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

Abstract

Argo (http://argo.nactem.ac.uk) is a generic text mining workbench that can cater to a variety of use cases, including the semi-automatic annotation of literature. It enables its technical users to build their own customised text mining solutions by providing a wide array of interoperable and configurable elementary components that can be seamlessly integrated into processing workflows. With Argo's graphical annotation interface, domain experts can then make use of the workflows' automatically generated output to curate information of interest.With the continuously rising need to understand the aetiology of diseases as well as the demand for their informed diagnosis and personalised treatment, the curation of disease-relevant information from medical and clinical documents has become an indispensable scientific activity. In the Fifth BioCreative Challenge Evaluation Workshop (BioCreative V), there was substantial interest in the mining of literature for disease-relevant information. Apart from a panel discussion focussed on disease annotations, the chemical-disease relations (CDR) track was also organised to foster the sharing and advancement of disease annotation tools and resources.This article presents the application of Argo's capabilities to the literature-based annotation of diseases. As part of our participation in BioCreative V's User Interactive Track (IAT), we demonstrated and evaluated Argo's suitability to the semi-automatic curation of chronic obstructive pulmonary disease (COPD) phenotypes. Furthermore, the workbench facilitated the development of some of the CDR track's top-performing web services for normalising disease mentions against the Medical Subject Headings (MeSH) database. In this work, we highlight Argo's support for developing various types of bespoke workflows ranging from ones which enabled us to easily incorporate information from various databases, to those which train and apply machine learning-based concept recognition models, through to user-interactive ones which allow human curators to manually provide their corrections to automatically generated annotations. Our participation in the BioCreative V challenges shows Argo's potential as an enabling technology for curating disease and phenotypic information from literature.Database URL: http://argo.nactem.ac.uk.