As reported here, Kairntech has been working together with Fraunhofer SCAI to generate Knowledge Graphs by automatic relation extraction from scientific content. The output is this project was BEL, a graph language, specifically designed for the expression of biomedical facts.
Another popular format to express, store and exchange Knowledge Graphs evidently is Neo4J’s Cypher: Kairntech has therefore proceeded to publish a sample of the project data from the work with SCAI as a public Neo4J demo in their catalog of public demos (scroll down to “Biomedical Knowledge Graph”).
One finding that the team at Fraunhofer SCAI had noted when analysing the results of the Kairntech analysis was that a number of proteins whose association to other neurodegenerative diseases such as Parkinson’s and Alzheimer’s appear in the newly created knowledge graph as also correlated with Schizophrenia – one of the two indications studied in the projects between SCAI and Kairntech. More on the project has been published here.
We’d like to highlight here that a key aspect of the demo scenario here is that the respective relations don’t come from the processing of already structure, tabular data, but from the creation of relation distilled from the analysis of free text (scientific publications in this case.)
We’d like to thank Alexander Jarasch and Niels de Jong from Neo4J for their support and their interest in setting up this public demo.