Harnessing the Power of the Community in a Library of Biomedical Ontologies

Biomedical ontologies provide essential domain knowledge to drive data integration, information retrieval, data annotation, natural-language processing, and decision support. The National Center for Biomedical Ontology is developing BioPortal, a web-based library of biomedical ontologies. As the biomedical community develops the ever growing set of ontologies of varying sizes, quality, purpose, level of logical rigor, it becomes more and more difficult to find the “best” and “most appropriate” ontologies in the domain. In BioPortal, we use the social approaches in the Web 2.0 style to bring structure and order to the collection of biomedical ontologies. BioPortal enables users to provide and discuss a wide array of knowledge components, from submitting the ontologies themselves, to commenting on and discussing classes in the ontologies, to reviewing ontologies in the context of their own ontology-based projects, to creating mappings between overlapping ontologies and discussing and critiquing the mappings. In this paper, we discuss the community features of the BioPortal ontology library and describe the infrastructure that supports these features. BioPortal is available online at http://bioportal.bioontology.org. 1 The library of biomedical ontologies in BioPortal As the number of ontologies available for Semantic Web applications grows, so does the number of ontology libraries that index and organize the ontologies. Some libraries get the ontologies by crawling the Web (e.g., Swoogle [4], Watson [3] and OntoSelect [1]). In other libraries, users submit their ontologies themselves (e.g., the Protégé ontology library1). Some libraries provide strict selection criteria for inclusion (e.g., OBO Foundry [14]). All these libraries provide a gateway for users and application developers who need to find ontologies to use in their work. In our laboratory, as part of the National Center for Biomedical Ontology (NCBO), we have developed BioPortal2—an open library of biomedical ontologies. Researchers in biomedical informatics submit their ontologies to BioPortal and community members can access the ontologies in their web browsers through the BioPortal user interface or through web services [12]. The BioPortal users can browse and search the ontologies, update the ontologies that they authored by uploading new versions, comment on any ontology (or portion of an ontology) in the library, evaluate it, describe their experience in using the ontology, or make suggestions to ontology developers. This focus on enabling members of the 1 http://protegewiki.stanford.edu/index.php/Protege_Ontology_ Library 2 http://bioportal.bioontology.org community to contribute actively to BioPortal content and to increase its value to other users, distinguishes BioPortal from other ontology libraries. At the time of this writing, BioPortal has 160 ontologies, covering a wide range of domains in biomedicine, from anatomy, to diseases, to protein descriptions. BioPortal supports several formats for representing ontologies: the OBO format, Rich Release Format (RRF) from the US National Library of Medicine (for terminologies comprising the Unified Medical Language System [8]), OWL, RDF(S), and the Protégé frames format. BioPortal uses the Mayo Clinic’s LexGrid system3 to store ontologies in OBO Format and to access standard biomedical terminologies, such as UMLS in RRF. Protégé4 serves as the backend for OWL and RDF ontologies. This paper makes the following contributions: – We define several types of contributions that a community-based ontology library can collect and aggregate, including ontologies and their successive versions, notes and threaded discussions on classes, reviews of ontologies in the context of specific projects, ontology mappings, and views and subsets of ontologies. – We describe the validation of our approach in the form of a production implementation of BioPortal. 2 Community Features of BioPortal With the open, community-based nature of the BioPortal library of biomedical ontologies, we are experimenting with the new ways of publishing, evaluating, and integrating the knowledge infrastructure that is essential to life sciences today. Specifically, BioPortal users and ontology developers can contribute a variety of information to the library, increasing its value to others. First, the library enables all members of the biomedical community to publish their ontologies (Section 2.1). Second, other members of the user community can provide feedback on specific elements of the ontologies or ontologies as a whole (Section 2.2). Third, ontology users can describe their experiences in using the ontologies from the library in their ontology projects, thus providing a novel way of evaluating ontologies (Section 2.3). Fourth, we open the process of declaring relationships between concepts in different ontologies (ontology mapping) to the community by enabling users to declare mappings between concepts and to comment on mappings created by others (Section 2.4). Finally, users can publish and describe subsets or different views of the ontologies in BioPortal (Section 2.5). 