Toward the accurate prediction of P450-mediated metabolism and adverse drug reactions using novel MM methods and QM-derived rules.

尽管对候选药物的毒理学和临床试验进行了大量投资，但药物不良反应仍然是停药的主要原因。这种形式的药物毒性可能只影响一小部分患者，并且在临床试验中未检测到。然而，如果这种药物已经上市并广泛使用，这种毒性可能是致命的。多年来，药物化学家一直依赖于“结构警报”来识别药物的潜在毒性，更具体地说是特异质毒性。这些官能团通常被代谢酶（例如，P450）转化为反应性代谢物，导致观察到的毒性。然而，这种官能团的存在并不总是与毒性相关，因为生物活化 仅在药物被处理时才发生（即，识别），并且尽管广泛使用，但这种方法并不可靠（C&EN 2012，90，34）。因此，计算预测已经成为研究的途径，在这种情况下，我们开发了一个程序，IMP 10（阿斯利康/NSERC CRD资助01/2010-02/2012），它预测药物代谢的位点，从而预测药物和结构警报是否可以被主要的P450处理。接下来，我们决定从两个方面进一步改进这个项目。一方面，我们认为使用量子力学技术来改善配体反应性规则，更具体地说，一旦被P450生物活化，它们倾向于导致反应性代谢物，重点是苯衍生物。其次，我们研究了一个概念新颖的力场的发展，以更准确地计算药物势能。在一项概念验证研究（化学计算组/NSERC CRD 08/2013-08/2014）中，我们已经证明了这两种优化确实是可能的。在当前的提案中，我们建议在这两个方面进一步开展工作，并完成预测反应性代谢物（即药物不良反应）的程序开发。

Despite the large investment in toxicology and clinical trials of drug candidates, adverse drug reactions remain a leading cause of drug withdrawal. This form of toxicity of drugs may only affect a small fraction of the patients and is not detected in clinical trials. However, if the drug is already on the market and widely used this toxicity may be fatal. Over the years, medicinal chemists have relied on "structural alerts" to identify potential toxicity of drugs and more specifically idiosyncratic toxicity. These functional groups are often activated by metabolic enzymes (e.g., P450) into reactive metabolites leading to the observed toxicity. However the presence of such a functional group does not always correlate with toxicity as the bioactivation can only occur if the drug is processed (i.e., recognized) by these enzymes and although widely used this approach is not reliable (C&EN 2012, 90, 34). Thus, computational prediction has become an avenue of research and in this context we have developed a program, IMPACTS (AstraZeneca/NSERC CRD funding 01/2010-02/2012) which predicts the site of metabolism of drugs, hence whether the drug and structural alert can be processed by the major P450s. We next decided to further improve this program in two ways. On one side, we thought to use quantum mechanical techniques to improve the ligand reactivity rules and more specifically their inclination to lead to reactive metabolites once bioactivated by P450s focusing on benzene derivatives. Second we investigated the development of a conceptually novel force field to compute drug potential energies more accurately. In a proof of concept study (Chemical Computing Group/NSERC CRD 08/2013-08/2014), we have shown that these two optimizations can indeed be possible. In the current proposal, we propose to further work on these two front and complete the development of a program predicting reactive metabolites hence, adverse drug reactions.