To promote data dissemination and integration of life sciences datasets produced by RIKEN, we developed an infrastructure database called "RIKEN MetaDatabase" that enables data publication and integration using the Resource Description Framework. In addition, we implemented a simple work ow for data management and a graphical user interface representing data links across laboratories. Consequently, inter-laboratory collaboration and coordination have been accelerated. Combined with global standardization activities, we expect that this infrastructure will contribute to worldwide data integration.
RIKEN is a comprehensive science research institute that covers physics, chemistry, and life sciences. Conventionally, RIKEN produces various individual life sciences databases. The integrated analysis of the internal and external research data is essential.
Semantic Web technologies based on the Resource Description Framework
(RDF) are powerful tools to realize distributed global data integration. Recently,
RDF-based public databases, such as Bio2RDF[
The RIKEN MetaDatabase assumes two user types: data viewers and data publishers. The main requirement of the system is to ensure simple and useful operations by both user types.
For data viewers, the requirements are as follows: 1) enabling users to nd datasets via a simple graphical interface (GUI) and 2) showing the interrelations of data entities across databases.
For data publishers, RDF data generation should be handled by biologists who are unfamiliar with Semantic Web technologies. The platform supports a tabular form RDF data, which is frequently used in life sciences, to enable biologists to create RDF data more easily. 3 3.1
Spreadsheet for RDF data generation As described above, the core methodology that allows biologists to generate RDF data is the introduction of a tabular form that can be implemented using a spreadsheet application. Figure 1 shows an example of a Microsoft Excel spreadsheet.
In the spreadsheet, a table, column names, and data cells correspond to a class, properties and instances/literals, respectively, and the spreadsheet corresponds to a database that can be represented as an RDF graph. In a single Microsoft Excel workbook, users can de ne or update a graph (database) with multiple classes (tables), properties, instances, and triples. The RIKEN MetaDatabase is implemented as a web server that connects a backend RDF triple store to a GUI accessed via a web browser. In addition, a SPARQL endpoint acts as an application programming interface (API). We employ two virtual machines(VMs) on our private cloud, i.e., the "RIKEN Cloud Service". One VM provides the GUI, and the other hosts Virtuoso as the backend RDF triple store. With cloud computing, database publishers using the RIKEN MetaDatabase do not require to con gure hardware or use external software. The GUI reproduces the tabular list de ned by the spreadsheet (Fig. 2).
The RIKEN MetaDatabase has a "built-in" database, i.e., the RIKEN Database Directory (http://metadb.riken.jp/metadb/db/DBcatalog), that list the metadata in order to easily nd the RIKEN databases. 4
As of July 2016, there were 110 individual databases, 21 ontologies, and 152,784,620 RDF triples registered in the RIKEN MetaDatabase. These databases as well as the non-RDF-based databases published by RIKEN are also listed in the specialized RIKEN Database Directory. The RIKEN Database Directory is designed to be data-compatible with the IntegBio Database Catalog, which is a portal site for life sciences database integration by four ministries of Japan.
This system realizes data-mediated collaboration among the RIKEN database developers. Figure 3 shows the examples of the common uses of URIs across databases, where multiple databases have multiple inter-reletions with other datasets (databases and ontologies) via common URIs.
Applying native RDF technologies, we have developed the RIKEN MetaDatabase,
a database infrastructure for the publication and integration of data produced
by RIKEN. The data-mediated collaboration shown in Fig. 3 bene ts both the
data developers as well as the data viewers, facilitating the integration of
published data and a viewer's private data. For the further reuse of existing URIs,
development of a retrieval function of public data realized in RightField[
In another case, laboratories have initiated developing common schemata,
including common upper-level classes and properties (e.g., mouse strains and
cell lines as experimental materials or images), to standardize the data format
of commonly used data entities. The standardization of such RDF schemata
should be consistent with global activities such as Open Biomedical Ontology
and BioSharing [