Computer model predicts side effects

Abstract

Drug candidates are usually found to be unsafe only late in the drug discovery process. A method for predicting the many biological targets of a given molecule might allow drug safety to be considered much earlier. See Article p.361 Adverse drug reactions can limit the clinical use of otherwise effective drugs, and are the leading cause of failure for new drugs in clinical trials. Drugs frequently interact with more than one protein target, and hundreds of such proteins are linked to the side effects of clinically used therapeutics. This large-scale computational study predicted the activity of 656 drugs against 73 protein targets that have been associated with adverse drug reactions. Of the 1,042 predictions tested, approximately half were confirmed. One prediction was that abdominal pain, a known side effect of the synthetic oestrogen chlorotrianisene, was mediated through its inhibition of COX-1, and the clinical relevance of this inhibition was borne out in whole human blood platelet-aggregation assays. The researcher's approach could be used to weed out drugs most likely to fail before they go to clinical trial.