Abstract:

Computational analysis of clinical events is a promising strategy for developing a comprehensive understanding of highly variable disease presentations. The development and validation of new methods appropriate for this general task are increasingly limited by the availability of carefully annotated, open, and diverse datasets of biomedical text. I will discuss our group’s recent efforts to enforce consistent structures and standards on the data within text documents written in the biomedical language. The standards support consistent data models and structures (i.e., knowledge graphs) for unifying heterogeneous observations and relationships as well as machine learning approaches for isolating biologically and clinically-relevant insights. I will also introduce our newly produced text datasets, each of which is richly annotated and freely available.

About Speaker:

J. Harry Caufield is a postdoctoral fellow in the NIH HeartBD2K Center of Excellence at UCLA, where he works with Prof. Peipei Ping of UCLA’s departments of Physiology, Medicine, and Bioinformatics. Before joining UCLA, Dr. Caufield earned his PhD in Integrative Life Sciences at Virginia Commonwealth University, where he studied microbial protein interactions and developed intuitive methods for working with large protein interaction networks. He continues to have an active interest in learning about biological relationships hidden within disparate data sources, particularly those with a direct impact on human health and disease.