脂肪醇临界温度与分子描述符的定量结构性质关系研究

Quantitative structure property relationship between critical temperature of aliphatic alcohols and molecular descriptors

采用Gaussian-16软件对脂肪醇分子进行了密度泛函理论计算,在6-311g(d,p)基组下进行几何优化和频率分析,得到每个醇分子最稳定的构象,将此构象的几何参数交给Alvadesc软件计算分子描述符。运用逐步多元线性回归技术找到了对脂肪醇临界温度有显著贡献的五个分子描述符,分别是piPC01,ATS6e,GATS2e,GATS3i和E2m;这些分子描述符描述了脂肪醇分子的结构、大小和连接信息;建立了五参数多元线性回归定量结构性质关系(QSPR)模型,模型的相关系数为0.9998,均方根误差仅为1.77 K,经Fisher检验、t-检验、方差膨胀因子检验、留一法交叉验证、留多法交叉验证以及外部测试集验证,所得QSPR模型稳健、可靠,能用于预测脂肪醇的临界温度。

Gaussian-16 software was used to calculate the aliphatic alcohol molecules with density function theory,geometry optimization and frequency analysis were performed on 6-311g(d,p)basis set to obtain the most stable conformation of each alcohol molecule,and the geometric parameters of the conformation were uploaded to Alvadesc software to calculate the molecular descriptors.Using stepwise multiple linear regression techniques,five molecular descriptors with significant contributions to the critical temperature of aliphatic alcohols were found and termed piPC01,ATS6e,GATS2e,GATS3i,and E2m,respectively.These molecular descriptors could describe the structure,size and connection information of alcohol molecules.The correlation coefficient of the proposed five-parameter quantitative structure property relationship(QSPR)model was 0.9998 with root mean square error of only 1.77 K.The results of Fisher test,t-test,variance inflation factor test,leave-one-out cross-validation,leavemany-out cross-validation and external test set verification show that the QSPR model is robust and reliable,which can be used to predict the critical temperature of aliphatic alcohols.