摘要
为定量结构/活性相关性,研究提取了量子化学参数、拓扑指数Am及分子连接性指数mX1,同时应用正交变换和最佳变量子集算法(Leaps-and-Bonds)进行了变量压缩和选择,进而实施了多元回归分析,并由此结果进行了2-甲酸吡啶缩氨基硫脲类化合物的结构/活性关系的理论解释.进行了人工神经网络法对于该类化合物的活性预测,其结果明显好于多元回归法.
The quantum-chemical descriptors, topological indices Am (Am1, Am2, Am3) and molecular connectivity indices mx1 of the 24 2-formylpyridine thiosemicarbazone compounds were calculated. The algorithms orthogonal descriptors and Leaps-and-Bonds regression were performed for variables selection. Multiregression and the quasi-Newton neural network calculations have been carried out, the results obtained by using neural network are much better than those obtained by using multiregression. The activity mechanisms have been briefly discussed.
出处
《河北工业大学学报》
CAS
2002年第5期80-84,共5页
Journal of Hebei University of Technology