Translating Diabetes Research through Automated Data Extraction and Knowledge Graph Induction from Published Literature

Our broad scope focuses on cardiovascular disease (CVD), the leading cause of death worldwide, and prediabetes, a modifiable risk factor for CVD which affects approximately one-third of Americans. However, for each of these complex diseases, hundreds of original research articles are published each year. In addition, reported estimates of risk for CVD among people with prediabetes use 5 commonly-used definitions of prediabetes. Current approaches to synthesis of risk cannot harmonize across these competing definitions, reducing sample sizes from hundreds to dozens. We aim to address both of these challenges, by developing NLP-based tools to extract numerical results from published research and to develop methods which  use all of the available data, correctly accounting for differences in definitions. We will use these tools to provide the NIA (and US taxpayers) with interpretable summaries of the research related to CVD among people with prediabetes as a function of age.

Zoom information will be shared with e-HAIL members.