Inducing Novel Gene-Drug Interactions from the Biomedical Literature

Motivation: Knowledge about the interactions between genes and drugs is important in determining the efficacy and toxicity of medications. We present a supervised learning algorithm for inducing previously unknown gene- drug interactions by text-mining the biomedical literature. This algorithm takes as its input a set of known gene- drug relationships and a literature source. New gene- drug interactions are induced based on similarity to known gene-drug interactions, where similarity is determined according to the biomedical literature. Results: Based on limited training data (258 gene- drug pairs), the algorithm induces correct unseen gene- drug relationships at precisions of over 60%, including gene-drug relationships that are not obvious from name similarity.