Background & Summary

The Ontology of Adverse Events (OAE) is a community-based biomedical ontology that is designed to be interoperable and logically well-formed and to incorporate accurate representations of adverse events (AEs) associated with medical interventions (e.g., drug administration and vaccination). As a comprehensive resource of AEs, OAE enables computational inference and sophisticated algorithms that allow AE data to be integrated with other data (e.g., phenotypic data and genomic data) and then support combined AE profile and pathogenetic mechanism analyses. Nowadays, OAE is available without restriction for public view at various ontology linked data resources including OntoBee1 (default OAE linked server), BioPortal2, OLS3, and AberOWL4.

To improve representation and organization of AE information, some vocabulary resources have been developed and widely used, including the Medical Dictionary for Regulatory Activities (MedDRA)5, the Common Terminology Criteria for Adverse Events (CTCAE)6, and the World Health Organization (WHO)’s Adverse Reaction Terminology (WHO-ART)7. The OAE differs from these available AE coding vocabularies in several crucial ways. First, the OAE has provided more logically well-formed definitions and a more structured classification of AEs than any other coding vocabulary. Second, the OAE is not a simple terminology, but rather a full community-based biomedical ontology that follows the Web Ontology Language (OWL) and the Open Biological and Biomedical Ontologies (OBO) Foundry principles8, thus serving as a computational resource that supports computer-assisted reasoning and analysis. Third, OAE represents AEs based on patient anatomic regions and clinical outcomes, as well as taking into account other elements (e.g., patient age, patient gender, and current illness) of the process that leads from initial medical intervention to subsequent outcomes, all of which assist in realizing precision medicine.

Generally speaking, OAE has experienced the following three stages of development. In the first stage, OAE was initially developed as an ongoing realism-based effort that aimed to reduce the confusion in existing AE terminologies and was subsequently published at the 2nd International Conference on Biomedical Ontology (ICBO-2011)9. In the second stage, OAE redefined the key term ‘adverse event’ (OAE_0000001) as a process subsequent to a medical intervention while not assuming any causal association between them. The term ‘causal adverse event’ (OAE_0000003) was newly created to represent the existence of data of a certain sort, namely data establishing a causal association between adverse events and medical interventions10. Within the OAE framework, a clear distinction is made between an adverse event and a causal adverse event, where the latter is defined as a specific subtype of the former. This change made the OAE more robust as a representation of the domain of AE reporting. In the third stage, OAE-based methods are widely used in the field of medical informatics and play an increasingly important role in representing, classifying, and analyzing the AE information. The OAE continuously evolves and improves through the integration of new terms derived from its practical applications.

This study systematically examines the structural and functional evolution of the OAE over the past decade, with a dual focus on delineating major ontological advancements and conducting a comprehensive review and summary of existing OAE-based research. Furthermore, we also report on recent efforts to translate the OAE into Chinese and introduce it into the monitoring and analysis for adverse events following immunization (AEFI) in China.

Extend the breadth and scope of OAE

The OAE is a community-driven biomedical ontology that ontologically organizes knowledge domains related to adverse medical occurrences. Its primary objective is to systematically represent AE manifestations across diverse anatomical locations and to logically classify adverse outcomes into distinct categories such as symptoms, clinical signs, and pathological processes (Fig. 1). As of 31 December 2024, the OAE has contained 10,589 AE-related terms (Fig. 2 and Table 1), reflecting an annual growth rate of 13.9% since the last Journal of Biomedical Semantics article published in 201410.

Fig. 1
Fig. 1
Full size image

General OAE design pattern.

Fig. 2
Fig. 2
Full size image

Evolution of OAE terminology. This visualization shows the distribution of top-level terms within the OAE. The dark-blue portion represents new terms introduced following the initial publication10, corresponding to the period from 20 April 2014 to 31 December 2024.

Table 1 Summary of ontology terms in the OAE, including those imported from external ontologies, as of December 31, 2024.
Full size table

AE data acquisition

The newly added terms were derived from mining the existing knowledge bases of the corresponding entries, including:

  1. 1)

    Clinical AE reports in the Vaccine Adverse Events Reporting System (VAERS) and the Food and Drug Administration (FDA) Adverse Events Reporting System (FAERS). VAERS is a national early warning system to detect possible safety problems in U.S.-licensed vaccines. It accepts and analyzes reports of AEs (possible side effects) after a person has received a vaccination. Anyone can report an AE to VAERS11. The FAERS database contains AE reports, medication error reports and product quality complaints resulting in AEs that were submitted to FDA. The database is designed to support the FDA’s post-marketing safety surveillance program for drug and therapeutic biologic products12.

