Discovering New Drug-Drug Interactions By Text-Mining the Biomedical Literature Beth Percha CS 224N Final Project March 17, 2011
Background ● Americans are on a lot of drugs. ● Drug interactions are bad. ● We don't test for them in clinical trials. ● We need ways to find likely interactions so we can watch out for them.
Methods (I) Start with indexed version of Medline: Two lexicons: 17 million abstracts Genes (43 VIP genes) AND 88 million sentences Drugs (602; at least one interaction)
Methods (II)
Results (I)
Results (II) Least likely interactions... Most likely interactions...
Recommend
More recommend