Argo: enabling the development of bespoke workflows and services for disease annotation

Area of Science:

• Computational biology
• Bioinformatics
• Natural Language Processing

Background:

• The increasing need for understanding disease etiology, diagnosis, and treatment necessitates efficient curation of medical literature.
• The Fifth BioCreative Challenge (BioCreative V) highlighted the importance of mining literature for disease-relevant information, including disease annotations and chemical-disease relations (CDR).

Purpose of the Study:

• To present Argo, a text mining workbench, and demonstrate its application in literature-based disease annotation.
• To evaluate Argo's suitability for semi-automatic curation of chronic obstructive pulmonary disease (COPD) phenotypes within the BioCreative V User Interactive Track (IAT).
• To showcase Argo's role in developing high-performing web services for normalizing disease mentions.

Main Methods:

• Utilizing Argo's generic, configurable components to build interoperable processing workflows.
• Applying Argo's graphical annotation interface for domain experts to curate automatically generated annotations.
• Developing and applying machine learning-based concept recognition models within bespoke workflows.
• Integrating various databases and enabling user-interactive correction of annotations.

Main Results:

• Argo facilitated the semi-automatic curation of COPD phenotypes, demonstrating its utility in the BioCreative V IAT.
• Argo supported the development of top-performing web services for the CDR track, specifically for normalizing disease mentions against the Medical Subject Headings (MeSH) database.
• The workbench proved versatile in supporting diverse workflows, from database integration to machine learning model application and interactive curation.

Conclusions:

• Argo is a powerful and flexible text mining workbench enabling the development of customized solutions for literature analysis.
• The workbench shows significant potential as an enabling technology for curating complex biological and phenotypic information from scientific literature.
• Argo's capabilities support both automated and human-in-the-loop approaches to information extraction and curation.