摘要
目的通过神经网络模型研究青霉素类药物的定量构动关系(QSPR),为药物设计和临床应用提供参考。方法利用AM1量子化学算法计算了19种青霉素类药物的11个结构参数,并从文献中获得其3种药代动力学参数:半衰期(t1/2)、表观分布容积(Vap)和血浆蛋白结合率(BRPP)。随机选取其中16种药物的结构参数作为人工神经网络的输入参数,药代动力学参数作为输出参数对QSPR模型进行训练,并用剩余的3种药物对模型进行验证。结果构建的QSPR模型预测值与实验值的差别较小,平均误差分别为<0.015、<0.02和<0.025。结论基于神经网络模型能有效预测青霉素类抗生素的上述药动学参数,可用于青霉素类药物的结构设计,为寻找发现更好的药物提供参考。
Objective To study the quantitative structure- pharmacokinetic relationship(QSPR)of penicillin antibiotics by neural network model,and provide reference for drug design and clinical application. Methods Eleven structural parameters of 19 penicillins were calculated by AM1 quantum chemistry algorithm,and three pharmacokinetic parameters(t1/2、Vap、BRPP)were ob. tained from the literature. The structure parameters of 16 drugs were randomly selected as input parameters of artificial neural net. work,and the pharmacokinetic parameters were used as output parameters to train the QSPR model. The remaining three drugs were used to validate the model. Results The mean errors between the predicted and experimental values were less than 0.015,0.02, 0.025,respectively. Conclusion The above pharmacokinetic parameters of penicillin antibiotics can be predicted effectively based on the neural network model. It can be used in the structural design of penicillins to provide a reference for finding better drugs.
作者
潘祎
黄小凤
黄忠朝
刘艺平
屈健
赵程程
PAN Yi;HUANG Xiao-feng;HUANG Zhong-chao;LIU Yi-ping;QU Jian;ZHAO Cheng-cheng(Department of Radiology,the Second Xiangya Hospital,Central South University,Changsha 410011,China;Hubei niversity of Medicine,Shiyan 442000,China;Department of Biomedical Engineering,School ofBaisic Medical Science,Central South University,Changsha 410000,China;Department of Pharmacy,the Second Xiangya Hospital,Central South University,Changsha 410011,China)
出处
《国际药学研究杂志》
CAS
北大核心
2019年第4期283-286,共4页
Journal of International Pharmaceutical Research
关键词
青霉素类药物
神经网络
反向传播法
定量构动关系
penicillin antibiotics
neutral network
back propagation
quantitative structure-pharmacokinetic relationship
