eGenPub, a text mining system for extending computationally mapped bibliography for UniProt Knowledgebase by...

Area of Science:

• Bioinformatics
• Computational Biology
• Proteomics

Background:

• UniProt Knowledgebase (UniProtKB) is a crucial resource for protein sequence and functional data.
• Expanding the bibliography for protein entries enhances research context and discoverability.
• Current methods for linking publications to protein entries may be limited.

Purpose of the Study:

• To develop and evaluate eGenPub, a text mining system for automatically identifying relevant literature for UniProt protein entries.
• To computationally map additional bibliography for plant proteins within the UniProtKB.
• To integrate this new bibliography into the UniProt production pipeline.

Main Methods:

• Development of eGenPub, a text mining system utilizing a gene normalization tool (pGenN).
• Training a support vector machine model to predict article relevance with 95.3% precision.
• Full-scale PubMed processing for eight common plant species.

Main Results:

• eGenPub identified 9025 relevant articles for 4752 UniProt entries.
• 5252 of these articles represent new, computationally mapped bibliography not previously in UniProtKB.
• The system demonstrated high precision in linking articles to specific protein entries.

Conclusions:

• eGenPub effectively expands the bibliography associated with UniProt protein entries.
• The computationally mapped additional bibliography enhances the informational scope of UniProtKB.
• This automated approach improves the discoverability of protein-related research literature.