药物清除率的定量构动关系研究

目的　利用量子化学和神经网络方法,建立药物分子结构参数与药物清除率之间的关联模型,以预测药物的清除率。方法　计算了已知清除率的100种药物的18个结构参数,并对其进行了主成分分析,获得3个独立结构参数,分别以3个结构参数和18个结构参数作为神经网络的输入,进行了神经网络的建模和预测。结果　采用3个结构参数作为输入参数的神经网络的预测能力明显优于采用18个结构参数的预测能力。结论　所建预测药物清除率的神经网络模型可行,建模时进行药物分子结构参数选择非常必要。

Study on quantitative structure-pharmacokinetics relationship between drug clearance and their structure parameters

OBJECTIVE To establish a model of correlative relationship between the total drug clearance and the drug structure parameters by using quantum chemistry and neural network, and predict the total drug clearance based on the model. METHODS Eighteen structural parameters of one hundred drugs, whose clearances had been determined, were calculated and three independent structural parameters were selected by using main component analysis. Two neural network models including two hidden layers were set up by inputting 3 or 18 independent structural parameters,respectively. RESULTS The learning and predicting results of two neural network models showed that three independent structural parameters neural network models were better than that of eighteen structural parameters one. CONCLUSION All results show that it is possible to predict total drug clearance basing on neural network model and it is necessary to select drug structural parameters by using main component analysis.