物理化学学报 >> 2002, Vol. 18 >> Issue (05): 385-388.doi: 10.3866/PKU.WHXB20020501

通讯    下一篇

应用反向传播神经网络预测化合物脑血分配系数

乔学斌;侯廷军;章威;徐筱杰   

  1. 北京大学化学与分子工程学院,北京　100871
  • 收稿日期:2001-12-26 修回日期:2002-02-28 发布日期:2002-05-15
  • 通讯作者: 徐筱杰 E-mail:xiaojxu@chem.pku.edu.cn

Predication of Brain-blood Partitioning Using Back Propagation Network

Qiao Xue-Bin;Hou Ting-Jun;Zhang Wei;Xu Xiao-Jie   

  1. College of Chemistry and Molecular Engineering, Peking University, Beijing　100871
  • Received:2001-12-26 Revised:2002-02-28 Published:2002-05-15
  • Contact: Xu Xiao-Jie E-mail:xiaojxu@chem.pku.edu.cn

PDF

3480

赞

摘要： 选用27种三维结构性质描述符对脑血分配系数预测建立神经网络模型.网络模型选用典型的适合函数逼近的两层结构神经网络对脑血分配系数（lgBB，BB为脑血浓度比）进行预测,计算中采用的模型具有一个双曲正切型激活函数的隐含层和一个线性激活函数的输出层.计算表明,使用小心选择的反向传播神经网络模型对化合物脑血分配系数具有较好的预测能力.

关键词: 脑血屏障, QSPR（定量结构性质分析）, 分配系数, 反向传播, 神经网络

Abstract: The back-propagation network based predication of brain-blood concentration ratio (BB) using 27 structurally derived descriptors was investigated. A typical biased two-layer network,including a sigmoid layer and a linear output layer, suited for arbitrary function approximation was chosen for the predication. The results show that fine-tuned back-propagation network models are very efficient for predication use.

Key words: Brainblood barrier, QSPR (Quantitative Structure Property Analysis)