A Quantitative Structure-activity Relationship（QSAR） model was developed to predict the hallucinogenic activity of phenylalkylamines by Artificial Neural Network（ANN） method.Each compound was represented by the calculated structural descriptors involving constitutional,topological,geometrical,electrostatic and quantum-chemical features of compound.The ANN method produced a nonlinear and seven-descriptor QSAR model with a standard error S = 0.0128 and a correlation coefficient R = 0.9752.The electronic properties of 75 phenylalkylamines were calculated with Gaussian 03 program at the DFT/B3LYP/6-311＋G（d,p） level.The quantum chemical analyses were performed from two aspects of frontier molecular orbital and charge distribution.The results show that seven structural describers are crucial to the hallucinogenic activity of phenylalkylamines and that the para-and ortho-positions could be active sites acting as electron donors.
DFT/B3LYP/6-311G＋（d,p） basis set including solvent effect was first used to calculate a set of molecular descriptors of 55 phenylalkylamine and 20 tryptamine compounds with hallucinogenic activity. Four quantitative structure-activity relationship （QSAR） models of the hallucinogenic activity for phenylalkylamines and tryptamines were obtained by employing multiple linear regression （MLR） method. The QSAR analysis indicated that electron-related descriptors were major contributors to the hallucinogenic activities of phenylalkylamines and tryp- tamines. In addition, electron-unrelated descriptors have some impact on the hallucinogenic activities of phenylal- kylamines. Based on the results of QSAR study, a novel Conformation Complementary Judgement, Transformation and Induction （CCJTI） model had been proposed to explain different action mechanisms of phenylalkylamines and tryptamines with their target receptors. It was concluded that phenylalkylamines might combine with receptor by electronic effect, but steric factor could affect it also, whereas tryptamines could act only through the electronic effect.
采用量子化学密度泛函B3LYP法,用6-311+G(d,p)基组,计算38个苯烷基胺类化合物的电子结构参数;利用多元线性回归(multiple linear regression,MLR)法,筛选出影响化合物迷幻活性显著的6个变量,并建立其结构参数与迷幻活性之间的定量关系(MLR模型);同时,利用人工神经网络(artificial neural network,ANN)法建立相应的QSAR模型(ANN模型)以资对比。所建MLR模型的相关系数R=0.9340,标准误差Se=0.2068;ANN模型的相关系数R=0.9992,标准误差Se=0.0036。结果表明人工神经网络法获得了比多元线性回归方法更精密的拟合效果,可望在QSAR研究中发挥重要作用。