2.1 Publishing Ontologies in BioPortal Any developer of an ontology that is relevant to biomedical domain can publish it in BioPortal. When submitting an ontology to BioPortal, the user must provide essential metadata about the ontology, such as its name and acronym, the domain that the ontology covers, keywords, links to additional information, and provenance information, including who developed the ontology, version details, dates of the release, and so on. 3 http://informatics.mayo.edu/LexGrid 4 http://protege.stanford.edu The ontology authors then have two choices in terms of submitting the ontology itself. They can choose not to submit the ontology content, thus providing BioPortal only with the ontology metadata. BioPortal users will then be able to see the ontology metadata in the library and to comment on the ontology as a whole. They will not be able to view and search the ontology content though. We designed this submission option for ontology authors who are not willing to make their ontology accessible to the community directly through BioPortal (e.g., because of licensing issues) but would still like the community to know about their resource. Most ontology authors, however, choose the second option: submit both the metadata and the ontology itself. After the ontology author submits an ontology, BioPortal parses and indexes it and makes it available for searching and browsing. All ontologies in BioPortal are publicly accessible. Ontology Web Services and Links. When ontology authors submit their ontology to BioPortal, they enable a wider user community to find and use their ontology. Furthermore— and in some cases more important—publishing in BioPortal is a very easy way to get a web presence for an ontology. Many ontology developers are neither interested nor willing to host their own web server. By uploading their ontology to BioPortal, they get a web link (a URL) that they can give to their users when inviting these users to browse the ontology, see details of specific concepts, visualize the ontology or any of its parts. BioPortal also provides URLs to access any class in the ontology directly. Thus, for example, an ontology author can send to her collaborators a URL for a specific class that she wants to discuss with them. The collaborator can then see the details of the class definition and all the related information by following the link. Any BioPortal user can subscribe to an RSS feed of changes to a specific ontology in order to get notified of any user-contributed content relevant to that ontology, such as comments or mappings. In addition to accessing an ontology and its components in a web browser, users can use the BioPortal RESTful API to access any ontology or its components through a web service. In fact, the BioPortal user interface itself uses this REST API to display most of the information that the users see on the BioPortal web site. There are web services to get metadata about an ontology, its root classes, details of any concepts, hierarchical information for any concept; there are web services to download an ontology, get a diff between two versions, get notes or mappings for an ontology [12].5 Ontology Versioning. Most, if not all, ontologies in BioPortal continue to evolve and authors continue to publish new versions. Thus, any ontology library must address the issue of ontology evolution [9]. Ontology authors can submit successive versions of their ontologies to BioPortal. Each version can have its own set of metadata, since any detail about an ontology (from its scope, to provenance details, to relevant links) can change from one version to another. Users can explore and use any version of any ontology in the collection (Figure 1). For each ontology, BioPortal provides two sets of services and links: one set resolves to a specific version requested by the user; another set resolves to the latest version of an ontology. Thus, if a user’s application relies on a 5 See http://bioontology.org/wiki/index.php/NCBO_REST_services for the details of REST services that NCBO currently offers. Fig. 1. Ontology details in BioPortal. The details page describes the information about the ontology, its provenance, lists the versions of the ontology and provides an overview of concepts that have notes and mappings. specific version, they can pass that version as the parameter to the services. If the user wants to get the latest version, whatever that version is, they can pass in the generic ontology id (we call it a “virtual” id); BioPortal redirects this call to use the latest version of the ontology. BioPortal and OBO Foundry The OBO Foundry initiative [14] aims at creating a set of well-documented and well-defined ontologies that are designed to work with one another. The OBO Foundry has an editorial process defining which ontologies beco