The volume of scientific literature being published has increased dramatically in the digital age. Ontologies and taxonomies are important tools to help researchers retrieve and understand this overwhelming amount of scientific literature, but using and managing ontologies can be challenging in itself.

In this whitepaper written by SciBite and our partner Copyright Clearance Center, we’re looking at the history of biomedical classification and how these systems have evolved to address new technology and use cases. We’ll explain the difference between taxonomies and ontologies, and discuss the challenges and successes that come with adopting and managing ontologies.

Challenges when using and managing ontologies

Ontologies can now be used for many applications, including search, data analysis, indexing and information sharing. The use of ontologies, potentially across structured and unstructured content, across a range of domains, opens possibilities far beyond Boolean searching of pooled citations. For example:

• Synonyms contributed by multiple vocabularies can be added to queries automatically, expanding retrieval while maintaining relevance.
• Synonyms and standardized or related terms suggested to a user can decrease abandoned queries and increase the length and activity of search sessions. It is this synonym support and expansion that enables the structuring of data, and makes ontologies more fit for text mining and analytics.
• Extending managed vocabularies to unstructured text increases the efficiency and reach of pharmacovigilance queries.
• The relationships embedded in advanced ontologies allow users to mine for unexpected co-occurrences and suggest novel uses for existing drugs, or similarities between diseases or functions that might point to underlying processes.

However, adopting and managing ontologies to achieve these benefits can be a challenge. To make use of public ontologies and taxonomies – such as MeSH, MedDRA, SNOMED, and NCI Thesaurus – in named entity recognition scenarios, for example, these must be tailored for use through expert manual curation or automated processes. These and private ontologies and taxonomies require high-throughput software to convert unstructured text into machine-readable information annotated with ontology terms.

Managing ontologies across an organization can also be a challenge. For example, if you use public ontologies modified locally, how do you then reconcile changes made in the original ontology with your modifications? How do you allow subject matter experts and users to contribute changes to your ontology safely and quickly?

Different types of ontology users need different permissions to suggest changes or make changes directly and immediately. Versioning, local changes, and audit trails must also be accounted for in productionized operations.

Summary

The volume of scientific literature has exploded. The value of searching across genetic, clinical, anatomy, and biochemical resources at once has shaped the new paradigm for research and development. Biomedical ontologies are a necessary tool to harness big data so that both humans and machines can advance our scientific understanding.