Large-scale automatic extraction of side effects associated with targeted anticancer drugs from full-text oncological articles

摘要: Display Omitted Innovative targeted anticancer drugs are often associated with unexpected toxicities.There exists no comprehensive toxicity knowledge base for drugs.Systematic studies of drug-associated toxicities can facilitate drug discovery and prediction.We developed an integrated approach to extract drug-SE pairs from full-text oncological articles. Targeted such as imatinib, trastuzumab erlotinib dramatically improved treatment outcomes in cancer patients, however, these innovative agents side effects. The pathophysiological mechanisms underlying effects not well understood. availability a has the potential illuminate complex pathways induced by drugs. While effect association multiple heterogeneous data sources, published articles represent important source pivotal, investigational, even failed trials variety patient populations. In this study, we present automatic process (drug-SE pairs) large number high profile articles.We downloaded 13,855 Journal Oncology (JCO) between 1983 2013. We text classification, relationship extraction, signaling filtering, signal prioritization algorithms extracted total 26,264 average precision 0.405, recall 0.899, F1 score 0.465. show that JCO is largely complementary US Food Drug Administration (FDA) labels. Through integrative correlation analysis, positively correlate their gene targets disease indications. conclusion, unique database built high-profile could development computational models understand toxic