Modeling Reactivity to Soft, Hard, and Biological Targets with a Deep Learning Network

Unexpected drug toxicity is a critical problem for the pharmaceutical industry. Toxicity problems cause around 40% of drug candidates to be discontinued, oftentimes only after significant resources have been invested. Furthermore, drug-induced liver injury (DILI) is the most common reason already approved drugs are withdrawn from the marker, and causes half of all cases of acute liver failure, as well as 15% of all liver transplants within the United States. Frequently, toxicity is caused by electrophilic drugs (and drug metabolites) that covalently bind to nucleophilic sites within biological macromolecules, including DNA and proteins. Conjugation to DNA can cause cancer, and conjugation to protein can cause a toxic immune response. For example, the well known hepatotoxicity of an acetaminophen overdose is due to metabolism of acetaminophen by Cytochromes P450 into the electrophilic metabolite N acetyl-p-benzoquinone imine (NAPQI). NAPQI is electrophilically reactive and covalently binds to nucleophilic sites within proteins, resulting in hepatotoxicity in high doses (Figure 1).