An Integrated Knowledge Graph for Microbe-Disease Associations.

Following the rapid advances of the human microbiome, the importance of micro-organisms especially bacteria is gradually recognized. The interactions among bacteria and their host are particulary important for understanding the mechanism of microbe-relate diseases. This article mainly introduces an explorative study to extract the relations between bacteria and diseases based on biomedical text mining. We have constructed a Microbe-Disease Knowledge Graph (MDKG) through integrating multi-source heterogeneous data from Wikipedia text and other related databases. Specifically, we introduce the word embedding obtained from biomedical literature into traditional method. Results show that the pre-trained relation vectors can better represent the real associations between entities. Therefore, the construction of MDKG can also provide a new way to predict and analyse the associations between microbes and diseases based on text mining.