relationships.

Key Features:

• Consensus Prediction Approach: Reports a single consensus prediction for gene-disease associations supported by multiple sentences within the literature.
• Predictive Model (four features): Employs a predictive model trained using four simple features to classify gene-disease relationships.
• Performance and Validation: Achieved an F-score of 0.77 on the curated fraction of DisGeNet and an F-score of 0.74 on a manually curated dataset.
• Discovery of Novel Associations: Identified associations not present in current databases when applied to a corpus of articles on chronic pulmonary diseases.

Scientific Applications:

• Database curation: Supports curation and expansion of databases such as DisGeNet by extracting and aggregating gene-disease associations from literature.
• Literature mining for novel associations: Enables discovery of previously unreported gene-disease associations in disease-specific article collections, e.g., chronic pulmonary diseases.
• Mapping genetic contributions: Aids efforts to map genetic mechanisms underlying human diseases by providing aggregated evidence from multiple sentences and articles.

Methodology:

Aggregates evidence across multiple sentences to produce a single consensus prediction and applies a predictive model trained on four features, with validation against the curated fraction of DisGeNet and a manually curated dataset.