摘要
采用量子化学的算法,计算了18个黄酮类化合物的13个结构参数,并采用逐步回归分析,筛选出3个电性结构参数,即A环的总电荷、B环的总电荷、双键氧上的电荷。使用Matlab7.0软件中的神经网络工具箱建立黄酮类化合物抗氧化活性的神经网络模型。交互验证法预报结果表明,预报误差平方和(PRESS)为0.0192,交叉验证系数(R)2为0.875,该神经网络模型预报结果可靠,神经网络可作为化合物活性研究的有效计算机辅助手段。
13 structural parameters of 18 flavonoids were calculated with the method of quantum chemistry.In addition,these structural parameters were selected by stepwise regression.Three electrical parameters were screened out.That is the total charge of the A ring and B ring and the charge of oxygen which had double bond.The neural network model of antioxidant activity in 18 flavonoids was set up successfully by using Neural network toolbox in matlab 7.0 software.The result of the leave-one-out-cross-validation sho...
出处
《化学分析计量》
CAS
2009年第5期51-54,共4页
Chemical Analysis And Meterage
关键词
黄酮
抗氧化活性
构效关系
AM1算法
神经网络
flavonoids
antioxidant activity
structure-activity relationship
AM1
neural network
