Building on prior developments in semantic and graph technologies, Knowledge Graphs (KGs) are increasingly being used to map together diverse, heterogeneous datasets in multiple domains. Algorithms based on KGs can be used to make predictions, and to extract patterns of insight. In this presentation, we will describe three kinds of knowledge graph that can be built (universal, domain and problem-specific), and will give examples of how link prediction and node embedding algorithms on top of domain and problem-specific graphs can be used as an alternate method of predicting drug-target activity (including off-target prediction) as well as identifying complex relationships between chemistry and biology that can be explored visually, and used to build hypotheses for biological effects of compounds, including off-target effects, adverse events, and phenotypic deconvolution.

Using knowledge graphs for prediction and visual hypothesis generation in drug discovery

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.