Wikidata is an open, Semantic Web-compatible database that anyone can edit. This 'data commons' provides structured data for Wikipedia articles and other applications. Every article on Wikipedia has a hyperlink to an editable item in this database. This unique connection to the world's largest community of volunteer knowledge editors could help make Wikidata a key hub within the greater Semantic Web. The life sciences, as ever, faces crucial challenges in disseminating and integrating knowledge. Our group is addressing these issues by populating Wikidata with the seeds of a foundational semantic network linking genes, drugs and diseases. Using this content, we are enhancing Wikipedia articles to both increase their quality and recruit human editors to expand and improve the underlying data. We encourage the community to join us as we collaboratively create what can become the most used and most central semantic data resource for the life sciences and beyond.
In the Stone Soup folktale [
Wikipedia is among the most visited sites on the Internet. Articles about medical
topics were viewed more than 4.88 billion times in 2013, a number on par with
http://nih.gov and significantly greater than WebMD [
Infoboxes provide the bridge between machine-readable structured data and
the unstructured text that forms the main body of each article. Since 2008, the Gene
Wiki project has automatically created and maintained the infoboxes for around
10000 articles about human genes [
The first application to use Wikidata extensively is Wikipedia but this could
be the tip of the iceberg. To give a preview of what Wikidata could become, it’s
useful to briefly examine its closest ancestor, DBpedia. The DBpedia project mines
content from Wikipedia by parsing infoboxes, maps this content to their own ontology,
and provides access to this data in the form of a large RDF database available both for
bulk download and SPARQL query. While enabling interesting queries on its own,
its most important function is as a global linking hub for the Semantic Web [
Within the biomedical domain, useful queries are already possible as a result
of the ‘single-pot’ nature of Wikidata. For example, it is possible to use Wikidata’s
SPARQL endpoint (https://query.wikidata.org/) to answer questions such as “what
clinically relevant drug-drug interactions are known for the drug methadone
(CHEMBL651)” [
This kind of serendipitous, automatic, cross-continental data integration is the primary goal of the Semantic Web, but is not yet commonplace. The key beauty and main challenge of the Semantic Web is its distributed nature. In order for this kind of integration to happen in the absence of a centralized resource like Wikidata, several major hurdles would need to be leaped. First, both teams would need to know enough about the fairly complex stack of semantic technologies to provide their data as RDF through a stable, public SPARQL endpoint. Second, they would have to work with overlapping identifier systems. Third, the would-be consumer of their data would need to discover both of their endpoints and be sophisticated enough with SPARQL to identify and issue the appropriate distributed query. All of this is possible and can work, but it is not easy.
By integrating data in a centralized, single community pot, Wikidata
provides a platform that addresses each of these problems. Data providers do not have to
set up and maintain their own SPARQL endpoint – a challenge that very few teams
have succeeded at doing for any length of time [
The fact that Wikidata is one centralized, community resource immediately surfaces the challenges incurred in any collaborative ontology development process. In Wikidata, the ‘ontology’ corresponds to its collection of linking properties used to describe items. A new property in Wikidata has to be proposed for community discussion and is only created after a consensus regarding the value of the property and its relation to existing properties has been established. For those used to controlling their own data and data models, this process can feel tedious. But this same fundamental process must be undertaken in any attempt at data integration. The fact that it happens up front, when data is first being loaded, should help to keep the data consistent and reduce the downstream identifier and ontological mapping problems that continue to plague bioinformatics.
Imagine the power of combining the structured data in Wikidata, the high accessibility and dedicated community of Wikipedia and the knowledge of the scientific community. Contemplate further that all of this data is freely available and accessible through a stable query interface and robust, read/write API. This makes important, high-quality information easily accessible by anyone and opens up scientific knowledge for public scrutiny. Further, the built-in provenance tracking can provide detailed chains of evidence to support or refute each claim and all of this can be discussed using the many social tools, such as ‘talk pages’ for every data item, baked into the MediaWiki infrastructure.
Aside from creating useful ways to disseminate data, this sociotechnical structure provides a framework for the broad community to broadcast feedback back to the original data owners. Even at this early stage of this project, this process has already led to improvements in source data. For example, in the Disease Ontology the term ‘Ollier disease’ had the synonym ‘Maffucci syndrome’. Upon importing the Disease Ontology into Wikidata, members of the Wikidata community pointed out that the two terms, though putative synonyms, linked to two different extant Wikidata items. Upon closer review it was determined that these two terms represent two different, albeit closely related, diseases, leading to the creation of a new term in the Disease Ontology. As Wikidata expands it is to be expected that additional differences in representation between it and other knowledge resources will surface. These will first be triaged by the Wikidata community to check for errors and, if consensus is achieved that there is an error in the original source, this will be relayed for consideration. In this way, the Wikidata community can become the ‘many eyes’ that make all ontology bugs shallow. ...Can Make a Delicious Soup
We can create a powerful commons of biomedical knowledge by building on established resources and the dedicated community to connect genes, proteins, drugs, diseases, phenotypes and symptoms. Wikipedia will be the first application to use the content in Wikidata, but certainly not the last. The fire is ready and the pot is starting to heat up. Some villagers are already peeking out of their windows ready to join us around the pot, but it will take the effort of the whole community to make a delicious biomedical data soup. We invite you to join us in this effort.