以人工神经网络误差反传算法对２羟基〔１Ｈ〕异吲哚二酮衍生物构效关系进行了探讨，获得了精密的拟合和准确的预测，拟合结果优于多元线性回归．该方法可作为一种有效的定量构效关系研究工具． The structure-activity relationship of 2hydroxy  [¹H] -isoindodione derivatives was studied by artificial neural network error back propagation algorithm, and the fitting precision and accurate prediction were obtained. The fitting result was better than multivariate linear regression . This method can be used as an effective quantitative structure-activity relationship research tool.