基于遗传算法的PLS分析在QSAR研究中的应用

Application of GA-Based PLS Analysis in QSAR Studies

将遗传算法和偏最小二乘法结合应用于定量结构活性关系研究中 ,进行变量选择和建立最终模型 .以较少的变量个数包含较多的变量信息 ,且变量间没有线性相关问题 ,因而得到较好的 QSAR模型 .将这种方法应用于氯代酚和单取代苯 2种系列化合物 ,可以得到几种常规多元回归分析方法不能得到的质量较高的 QSAR模型 .

Genetic algorithms(GA)are an effective optimization method,and partial least squares(PLS)are a well\|known multivariate statistical method.The combination of two methods is proposed to construct GA\|based PLS analysis,which is applied to QSAR studies for variable selection and model construction.By the method,a few latent variables can include the information of more variables,thus better QSAR models can be obtained.The method is applied to chlorophenols and mono\|substituted benzene derivatives,and some models are better than the best multivariate regression results.