  2. 2)

    Symptoms, signs, and adverse drug reactions that are cataloged within the MedDRA terminology system. MedDRA is a rich and highly specific standardized medical terminology developed by the International Conference on Harmonisation (ICH) to facilitate sharing of regulatory information internationally for medical products (e.g., pharmaceuticals, vaccines, and drug-device combination products) used by humans5. We have referenced MedDRA in our OAE development by cross-referencing related MedDRA identifiers (Fig. 3). The data models of AE analysis provided by the Clinical Data Interchange Standards Consortium (CDISC) were also referenced13.

    Fig. 3
    Fig. 3
    Full size image

    Demonstration of standardized annotation for a new term. The annotation for ‘acute abdomen AE’ (OAE_0004600) is presented as an example, following the defined annotation schema of the OAE.

  3. 3)

    Adverse events identified from OAE-based scientific researches and peer-reviewed journal articles. Numerous pharmacovigilance investigations have employed OAE methodologies for systematic analysis. Some AEs revealed in these studies but not yet included in the OAE terminology system would be identified when OAE-based hierarchical classification was conducted. For example, in our researches on AEs associated with hepatitis A/B and brucellosis vaccines14,15, we successfully identified and subsequently incorporated more than 70 novel AE terms into OAE. To supplement and validate term definitions, we drew on AE information from journal articles and expert consultations, particularly for rare events, which was then incorporated into the OAE as supporting references.

AE data annotation

The precise annotation of terminology constitutes a critical element in quality control during ontology development. The OAE provides annotations for AE terms defined by two major patterns:

  1. 1)

    Import terms that are available in established ontologies. To ensure alignment with the OBO Foundry Principles and facilitate ontology term reuse and interoperability16, the existing ontological terms are systematically integrated into the OAE framework through direct importation mechanisms (Table 1). For example, the complete ontological architecture of the Basic Formal Ontology (BFO) has been imported into the OAE as its foundational upper-level ontology17, demonstrating methodological consistency with established ontological integration protocols. The annotations and original identifiers (IDs) of imported individual ontology terms are preserved through implementation of the OntoFox software18, while maintaining the semantic relationships between these entity terms and their corresponding property associations.

  2. 2)

    Manual annotation of new AE terms. As mentioned above, the annotation comprehensiveness of AE terms should be maximized following established epistemological principles. Figure 4 outlines the foundational annotation scheme governing newly incorporated AE terms within the OAE framework, including label (required), definition (required), definition source (optional), see also (optional), term editor (required), alternative name (optional), subclass of (required), and occurs in (required). Furthermore, to facilitate communication and utilization of OAE among Chinese science and technology professionals, we have developed a “Chinese Translation” annotation option and translated the newly added AE terms into Chinese (Fig. 3). It is noteworthy that some imported AE terms remain undefined, requiring systematic redefinition of these terms by our research team. The definitions of AE terms within OAE are predominantly sourced from established knowledge bases and scholarly repositories, including but not limited to domain-specific databases (e.g., PubMed) and collaboratively curated knowledge platforms (e.g., Wikipedia). For AE-specific terms in OAE, new identifiers incorporating the “OAE_” prefix followed by a seven-digit auto-incremental numerical sequence were generated. For example, the term ‘acute abdomen AE’ is assigned the identifier OAE_0004600, with its complete metadata accessible through the OntoBee ontology browser via its Internationalized Resource Identifier (IRI)1: http://purl.obolibrary.org/obo/OAE_0004600.

    Fig. 4
    Fig. 4
    Full size image

    Annotation schema and content of OAE terminology.

As presented in Table 1, OAE comprises 10,829 formally defined representational units. The OAE-specific terms include 9,397 distinct classes alongside 13 ontological properties, specifically categorized as eight object properties, two datatype properties, and three annotation properties. Comparative analysis reveals substantial terminological expansion, with 7,743 novel terms (representing a 251% increase) having been incorporated since the previous 2014 publication documenting 3,086 ontology terms. This quantitative growth demonstrates marked ontological development, reflecting both conceptual granularity enhancement and domain coverage extension.

