Key Lab of Green Processing & Functional Textiles of New Textile Materials, Ministry of Education, Wuhan Textile University, China.
SAR QSAR Environ Res. 2013 Jan;24(1):47-59. doi: 10.1080/1062936X.2012.728997. Epub 2012 Oct 16.
A quantitative structure-property relationship (QSPR) model was proposed between the molecular descriptors representing the molecular structure and the Freundlich adsorbability parameter (K) for a set of 55 organic compounds onto activated carbon cloth. The best linear model was composed of three descriptors, which were selected by stepwise multiple linear regression (MLR) analysis. The statistical parameters provided by the linear model were r² = 0.7744, r²(adj) = 0.7551, s = 0.169 for the training set; and r² = 0.6725, r²(adj) = 0.6316, s = 0.196 for the external test set, respectively. The stability and predictive power of the proposed model were further verified using Y-randomization tests, five-fold cross-validation and leave-many-out cross-validation. The model may give some insight into the main structural features that affect the adsorption of the investigated compounds onto activated carbon cloth.
提出了一个定量构效关系（QSPR）模型，用于将代表分子结构的分子描述符与活性炭布上 55 种有机化合物的 Freundlich 吸附参数（K）联系起来。最佳线性模型由三个描述符组成，这些描述符是通过逐步多元线性回归（MLR）分析选择的。线性模型提供的统计参数为：训练集的 r²=0.7744，r²（adj）=0.7551，s=0.169；外部测试集的 r²=0.6725，r²（adj）=0.6316，s=0.196。使用 Y 随机化测试、五重交叉验证和留一法交叉验证进一步验证了所提出模型的稳定性和预测能力。该模型可以深入了解影响所研究化合物在活性炭布上吸附的主要结构特征。
