优化的径向基-循环子空间网络为药物定量构效关系建模  

Optimized Radial Basis Functions-Cyclic Subspace Regression and Its Application to Quantitative Structure-activity Relationships

作　　者：李剑[1] 陈德钊[1] 吴晓华[1] 叶子清[1] 

机构地区：[1]浙江大学化工系仿真中心,杭州310027

出　　处：《分析化学》2005年第6期767-771,共5页Chinese Journal of Analytical Chemistry

基　　金：国家自然科学基金(No.20276063);浙江省重点科技(No.2004C21054);杭州市科技(No.2003131B07)资助项目

摘　　要：径向基循环子空间回归(RBFCSR)网络,保留了径向基偏最小二乘(RBFPLS)网络的优点,且可在更广的范围内选择最优模型,但仍存在着参数难以确定,计算量大等问题。对此,本研究兼顾网络模型的拟合与预测性能,采用具有高效全局搜优能力的优进遗传算法(EGA)优化网络参数,构建为EGARBFCSR方法,并将其成功应用于苯乙酰胺类除草剂的构效关系(QSAR)建模,效果良好,显示出很强的学习能力,所建模型具有良好的预报性能和稳定性,并优于其他方法。A new approach, radial basis functions-cyclic subspace regression ( RBF-CSR), was proposed based on the analyzing radial basis functions-patial least squares ( RBF-PLS). The approach has the merit of RBF-PLS, and it can select the optimal model in wider range. But selecting the parameter of the model is hard work. In this article, in order to solve the problem, an RBF-CSR base on eugenic evolution strategy genetic algorithm( EGA -RBF-CSR) was proposed, in which EGA was applied to optimize the parameter of RBF-CSR model for improving the performance of fitting and predicting. Finally, EGA-RBF-CSR was applied successfully to modeling quantitative structure-activity relationships. Compared to the other nonlinear models and linear models, The EGA-RBF-CSR model not only holds on fine learning ability but also gives better prediction performance and steady capability.

关 键 词：定量构效关系 径向基 空间网络 建模 优化 循环子空间回归 药物 酰胺类除草剂 偏最小二乘 最优模型 预测性能 网络模型 网络参数 遗传算法 学习能力 预报性能 计算量 稳定性 苯 

分 类 号：TQ460.1[化学工程—制药化工]