Deep Learning Prediction of Adverse Drug Reactions in Drug Discovery Using Open TG–GATEs and FAERS Databases

With the advancements in Artificial intelligence (AI) and the build up of health-related big data, it has become increasingly feasible and commonplace to leverage machine learning techniques to analyze clinical and omics data to assess the likelihood of adverse drug reactions or events (ADRs) in the course of drug discovery. Here, we have presented a novel machine learning-based framework for predicting the likelihood of ADRs; it combines two distinct datasets, drug-induced gene expression profiles from Open TG–GATEs (Toxicogenomics Project–Genomics Assisted Toxicity Evaluation Systems) and ADR occurrence information from FAERS (FDA [Food and Drug Administration] Adverse Events Reporting System) database, and can be applied to many different ADRs. Using this framework with Deep Neural Networks (DNN), we built a total of 14 predictive models; in the validation tests, our models achieved a mean accuracy of 89.94%, indicating that our approach successfully and consistently predicted ADRs for a wide range of drugs. As examples, we have described the models in the context of Duodenal ulcer and Hepatitis fulminant, highlighting mechanistic insights into those ADRs. Our models should help to assess the likelihood of ADRs in testing novel pharmaceutical compounds, and will be useful for researchers in drug discovery.