Software Solutions Product team connects communities, improves resources, and boosts potential for vaccine R&D

With the increasing use of AI and large language models (LLMs) to assist in analysis, clean data is essential to ensure models are grounded in evidence-based knowledge. The use of ontologies is key to transforming unstructured data into high-quality material. Only then can decision-makers have confidence to pursue a line of research that is based on reliable, reproducible, and accurate information.

Scientific curators from Elsevier’s Software Solutions SciBite team have helped connect the biomedical, pharmaceutical, and regulatory communities to a robust, standards-based data resource for vaccines. Paola Roncaglia and Rachael Huntley are co-authors on a Nature Scientific Dataopens in new tab/window paper detailing the Vaccine Ontologyopens in new tab/window (VO). The work highlights how VO provides a community-curated, machine-readable map that standardizes vaccines, their components, vaccine-induced immune responses, and related vaccine investigations.

The goal: enable data from diverse sources to be interoperable, to accelerate discovery, and to support safe and effective vaccine development, made even more relevant by the recent outbreaks of Hantavirus, Ebola virus, and coronavirus infections.

Ontologies are essential, and new ones need expert curation

Ontologies – representations of the types of ‘things’ in a given domain, and the relationships between them – are designed in a language that both humans and computers can understand. They have a fundamental role in grounding LLMs with the structured terminology providing a foundation of scientific knowledge. LLMs leveraging reference ontologies promotes evidence-based, reproducible outputs in return.

Life sciences have a rich coverage of publicly available ontologies serving many domains, but creating new ontologies is time-consuming and involves extensive manual effort by expert curators to make an accurate and complete representation of a domain. Therefore, the development of a large ontology, such as VO, benefits from a shared effort, in return generating a community-agreed standard.

Why is the Vaccine Ontology so important?

VO is a comprehensive, evolving ontology that covers more than 10,000 vaccines, targeting hundreds of pathogens and cancers in humans and more than 30 non-human species. VO links to external resources such as RxNorm (the U.S. drug naming and mapping system), CVX (the vaccine-administered code set), FDA, and USDA. By providing a shared vocabulary and structure, VO serves as a powerful backbone for data standards, literature mining, and AI-driven analyses across major resources like the VIOLIN vaccine database, the ImmPort immunology data portal, and the Vaccine Adjuvant Compendium (VAC).

How Elsevier is enabling access to VO

Organizations can access the Vaccine Ontology through multiple formats via Elsevier Software Solutions, enabling a range of downstream applications—from electronic lab notebooks (ELNs) to form filling and automated annotation. The original source ontology is provided in the ontology management platform CENtree, allowing researchers to augment VO content as needed while preserving a shared core.

A targeted vocabulary for rapid information extraction

To support fast, accurate data extraction, Elsevier built a Named Entity Recognition (NER) vocabulary focused on the “vaccine adjuvant” branch of VO, which contains information about substances that increase or modulate the immune response to a vaccine. This enables customers to quickly identify relevant vaccine information during development, safety monitoring, and post-production activities.

The Elsevier vaccine ecosystem now also includes a larger vocabulary based on the wider vaccine-specific content of VO, including vaccines, vaccine components, and vaccine allergens. This new vocabulary has also been NER-optimized to extend semantic search capabilities and complement the adjuvant vocabulary.

Community-driven curation and feedback

The Vaccine Ontology project thrives on community input, and in turn, high-quality resources allow Elsevier to deliver aligned, quality content. The Elsevier curators have augmented VO with custom entities and synonyms and have fed back suggestions for new classes to VO developers for inclusion in the public ontology. This collaborative loop helps VO stay aligned with real-world needs while remaining interoperable with other public ontologies.

Why this matters now

In an era of rapid vaccine development and complex data landscapes, a shared, machine-readable ontology like VO helps ensure that data from different studies, databases, and regulatory environments can be connected and analyzed together. By combining community curation, practical tooling, and alignment with widely used external vocabularies, VO supports safer vaccines, faster discoveries, and more transparent, sharable science.

Ongoing partnerships and broader impact

Beyond VO, the Elsevier Scientific Curation team contributes to improvements in multiple public ontologies and data resources, including HGNC (the resource for approved human gene nomenclature), Human Phenotype Ontology (HP), Experimental Factor Ontology (EFO), Gene Ontology (GO), and the Mondo disease ontology.

These collaborations support FAIR data integration and enable computer-assisted analyses that can provide high-quality, reliable insights to support decision making and thus accelerate the development of medicines and therapies.

Related articles

Why ontologies – and fruit flies – matter in the age of AI

Harnessing ontologies for pharma: Dr Jane Lomax on the synergy of AI and scientific expertise