Annotating gene sets by mining large literature collections with protein networks

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Identifying gene functions is crucial for understanding human disease.
• Current methods for analyzing gene sets often lack comprehensive functional insights.

Purpose of the Study:

• To develop an integrative natural language processing (NLP) system for inferring gene set functions.
• To improve the discovery of novel disease-associated pathways using literature and network analysis.

Main Methods:

• Developed an NLP system integrating scientific literature mining with biological networks.
• Constructed a heterogeneous network linking genes, literature phrases, and protein interactions.
• Inferred multiscale functional annotations based on network distances and visualized them as a biological concept ontology.

Main Results:

• The system significantly improved function prediction for known gene sets compared to baseline text-mining.
• Discovered novel functional annotations for gene sets and pathways lacking prior functional information.
• Demonstrated utility through case studies identifying new cancer-related pathways with ontological annotations.

Conclusions:

• The integrative NLP system effectively infers gene functions and aids in discovering novel disease pathways.
• This approach offers a powerful tool for advancing biological discovery, particularly in cancer research.