QSAR Study Of Thiophenyl-carboxylates Derivatives

In organic contamination chemistry and ecological toxicology, the quantitative structure-activity relationship (QSAR) technology may have two basic functions. First of all, it can be used for predicting or assessing the unknown environmental behavior and ecological toxicity of some chemicals. On the other hand, QSAR is very helpful in studying the mechanism of toxicity for a variety of pollutants and is of theoretical significance in pollution control and risk minimization. Therefore, optimum QSAR models may not only have good predictive power, but also possess explicit physical interpretations. Usually quantum chemical calculation is an important way to get structural parameters of specific molecules in study of the structure-activity relationship of organic pollutants. With quantum chemical calculation, all the information about electronic and spatial structures of specific compounds can be obtained, including molecular orbital energy level, atomic charge density, polarizability, and electrostatic charge of molecules. As widely considered, ab initio is theoretically precise, and thus it has become an effective scientific tool in quantum chemical calculation worldwide, due to the rapid development of computer science. In the present study, Hartree-Fock (HF) and Density Functional Theory (DFT) with different basis sets were employed to calculate the structural and thermodynamic parameters of 66 thiophenyl, phenyl sulfoxide and phenylsulfonyl cycloalkane carboxylates. With the statistic software SPSS11.5 for windows and GQSARF2.0, the QSAR models relating physicochemical properties of these compounds to their structural and thermodynamic parameters were successfully obtained, using multiple stepwise regression method. The physicochemical properties observed in this work include octanol/water partition coefficient (Kow), chromatographic capacity index (Kw), toxicity (EC50 and LC50), and solubility (Sw) etc. To verify the reliability of the present QSAR models, collinearity of the variables in the model was examined first, and then partial example-based approach and leave-one-out (LOO) cross validation method was used for predicted checkout of these models. The result showed that in all the correlation models achieved from diverse basis sets and calculation methods, the models obtained from B3LYP/6-31G* method are advantageous to those from other calculation methods. So, it is suggested that high-precision method and relatively large basis set may result in high-correlation equations. In addition, the activity mechanism of phenylsulfonyl cycloalkane carboxylates was also investigated based on the relevant QSAR models. With observation of these models, the standard regression coefficient of polarizability in all models is obviously higher than that of other variables, implying that polarizability (or molecular volume) may have most effect on the activity of phenylsulfonyl cycloalkane carboxylates. It is also found that if structural parameters together with thermodynamic ones are taken as theoretical parameters, the QSAR models would have relatively higher predictive capacity than those with structural parameters or thermodynamic parameters, respectively.