Science moves fast, and it can be difficult to keep pace. Reading, understanding, and being able to quickly draw on scientific advances in one's own work is a crucial, but extraordinarily time-intensive, part of scholarly work.
The main goal of this project is to reconstruct causal models from text (e.g., scholarly papers) within the biological sciences. The researchers leverage a database called Semantic Medline that contains millions of "predications" (essentially correlative links between ideas) from abstracts of biomedical literature available on PubMed.
The specific goals of this research are:
- To construct graphical causal models from this text data in specific biology focus areas (e.g., specific human diseases)
- To validate these models using high-throughput measurements (e.g., genomics, transcriptomics, metabolomics, proteomics)
- To facilitate the proposal of key experiments that would most substantially reduce the uncertainty in learned models