论文部分内容阅读
根据分子顶点特征值构建了一种分子结构描述符,采用该描述符对86个醛、酮类化合物进行了结构表征,并与沸点(BP)建立了定量结构—性质相关(QSPR)模型。多元线性回归(MLR)的复相关系数(R)为0.989,交互检验复相关系数(RCV)为0.985;偏最小二乘回归(PLS)模型复相关系数(R)为0.983,交互检验复相关系数(RCV)为0.967。结果表明,构建的结构描述符能较好地表征醛酮类化合物分子的结构信息,所建模型具有良好的稳定性与预测能力。
A molecular structure descriptor was constructed based on the molecular vertex eigenvalues. The structures of 86 aldehydes and ketones were characterized by this descriptor, and the quantitative structure-property correlation (QSPR) model was established with the boiling point (BP). The multiple correlation coefficient (R) of multiple linear regression (MLR) was 0.989 and the RCV of interaction test was 0.985. The complex correlation coefficient (R) of partial least squares regression model was 0.983, and the correlation coefficient (RCV) was 0.967. The results show that the constructed structure descriptors can better characterize the structural information of aldehydes and ketones. The model has good stability and predictability.