Knowledge Graphs for Biomedical Knowledge Discovery

Knowledge graphs are heterogeneous graphs that provide a powerful, expressive, and machine comprehensible way to model interconnected multi-relational heterogeneous entities to describe complex biological process. Due to the proliferation in multi-omics databases, knowledge graphs are becoming main stream in biomedical research by designing AI approaches to infer important insights and drive new discoveries . Knowledge graph embedding is the popular way of predicting hidden links from incomplete knowledge graphs by projecting entities and relations into low-rank embedding space. These embeddings are then used in many downstream tasks, for example, link prediction, entity resolution and node classification. Knowledge graphs are increasingly being used in biomedical knowledge discoveries such as protein structure, function and interaction prediction, drug repurposing and drug discovery. In this tutorial, I will focus on, 1) listing biomedical data sources, 2) constructing knowledge graphs using multi-omics data, 3) solving a case study on drug repurposing using knowledge graphs.