中文摘要：

针对药物构效关系呈非线性的特征，提出一种径基函数（radial basis function，RBF）-自适应偏最小二乘回归（adaptive partial least squares regression，APLSR）相结合的建模方法。该组合方法应用RBF实现自变量非线性变换，应用APLSR方法消除非线性变换后输出变量间存在的复共线性，并以模型的预报能力为目标，自适应地确定PLSR模型的最佳隐变量个数，从而获得预报性能良好的模型。本文将RBF-APLSR方法应用于含硫苯衍生物的定量构效关系建模，取得了令人满意的效果，其预报精度高于PLSR方法。

英文摘要：

There usually exist the nonlinear quantitative structure-activity relationships （QSAR） of drug and the significant correlation among structure parameters of the drug. Sometimes, the multicollinearity is even formed among the structure parameters. A novel modeling method integrating the radical basis function （RBF） with adaptive partial least squares regression （ APLSR）, which can describe complex nonlinear system, was proposed. Firstly, the method applies RBF to carry out the nonlinear transformation for independent variables. Secondly, APLSR is applied to remove the correlation among the nonlinear transformed variables and the optimal number of latent variables is obtained by APLSR according to the predicting correctness of the model. Further, RBF-APLSR was applied to model QSAR of the substituted aromatic sulfur derivatives. Satisfactory results were obtained.