Ferret: a sentence-based literature scanning system.

Area of Science:

• Biomedical Informatics
• Genomics
• Bioinformatics

Background:

• The exponential growth of bioscience literature presents a significant challenge for researchers seeking relevant information.
• MEDLINE contains over 20 million citations, with new additions annually, necessitating advanced retrieval systems.
• Existing systems struggle to pinpoint specific gene-centric relationships within this vast dataset.

Purpose of the Study:

• To develop and evaluate Ferret, a prototype retrieval system for identifying and ranking gene-centric relationships.
• To assist scientists in navigating complex biomedical literature for gene discovery and analysis.
• To address challenges like gene name ambiguity and facilitate query expansion.

Main Methods:

• Ferret processes inputs including gene lists and keywords to retrieve relevant sentences and documents.
• The system handles gene name ambiguity and expands queries for broader search capabilities.
• Results are visualized using a heatmap, allowing users to explore species-specific gene information across different research stages.

Main Results:

• Ferret successfully identified gene families (e.g., bZIP) related to specific phenotypes in Arabidopsis thaliana.
• The system uncovered cross-species associations, linking Arabidopsis genes with information from potato and jasmine.
• Case studies demonstrated Ferret's ability to generate novel research questions, such as those involving salicylic acid in plant stress responses.

Conclusions:

• Ferret effectively retrieves relevant gene-centric sentences from PubMed, demonstrating its utility.
• The system's performance in three plant biology case studies indicates high end-user satisfaction.
• Ferret shows promise as a valuable tool for accelerating bioscience research and discovery.