摘要
结合逐步线性回归方法与人工神经网络方法，研究了含硫芳香衍生物对发光细菌毒性的构效关系。充分表明了两个方法的互补，回归方法为网络方法提供变量的物理解释，网络方法建立精确的构效关系模型。人工神经网络方法在非线性较强的构效关系研究中起到重要的作用。基于交叉检验，本文还提出了防止人工神经网络方法过拟合现象发生的ｒ判据，对于建立较好的预报模型，具有一定的普遍意义。
With the artificial neural network (ANN) method combined with the multiple linear regression (MLR). a quantitative structure-activity relationship (QSAR) was carried out on a set of 29 substituted aromatic sulfur derivatives. The compounds were tested on photobacteria (photobacterium phosphoreum) for the toxicity. It is shown that the two methods are complementary in the calculations. The regression method gives support to the neural network method with the physical explanation, and the neural network method gives out a more accurate model for QSABR. The ANN method can Play an important role in non-linear QSAR. Based on cross-validation, here we suggested a r-criterion to avoid the overfitting phenomena in ANN.
出处
《计算机与应用化学》
CAS
CSCD
1996年第1期20-24,共5页
Computers and Applied Chemistry
基金
国家自然科学基金
关键词
神经网络
构效关系
含硫芳香衍生物
Artificial neural network, Structure-activity relationship, Toxicity to photobacteria, r-criterion
