Abstract

Knowledge Graphs have in recent years gained a lot of prominence within biomedical AI and applications. This partnership holds tremendous potential given the highly complex and sparse nature of biomedical data, along with the need for prior knowledge to be integrated with the world’s knowledge to obtain a deep and comprehensive view of complex disease landscapes such as in Oncology. The ‘unFAIR’ (Findable, Accessible, Interoperable, Reusable) nature of this research field makes healthcare AI technically and scientifically challenging, and Knowledge Graphs driven by NLP and GraphML (Graph Machine Learning) could greatly influence the drug discovery and development processes.

In this talk, I will discuss AstraZeneca’s Knowledge Graph (BIKG) focussing on a selection of real-world applications answering day-to-day drug discovery questions. These use cases cover accelerating target discovery to predicting drug efficacy in disease models. In addition, I will also discuss how Graphs can help bridge the critical bench-to-bedside gap in biomedical R&D by translating Discovery knowledge to Clinical applications, and vice versa. I will conclude by discussing key bottlenecks and pain-points prevalent in the scientific community that will need to be addressed for Knowledge Graphs to drive and influence key decisions in the drug discovery pipeline.