关键词：
Computational chemistry
Artificial intelligence
Computer science
摘要：
Cytochrome P450 enzymes aid in the elimination of a preponderance of small molecule drugs, but can generate reactive metabolites that may adversely conjugate to protein and DNA, in a process known as bioactivation, and prompt adverse reaction, drug candidate attrition, or market withdrawal. Experimental assays are low-throughput and expensive to perform, so they are often reserved until later stages of the drug development pipeline when the drug candidate pools are already significantly narrowed. Reactive metabolites also elude in vivo detection, as they are transitory and generally do not circulate. In contrast, computational methods are high-throughput and cheap to screen millions of potentially toxic molecules during early stages of the drug development pipeline. This work computationally models sequences of metabolic transformations, i.e., pathways, between an input molecule and a corresponding, optional reactive metabolite(s). Additionally, an accurate graph neural network model was developed to assess importance of intermediate metabolites and extract connected subnetworks of relevance to bioactivation. Connecting multiple site of metabolism and structure inference models, we developed an integrated model of metabolism and reactivity to evaluate bioactivation risk driven by epoxidation, quinone formation, thiophene sulfur-oxidation, and nitroaromatic reduction. We applied this framework to an understudied substructure, the isoxazole ring, that is gaining traction in a class of drugs known as bromodomain inhibitors that may potentially drive quinone formation. Finally, we attend to toxicity associated with drug-drug interactions, particularly with NSAID usage reported in electronic health records.
欢迎来到三峡大学图书馆！
