Introduction
In the modern research ecosystem, the ability to discover both human expertise and physical resources is critical for scientific collaboration. However, information regarding "who" a researcher is and "what" resources they utilize or manage often resides in disparate systems. This paper, originally presented at the 2011 International Conference on Biomedical Ontology (ICBO), details the technical and conceptual integration of two prominent academic networking platforms: VIVO, a researcher profiling system, and eagle-i, a research resource discovery platform. By aligning their underlying ontologies, the authors demonstrate a path toward a unified map of institutional capacity.
The Challenge of Information Silos
Academic data is frequently compartmentalized. Traditional profiling systems like VIVO excel at tracking "visible" outputs such as publications, grants, and teaching roles. Conversely, the "hidden" infrastructure of science—including specialized reagents, core facilities, and high-end instrumentation—is often managed in separate databases or remains entirely undocumented in digital profiles.
This artificial separation results in two primary inefficiencies:
- Discovery Gaps: Researchers often lack the tools to identify local resources, such as a specific microscope or a rare antibody, and the individuals who manage them.
- Incomplete Expertise Profiles: The labor involved in maintaining a core facility or a unique biospecimen collection is a significant indicator of scientific expertise. Without integrating these resources into researcher profiles, institutional records fail to capture a comprehensive view of a scientist's contributions.
Methodological Approach: Ontology-Driven Interoperability
The integration of VIVO and eagle-i relies on Ontology-Driven Application design. Both platforms utilize the Resource Description Framework (RDF) and Web Ontology Language (OWL) to structure data, but their specific models were developed independently. To achieve interoperability, the authors employed several key strategies:
Adherence to OBO Foundry Principles
The teams followed the principle of "orthogonality," ensuring that their respective ontologies did not overlap unnecessarily. This allowed the systems to be combined without creating redundant or conflicting definitions.
Shared Upper Ontology
Both projects aligned their models under the Basic Formal Ontology (BFO). By using a consistent top-level structure, the systems gained a common semantic foundation for defining classes and properties, facilitating smoother data exchange.
The MIREOT Technique
To manage complexity, the authors utilized the Minimum Information to Reference an External Ontology Term (MIREOT) approach. Rather than importing entire external ontologies, which can introduce unnecessary overhead and logical conflicts, they pulled in only the specific terms required for the integration.
Local Extensions
The implementation allowed for local flexibility. For instance, Cornell University implemented a local extension process that integrated eagle-i resource types into their VIVO instance. This enabled site-specific customization while maintaining the ability to participate in broader, multi-site federated searches.
Structural and Semantic Insights
The alignment process revealed fundamental differences in how the two systems represent data. The authors observed that eagle-i requires a high density of classes to categorize thousands of distinct resource types (e.g., specific cell lines or instruments). In contrast, VIVO requires a high density of properties to describe the intricate web of relationships between people, organizations, and their outputs.
A notable semantic challenge involved the "Homo Sapiens" problem. While VIVO might use foaf:Person to represent a researcher as a social and professional entity, eagle-i might use eagle-i:HomoSapiens to describe a human as a source of a biological specimen. Reconciling these different conceptualizations of the "human" was a central task in the alignment effort.
Results and Impact
The project successfully demonstrated the inclusion of eagle-i resource types within the VIVO instance at Cornell University. This integration allows users to navigate seamlessly from a researcher’s profile to the facilities they manage, effectively treating resource provision as "compelling evidence of expertise" on par with traditional publications.
The systems involved handle significant data volumes, often exceeding tens of millions of statements per university. By bridging these platforms, the authors established a scalable framework for Linked Open Data (LOD) that encompasses people, organizations, grants, publications, laboratories, equipment, and biological specimens.
Future Directions
While the integration was successful, several challenges remain. Contextual ambiguity persists in certain terms; for example, hasLocation may refer to a physical building in a geographical context or a specific anatomical region in a biomedical context. Future work aims to improve metadata consistency to assist automated systems in interpreting these terms. Additionally, the authors seek to bridge the "bench-to-bedside" gap by mapping clinical expertise (using vocabularies like MeSH or UMLS) to the research resources described in the VIVO/eagle-i ecosystem.
Conclusion
The alignment of VIVO and eagle-i represents a significant step toward a more transparent and interconnected research environment. By leveraging shared ontologies and standardized integration techniques, institutions can better represent the full spectrum of scientific activity, from the researchers themselves to the essential infrastructure that makes their work possible.
Summary updated