Interoperability applications of OAE with other knowledge bases

In conjunction with other standardized terminologies and classification systems, the OAE and its user-friendly annotation framework enable semantic interoperability within the domain of digital medicine. The Disease Specific Ontology of Adverse Events (DSOAE) has been developed to understand and represent disease-intervention specific AEs for a specific domain such as chronic kidney disease (CKD) through the extension and adaptation of general ontologies (e.g., OAE) by incorporating domain-specific information and usage requirements19. With the support of DSOAE, a data-driven knowledge base for CKD-specific AEs has been established, and successfully applied in data analysis and integration.

Drug adverse events (DAEs) represent a critical focus area in pharmacovigilance, and substantial progress has been made in the development of domain-specific knowledge bases to systematically document these phenomena. Notably, ontology-driven frameworks have emerged as pivotal tools in this domain. For example, OAE-based DAE-specific ontologies, such as the Ontology of Drug Neuropathy Adverse Events (ODNAE) and the Ontology of Cardiovascular Drug Adverse Events (OCVDAE)20,21, have been developed to standardize the characterization of drug-associated neuropathy-inducing and cardiovascular-related DAEs, respectively. To logically represent the complex relations among drugs, drug components, mechanisms of action, AEs, age, diseases, and other related factors, an ontology design pattern was developed and implemented in the community-driven open-source Ontology of Drug Adverse Events (ODAE)22. The ODAE expands both the OAE and the National Drug File-Reference Terminology (NDF-RT), providing a generic representation of DAEs across varying conditions, which enables the systematic querying of scientific questions.

Since its establishment in 2011, the annotated terms and hierarchical architecture of OAE have been extensively cross-referenced by diverse domain-specific ontologies and databases23,24,25,26. The development and application of OAE-based ontology extensions, coupled with the sharing and integration of interoperable data with other ontologies or knowledge repositories, have collectively contributed to enhancing the depth, coverage, and sophistication of the OAE system.

Usage

The OAE has been extensively applied in research on adverse events associated with medical interventions. As of December 31, 2024, a total of 148 studies indexed in Google Scholar have cited the OAE methodology. In this section, we systematically categorize and review these studies, with a focus on representative cases to illustrate their applications.

OAE-based AE studies for vaccines

Vaccines are mostly preventive measures administered to healthy individuals for disease prevention, consequently imposing higher safety requirements and eliciting heightened public sensitivity towards associated adverse events. As mentioned above, the VAERS is a pioneering, comprehensive, and widely utilized surveillance system for monitoring vaccine adverse events (VAEs). The first OAE-based VAEs study was performed in 201227, which developed a Combinatorial, Ontology-based Detection of AE (CODAE) method to analyze the AE data associated with killed and live influenza vaccines sourced from the VAERS database. In this study, the statistically enriched VAEs were classified and analyzed using the hierarchical structure of OAE, MedDRA, and SNOMED-CT respectively. The OAE method provided better classification results than the other two methods.

The advantages of utilizing the OAE-based semantic web representation and data analysis approach were demonstrated to be remarkably effective in subsequent VAE studies, which incorporated more sophisticated statistical algorithms. For example, Xie et al. conducted an OAE-based systematic analysis of AE profiles associated with Hepatitis A/B vaccines, employing a logistic regression model accompanied with Markov Chain Monte Carlo (MCMC) sampling, and then identified 13 AEs (e.g., hepatosplenomegaly) attributable to the synergistic effects of vaccine-vaccine interactions (VVIs)14. This study represents the first report to analyze VVIs using clinically reported AE case data based on an ontological approach integrated with statistical modeling. Zi et al. adopted the CODAE method to evaluate AE profiles of three licensed human papillomavirus (HPV) vaccines (i.e., Cervarix, Gardasil, and Gardasil 9) using the reported data from VAERS28. Through application of a gender/age-stratified Gamma-Poisson Shrinker (GPS) model, 46 statistically significant serious adverse events (SAEs) were found to be associated with at least one HPV vaccine. The SAE profile was represented and classified via the OAE to enable mechanistic analysis.

It is noted that the findings derived from these AE surveillance studies should be interpreted with caution regarding causal attribution between vaccines and AEs owing to the methodological constraints inherent in spontaneously reported AE data29. Future research requires rigorous design and implementation of population-based epidemiological studies alongside controlled clinical trials to establish causal relationships between vaccines and specific AEs, with particular emphasis on characterizing the occurrence dynamics of SAEs.

