论文部分内容阅读
【目的】驱避剂可使害虫不敢接近受用者从而保护受用者免遭其害。建立高精度、可解释性强的非线性定量构效关系(quantitative structure-activity relationship,QSAR)模型对设计合成新的高效昆虫驱避剂有重要意义。【方法】基于37个芳香羧酸类化合物对家蝇Musca domestica的驱避活性,以量子化学计算软件PCLIENT获取每一化合物初始描述符,以二元矩阵重排过滤器、多轮末尾淘汰实施特征非线性筛选,以支持向量回归(support vector regression,SVR)建立非线性QSAR模型,以SVR非线性解释体系分析各保留描述符对驱避活性的影响。【结果】1 542个初始描述符的SVR模型F=1.2,特征筛选后6个保留描述符的SVR模型F=184.6,特征筛选对QSAR模型精度有重要影响。6个保留分子描述符的重要性依次为p4BCD>GATS7v>T(O..O)>JGI8>SssO>nArCONR2。【结论】保留描述符与芳香羧酸类化合物对家蝇驱避活性的非线性关系明显,获得了高精度、普适性强的非线性SVR-QSAR模型。
[Purpose] Repellent can prevent pests from approaching the user to protect the user from harm. The establishment of a high-precision and interpretative model of quantitative structure-activity relationship (QSAR) is of great significance for the design and synthesis of new insect repellent. 【Method】 Based on the repellent activity of 37 aromatic carboxylic acids on housefly Musca domestica, the initial descriptors of each compound were obtained by using PCLIENT, a quantum chemical calculation software. The filter was rearranged by binary matrices. Non-linear screening was used to establish a nonlinear QSAR model with support vector regression (SVR). The SVR nonlinear interpretation system was used to analyze the effect of each retention descriptor on repellent activity. 【Result】 The results showed that the SVR model F = 1.2 for 1 542 initial descriptors and F = 184.6 for the six retention descriptors after feature selection. Feature filtering has an important impact on the accuracy of QSAR models. The importance of six reserved molecule descriptors is p4BCD> GATS7v> T (O..O)> JGI8> SssO> nArCONR2. 【Conclusion】 The non-linear relationship between retention descriptors and repellent activity of aromatic carboxylic acids on housefly was significant. A nonlinear SVR-QSAR model with high precision and universality was obtained.