【摘要】 目的应用量子化学和神经网络方法,研究喹诺酮类化合物的的定量构动关系。方法计算了21个喹诺酮类化合物的结构参数,建立了含两个隐含层的反传神经网络定量构动关系模型。输出层为试验获得的喹诺酮类化合物的3个药动学参数Cm ax、AUC和t1/2;输入层为筛选出的喹诺酮类化合物的6个结构参数。结果随机挑选20个化合物作为训练集并进行leave-one-out分析。剩下1个化合物作为预测集,预测结果能够很好地与试验数据相吻合。结论所建立的定量构动关系模型能够有效地分析喹诺酮类化合物的定量构动关系。更多还原

【Abstract】 OBJECTIVE To use quantum chemistry and neural network to study the quantative structure-pharmacokinetic relationship（QSPR） on quinolones compounds.METHODS 16 molecular structure parameters of 21 quinolones drugs have been calculated and a QSPR model of neural networks has been built.6 molecular structure parameters were chosen as the inputs of the networks and 3 pharmacokinetics parameters of quinolones(C_max,AUC and t_1/2) as the output.RESULTS 20 compounds were selected stochastically as the training sets and tested by leave-one-out method.The residual compound was selected as the predicted sets.Pharmacokinetics parameters have been predicted by the QSPR models.While the predicted values of the residual compound showed good agreemenu with the experimental ones.CONCLUSION The QSPR models can be use to predict the pharmacokinetics parameters of quinolones compounds.更多还原