Development And Validation Of Qsar Predictive Model On(Subcooled) Liquid Vapor Pressure Of Organic Chemicals

(Subcooled) liquid vapor pressure (Pl) is an key physicochemical property govinging the environmental fate of organic chemicals. However, it will be impossible if we measure the P^data one by one due to the huge number of chemicals, as well as the analytical difficulties. Quantitative structure-activity relationship (QSAR) is adopted to predict the Pi of organic chemicals. It will provide the basic data for the description of environmental fate of chemicals. Also this will make the ecological risk assessment more effective and convient.The linear relationships between the molecular parameters characterizng the structural information and_PL for organic chemicals were developed and the main results are summarized as follows:(1) The experimental Pl data of439organic chemicals at298K were collected from peer-reviewed scientific publications. The quantum chemical parameters were caculated at the semi-empirical PM6level and the predictive model were developed employing partial least square (PLS) regression method. Satisfactory goodness-of-fit, robustness and the external predictive performance were demonstrated by the value of square of multiple correlation coefficient (R2=0.906), root mean square error (RMSE=0.776), coefficient of cross validation (Q2CV=0.899) and coefficient of external validation (Q2EXi=0.863). Moreover, the mechanistic interpretation, which was of significance for the reliability of model, was also discussed after the model trained. The main factors influencing Pl were dispersive force, which was dependent on the size of the molecular, polarizability, and ability of formatting intermolecular hydrogenbon bond. The bigger the dispersive force is, the higher the Pl will be.(2) The10396experimental Pl data of661chemicals under various temperatures were collected from peer-reviewed scientific publications. The temperature-dependent predictive model for vapor pressure was developed applying PLS regression method. The optimal model involved eight descriptors including I IT. Satisfactory goodness-of-fit, robustness and the external predictive performance were demonstrated by the value of square of multiple correlation coefficient (R2=0.923), root mean square error (RMSE=0.447), coefficient of cross validation (^v-0.921) and coefficient of external validation (Q2EXi=0.919). Moreover, the definition of applicability domain and mechanistic interpretation, which were of significance for the reliability of model, were also discussed after the model trained. The representation of applicability domain showed that the model was suitable for the aliphatic hydrocarbons and aromatic hydrocarbons, as well as their nitrogenous, oxygenous, and sulfides substitutes. The main factors influencing the vapor pressure wereI/T, van der Waals volume, distribution of the charges, number of hydrogen bonds and polar surface area.