PubTator central: automated concept annotation for biomedical full text articles.

Area of Science:

• Biomedical Informatics
• Text Mining
• BioNLP

Background:

• PubTator Central (PTC) is a web service for retrieving bioconcept annotations from biomedical literature.
• Existing PTC annotates PubMed abstracts and PMC full text articles using text mining systems.
• Previous versions focused on abstracts, limiting comprehensive data retrieval.

Purpose of the Study:

• To introduce the enhanced PubTator Central (PTC) web service with expanded full text article annotations.
• To improve the accuracy and accessibility of biomedical concept annotations.
• To enable new downstream applications and enhance existing research workflows.

Main Methods:

• Automated annotation of genes/proteins, genetic variants, diseases, chemicals, species, and cell lines.
• Integration of state-of-the-art text mining and deep learning-based disambiguation modules.
• Development of a new web interface for document collection building and annotation visualization.
• Provision of annotations via online interface, RESTful API, and bulk FTP in multiple formats (XML, JSON, tab-delimited).

Main Results:

• PTC now annotates 3 million full text articles from the PMC Text Mining subset, significantly increasing biomedical concept coverage.
• Enhanced annotation accuracy achieved through improved concept identification and a new deep learning disambiguation module.
• A faster server-side architecture and daily synchronization with PubMed and PubMed Central ensure up-to-date data.

Conclusions:

• The expanded full text annotation in PTC substantially increases biomedical concept coverage.
• This enhancement is anticipated to improve existing applications like biocuration and gene prioritization.
• The expanded PTC is expected to enable novel use cases in literature-based knowledge discovery and biomedical research.