论文部分内容阅读
为提高定量构效关系(quantitative structure-activity relationship,QSAR)模型预测的精度,以支持向量回归(support vector regression,SVR)全局与局部核函数,发展出1种非线性组合方法GK-LK-SVR,其基本思路为:依均方误差(MSE)最小原则,分别基于SVR的全局与局部核函数筛选描述符后预测,实测值与不同核函数的预测值组合成混合样本,然后再依MSE最小原则基于SVR对混合样本实施核函数寻优及子模型筛选,最后以留一法完成预测。对2种化合物QSAR建模结果表明:GK-LK-SVR方法预测精度高,有望在QSAR研究中得到广泛应用。
In order to improve the prediction precision of quantitative structure-activity relationship (QSAR) model, a nonlinear combination method, GK-LK-SVR, is developed by using global and local kernel functions of support vector regression (SVR) , The basic idea is as follows: According to the minimum mean square error (MSE) principle, the global and local kernel function descriptors based on SVR are respectively used to predict, and the measured values are combined with the predicted values of different kernel functions into a mixed sample, The principle is based on SVR to perform the optimization of the kernel function and the screening of the sub-model for the mixed samples. Finally, the prediction is made by leaving one method. The modeling results of two compounds QSAR show that the GK-LK-SVR method has high prediction accuracy and is expected to be widely used in QSAR research.