OAE-based AE studies for drugs

In contrast to vaccines, which are typically administered to healthy individuals as prophylactic measures, the majority of drugs are prescribed to patients with pre-existing medical conditions. Responses to drug treatments, particularly with respect to AEs, may be significantly influenced by both the underlying disease pathophysiology and its clinical progression. Consequently, the systematic monitoring and analytic evaluation of drug adverse events (DAEs) present substantially greater methodological challenges compared to vaccine safety surveillance.

In the USA, the DAEs are monitored by the FAERS system, which mandates the application of the MedDRA as a standardized medical terminology system for classification and coding of reported DAEs. The FDA has been actively exploring the application of ontology-based systems pharmacology to enhance mechanistic analysis of DAEs and improve predictive toxicology assessments30,31. There are a number of specific projects and/or activities that have been launched or are currently in progress, such as creation of a standardized ontology for the universe of “molecular toxic targets”32, and development of a multiple-scale network linking both the molecular targets and organ toxicities to clinical drug toxicities using the MedDRA ontology25.

Given that the application of OAE in vaccine research has demonstrated promising interpretable outcomes, our research team has prioritized extending this analytical framework to the more complex domain of DAE analysis. A scalable auto-assistance platform for extraction of potential DAEs from medical case reports has been built by Gurulingappa et al.33. It relies on a high-quality corpus that was manually annotated using the OAE-driven methodology, and consequently has the ability to capture valid, under-reported, and novel potential DAEs not yet present in product leaflets. Sarntivijai et al. established a semantic framework to link biological mechanisms to phenotypes of DAEs by combining OAE with MedDRA in FAERS data analysis34. Through MedDRA-OAE term mapping, they integrated MedDRA-based DAEs associated with tyrosine kinase inhibitors (TKIs) and monoclonal antibodies (mAbs) into the OAE terminology system. This enabled subsequent semantic mapping to established biomedical ontologies, including the Human Phenotype Ontology (HP) for clinical manifestations35, the Uberon Anatomy Ontology (UBERON) for anatomical localization36, and the Gene Ontology (GO) for molecular functional annotations37. This multi-ontology integration approach effectively bridged pharmacological adverse event documentation with fundamental biological understanding. Hur et al. developed an ontology-driven literature mining system to detect DAEs from drug labels and investigate mechanistic pathways of neuropathy-inducing drugs (NIDs)38. Their OAE-based analytical approach further found that the identified 43 benzimidazole-related DAEs exhibited multisystem distribution patterns, with predominant manifestations observed in the behavioral and neurological, digestive, skin, and immune systems. This investigation substantiates that the integration of ontological text mining with network pharmacology constitutes a robust methodological framework for identifying drug-class-specific DAEs and elucidating their multiscale biological interconnections.

Advancing precision medicine implementation

It is widely acknowledged that precision medicine has significantly transformed modern medical practice. This innovative paradigm, originally coined to describe therapeutic strategies (e.g., routes of drug administration and vaccination) tailored to individual patients based on their unique genetic profiles, biomarker signatures, phenotypic characteristics, and psychosocial determinants, fundamentally differs from conventional approaches targeting patients with similar clinical manifestations39. The foremost and most formidable challenge in implementing precision medicine involves establishing a robust multidisciplinary framework and an integrated knowledge base. Such infrastructure must effectively bridge the translational gap between basic research discoveries and clinical applications, necessitating unprecedented collaboration among biomedical researchers, diagnostic laboratories, healthcare providers, and patient communities40.

Arguably, ontologies as formal knowledge representation systems serve as frameworks for integrating and analyzing heterogeneous clinical datasets alongside scientific evidence derived from diverse healthcare stakeholders. These structured vocabularies enable systematic data harmonization across disparate sources, crucially facilitating the standardization and interoperability of domain-specific data resources required for advancing precision medicine initiatives. Haendel et al. conducted a seminal investigation of ontologies, including the OAE, demonstrating their critical role in computational reasoning systems for enabling granular patient stratification41. Their research substantiated the utility of these structured knowledge systems across three key domains: diagnostic precision optimization, clinical care pathway personalization, and translational research methodologies. They conclusively established that ontologies facilitate multidimensional patient characterization through standardized semantic interoperability, thereby enabling phenotype-driven classification at clinically actionable resolutions. Furthermore, Xie et al. performed a systematic review examining precision vaccinology—an emerging discipline applying precision medicine principles to vaccine development—and its potential enhancement by ontological engineering, such as Vaccine Ontology (VO), OAE, and Ontology of Vaccine Adverse Events (OVAE)42. Based on these ontologies, they formulated a novel ontological framework strategy to advance precision vaccinology, proposing its systematic integration into vaccine target identification, mechanistic modeling of immune responses, and AE prediction during a vaccination campaign.

