人工神经元网络辅助酚类化合物构效关系研究

 分子连接性指数（^tXv）与在正辛醇和水之间的分配系数logP是反映化合物性能的重要结构参数。本文计算了酚类化合物的分子连接性指数及分配系数，运用新型的模式识别方法──人工神经元网络对酚类化合物的构效关系进行了研究。所得结果优于逐步判别法，对于预测未知化合物的毒性，具有重要意义。

ARTIFICIAL NEURAL NETWORKS AIDED STRUCTURE-ACTIVITY RELATIONSHIP STUDY OF PHENOLIC COMPOUNDS

t is well known that molecular connectivity index(tXv) and logP(the logarithm of partition coefficent between n-octanol and water)are important structure parameters that are relative to properties of compounds.Molecular connectivity index and logp of phenolic compounds are calculated. A new pattern recognition method-artificial neural networks has been used to investigate the structure-activity relationship (SAR) of phenolic compounds.The results obtained with these methods compared favourably with that by using conventional stepwise discrimination method.It is useful for the prediction of toxicity of unknown compounds.