GeneCup

GeneCup mines PubMed abstracts to extract sentences where user-specified gene symbols and ontology-organized keywords co-occur, facilitating integration and interpretation of omics results and incorporation of human GWAS associations.

Key Features:

• Ontology-Based Keyword Organization: Accepts a hierarchical, extensible ontology of keywords to structure searches and improve relevance of retrieved sentences.
• Integration with GWAS data: Incorporates results from human genome-wide association studies into the keyword ontology to link genetic associations with literature evidence.
• Sentence Co-occurrence Extraction: Automatically extracts sentences from PubMed abstracts where specified gene symbols and keywords co-occur.
• Convolutional Neural Network Classification: Uses a convolutional neural network to distinguish sentences describing systemic stress from those describing cellular stress.
• PubMed querying and local archive retrieval: Queries the PubMed server via a programming interface and retrieves abstracts from a local copy of the PubMed archive.

Scientific Applications:

• Integration of omics results with literature: Links omics study findings to relevant PubMed evidence through gene–keyword sentence co-occurrence and GWAS incorporation.
• GWAS annotation and interpretation: Associates GWAS-derived genetic signals with literature contexts to support interpretation of genetic associations.
• Drug addiction research: Supports domain-specific investigations, exemplified by a keyword ontology containing over 300 terms for addiction-related research.

Methodology:

Queries the PubMed server via a programming interface to retrieve abstracts from a local copy of the PubMed archive, processes abstracts to extract sentences where specified gene symbols and keywords co-occur, and applies a convolutional neural network to classify sentences as describing systemic versus cellular stress.