论文部分内容阅读
目的评价支持向量机回归算法(SVR)和Tanagra软件用于计算抗丝虫药物定量构效关系(QSAR)的性能。方法设计了基于Tanagra软件的SVR模型,优化模型参数计算抗丝虫药物的QSAR,并将计算结果与多元线性回归方法的结果进行比较。结果所用算法计算精度高,预测能力强,计算结果接近活性实验值,训练集和测试集的Pseudo-r2分别为0.9842、0.9057,结果明显优于多元线性回归方法。结论基于Tanagra软件的SVR可有效地改善抗丝虫药物的QSAR计算,也适用于其他药物的QSAR分析。
Objective To evaluate the performance of SVR and Tanagra software in the calculation of quantitative structure-activity relationship (QSAR) of antifungi drugs. Methods An SVR model based on Tanagra software was designed and the model parameters were calculated to calculate the QSAR of antifungi drugs. The calculated results were compared with the results of multivariate linear regression. Results The algorithm used in this paper has high computational accuracy and strong predictive ability. The calculated results are close to the experimental ones. The Pseudo-r2 of the training set and the test set are 0.9842 and 0.9057, respectively. The result is obviously better than the multivariate linear regression method. Conclusions The SVR based on Tanagra software can effectively improve the QSAR calculation of anti-filarial drugs and QSAR analysis of other drugs.