A large part of the daily practice of a researcher doing in vitro Drug Discovery is comparing and manually matching high-quality information from multiple disciplines in the Life and Biomedical Sciences. The Open PHACTS Discovery Platform4 is an initiative to integrate publicly available data relevant for both academia and the pharmaceutical industry. It integrates numerous datasets including for example ChEBI, ChemSpider, DrugBank and the GeneOntology. The platform provides an easy interface that allows researchers to consult the database without being confronted with the complexity of de ning e cient Linked Data queries. A set of services are accessible via a RESTful interface. The Open PHACTS Discovery Platform provides an interpretation of biomedical research activities (identi ed by domain experts) as work ows that are authored using visual tools. Work ows retrieve data via API calls. The platform executes the resulting instantiated queries at an endpoint that serves relevant data.Currently, the infrastructure uses commercial software to reason over the vast amount of RDF data and the Big Data Europe (BDE) project took up the challenge to get the same functionality but via open source Big Data technology.
The BDE project5 is developing a re-usable Big Data infrastructure (BDI) needed by data-intensive science practitioners tackling a wide range of societal challenges. The infrastructure is designed to cover aspects of publishing and consuming semantically interoperable, large-scale, multi-lingual data assets and knowledge. This BDE infrastructure is designed to minimize the disruption of current work ows, and maximizes the opportunities by taking advantage of the latest European RTD developments, including multilingual data harvesting, data analytics, and data visualization. To test the e ectiveness of the platform, 4 http://www.openphactsfoundation.org 5 https://www.big-data-europe.eu multiple pilot implementations are developed in the various domains. The rst of these pilots is the Drug Discovery Pilot implementation, which replicates much of the functionalities of the Open PHACTS platform. The infrastructure relies heavily on the Docker containers6 and con guration via Docker Compose where generic 3rd party Docker containers (e.g. MemCached, MySQL, SPARK, HDFS) are combined with custom made pilot speci c Docker containers. 3
In the pilot we propose to demonstrate7, the Open PHACTS functionality is implemented on the BDI. One goal of this pilot is to investigate dealing with the signi cant diversity of the entity name space in the bio-medical domain and exploring how this issue a ects a generic big data infrastructure. Mapping this vast amount of entities leads to a signi cant increase of triples. A second goal is covering data and query security and privacy requirements and exploring how the methods used to handle this in the current implementation of the Open PHACTS Discovery Platform can be used to guide development of the generic BDE platform. An important challenge for this pilot is to replace the commercial cluster version RDF store, with an open source variant version: 4Store. To this end, we are implementing a 4Store BDE docker component and improving it in such a way that it can serve as a generic component on the BDE infrastructure8.
The pilot integrates multiple datasets, available in RDF. The mappings between the identi ers used in the various datasets are freely available as RDF linksets9. Most datasets have a metadata description published in VoID. The functionality of the Open PHACTS services is described in SWAGGER. The following processing is carried out: { Real time processing: Using an external service (such as the Scienti c Lenses keyword expansion service) to process a query and then to execute the processed query on the data stored in the infrastructure. { Batch processing: Data transformations that align and link datasets at ingestion time. The datasets above are regularly updated and must be periodically re-ingested.
The pilot implementation exposes a querying endpoint as well as a data ingestion endpoint for visualization or further processing.
The pilot itself is available in its entirety as Open Source software10. Both BDI and the pilot-speci c components are implemented as Docker components. Acknowledgements. This work is supported by European Union's Horizon 2020 research and innovation programme under grant agreement No 644564 www.bigdata-europe.eu. We thank our BDE collaborators for their support. 6 https://www.docker.com/what-docker 7 https://github.com/big-data-europe/pilot-sc1-cycle1 8 https://github.com/big-data-europe/docker-4store 9 https://www.openphacts.org/2/sci/data.html 10 Download and instructions at https://github.com/big-data-europe/pilot-sc1-cycle1