The significance of OAE in the surveillance of AEs associated with medical interventions in China

The current AE reporting systems for medical interventions in China exhibit limitations in terminology classification. Exemplified by the National Adverse Events Following Immunization Information System (CNAEFIS)43, it categorizes adverse events following immunization (AEFI) into seven broad types based on potential causative factors: common AEs, abnormal AEs, vaccine quality events, program errors, coincidental events, psychogenic reactions, and unexplained reactions. In addition, cases that cannot be definitively classified under these predefined categories are arbitrarily designated as “other” AEs. This oversimplified approach has resulted in the accumulation of substantial AEFI data without logical categorization or analytical coherence. In reality, AEFI manifest through diverse clinical presentations and etiological mechanisms. While partial adoption of ICD-9/10 coding standards has been implemented, the existing framework demonstrates insufficient systematic integration of diagnostic terminology, thereby compromising its capacity to support rigorous post-hoc analyses of AEs. This critical infrastructure gap underscores the urgent need to establish a standardized medical terminology classification system to enhance AE surveillance research.

Over the past decade, ontological engineering has evolved into a pivotal methodological paradigm for biomedical big data research and precision medicine applications. Its demonstrated capabilities in semantic integration and knowledge representation have attracted sustained attention from the Chinese scientific community. In 2018, the National Population and Health Science Data Sharing Platform spearheaded the formation of the China Biomedical Ontology Consortium (OntoChina, http://www.ontochina.org/). This initiative aims to construct ontology repositories for Chinese biomedical big data research through the adoption of international best practices in ontological engineering and cross-domain collaboration, while promoting ontological applications in national health informatization. Of particular significance is the foundational work by Dr. Yan Zhu’s research team at the China Academy of Chinese Medical Sciences, which completed the Chinese localization of the BFO44, thereby establishing a standardized upper-level framework for developing Chinese-language biomedical ontologies.

The OAE has been successfully applied in multiple studies analyzing medical intervention-associated AEs. Notably, our research team has methodologically integrated standardized Chinese terminological equivalents during OAE’s iterative development cycles. Building upon recent advancements in the localization of the BFO, the comprehensive sinicization of OAE, encompassing both granular terminological mapping and logical relationship adaptation, has now attained technical viability. In recent years, an increasing number of Chinese researchers have adopted the OAE as a standardized terminological framework to advance related studies. For instance, Li et al. utilized the OAE to perform a comprehensive analysis of DAE case reports associated with domestic anti-infection drugs, as illustrated in Fig. 545. Their study delineated the epidemiological characteristics of these DAEs, including demographic trends, drug administration routes, and severity profiles. These findings offer actionable recommendations to optimize evidence-based clinical practices for anti-infection therapies. By strategically utilizing OAE’s open-source architecture in conjunction with established ontological engineering platforms (e.g., OntoBee for ontology browsing1, OntoFox for term extraction18, and OntoRat for collaborative editing46), this initiative enables semantic interoperability between OAE and existing Chinese biomedical ontology systems. Consequently, the systematic implementation of OAE in China would not only optimize the technical infrastructure for medical intervention safety surveillance but also catalyze paradigm-shifting advancements in causality analysis frameworks for medical intervention-adverse event investigations.

Fig. 5
Fig. 5
Full size image

OAE-based analytical framework for anti-infection DAE case reports in China.

Conclusion

In recent years, the terminology system of OAE has undergone substantial updates and refinement through the collaborative efforts of our research team combined with expert contributions from relevant domains. We will broaden our efforts to extend the OAE terminology into multiple specialized subontologies, particularly those addressing drug- and vaccine-related adverse events. Concurrently, we will strengthen interoperability frameworks with established ontologies, with the goal of streamlining the creation of novel domain-specific knowledge systems.