ABSTRACT

The Monte Carlo algorithm was applied to formulate a robust quantitative structure–property relationship (QSPR) model to compute the reactions rate constants of hydrated electron values for a data set of 309 water contaminants containing 125 aliphatic and 184 phenyl-based chemicals. The QSPR models were computed with the hybrid optimal descriptors which were procured by combining the SMILES and hydrogen-suppressed molecular graph for both classes of compounds. Approximately 75% of the total experimental data set was randomly divided into training and invisible training sets, while approximately 25% was divided into calibration and validation sets. The authenticity and robustness of the developed QSPR models were also judged by the Index of Ideality of Correlation. In QSPR modelling of aliphatic compounds, the numerical values of $r_{\mathrm{T}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}}^2$, $r_{\mathrm{V}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}^2$, $Q_{\mathrm{T}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}}^2$and $Q_{\mathrm{V}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}^2$ were in the range of 0.852–0.905, 0.815–0.894, 0.839–0.897 and 0.737–0.867, respectively. Whereas, in the QSPR modelling of phenyl-based compounds, the numerical values of $r_{\mathrm{T}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}}^2$, $r_{\mathrm{V}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}^2$, $Q_{\mathrm{T}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}}^2$and $Q_{\mathrm{V}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}^2$ were in the range of 0.867–0.896, 0.852–0.865, 0.816–0.850 and 0.760–0.762, respectively. The structural attributes, which are promoters of $\mathrm{l}\mathrm{o}\mathrm{g}{K}_{e_{aq}^{-}}$increase/decrease are also extracted from the SMILES notation for mechanistic interpretation. These QSPR models can also be applied to compute the reaction rate constants of organic contaminants.

摘要

蒙特卡罗算法用于建立鲁棒的定量结构-性质关系（QSPR）模型，以计算309种水污染物的数据集的水合电子值的反应速率常数，该数据集包含125种脂族和184种基于苯基的化学物质。QSPR模型是用混合最佳描述子计算的，该描述子是通过将两种化合物的SMILES和氢抑制分子图相结合而获得的。总实验数据集中约有75％被随机分为训练和隐形训练集，而约25％被分为校准和验证集。所开发的QSPR模型的真实性和鲁棒性也由相关理想指数来判断。在脂族化合物的QSPR建模中，${\mathrm{[R}}_{\mathrm{Ť}\mathrm{[R}\mathrm{一种}\mathrm{一世}\mathrm{ñ}\mathrm{一世}\mathrm{ñ}\mathrm{G}}^2$， ${\mathrm{[R}}_{\mathrm{V}\mathrm{一种}\mathrm{升}\mathrm{一世}\mathrm{d}\mathrm{一种}\mathrm{Ť}\mathrm{一世}\mathrm{Ø}\mathrm{ñ}}^2$， ${问}_{\mathrm{Ť}\mathrm{[R}\mathrm{一种}\mathrm{一世}\mathrm{ñ}\mathrm{一世}\mathrm{ñ}\mathrm{G}}^2$和 ${问}_{\mathrm{V}\mathrm{一种}\mathrm{升}\mathrm{一世}\mathrm{d}\mathrm{一种}\mathrm{Ť}\mathrm{一世}\mathrm{Ø}\mathrm{ñ}}^2$分别在0.852-0.905、0.815-0.894、0.839-0.897和0.737-0.867范围内。而在基于苯基的化合物的QSPR建模中，${\mathrm{[R}}_{\mathrm{Ť}\mathrm{[R}\mathrm{一种}\mathrm{一世}\mathrm{ñ}\mathrm{一世}\mathrm{ñ}\mathrm{G}}^2$， ${\mathrm{[R}}_{\mathrm{V}\mathrm{一种}\mathrm{升}\mathrm{一世}\mathrm{d}\mathrm{一种}\mathrm{Ť}\mathrm{一世}\mathrm{Ø}\mathrm{ñ}}^2$， ${问}_{\mathrm{Ť}\mathrm{[R}\mathrm{一种}\mathrm{一世}\mathrm{ñ}\mathrm{一世}\mathrm{ñ}\mathrm{G}}^2$和 ${问}_{\mathrm{V}\mathrm{一种}\mathrm{升}\mathrm{一世}\mathrm{d}\mathrm{一种}\mathrm{Ť}\mathrm{一世}\mathrm{Ø}\mathrm{ñ}}^2$分别在0.867-0.896、0.852-0.865、0.816-0.850和0.760-0.762范围内。结构属性，是$\mathrm{升}\mathrm{Ø}\mathrm{G}{ķ}_{{Ë}_{\mathrm{一种}q}^{-}}$还从SMILES标记中提取增加/减少以进行机械解释。这些QSPR模型也可以用于计算有机污染物的反应速率常数。