Prediction For Gas-Particle Partition Coefficient And Octanol-Air Partition Coefficient Of Selected Organic Compounds

The partition coefficient between particulate and gas phases (K_p) for organic pollutants areof great importance to characterize the behavior of organic pollutants in atmosphere, and thevalue of K_p is notably correlated to K_OA, octanol/gas partition coefficient. K_OA is a veryimportant physicochemical parameter for describing the partition of organic chemicalsbetween air and environmental organic phases and is basic data needed by ecological riskassessment. Experimental determination of both K_p and K_OA are time- consuming, expensiveand limited by standard sample, which lag behind the synthesis and use of organic chemicals.It is thus of theoretical and practical importance to develop predicting models for K_p and K_OA.Partial least squares (PLS) and support vector mechanism (SVM) regression with 16theoretical molecular structural descriptors calculated by PM3 arithmetic were used todevelop quantitative structure-activity relationship (QSAR) model for K_p of 18 aliphatichydrocarbons, 21 polycyclic aromatic hydrocarbons (PAHs), 16 polychlorinated biphenyls(PCBs) and 13 polychlorinated dibenzo-p-dioxins/dibenzofurans (PCDD/Fs). The temperaturedependence model of PCDD/Fs was also developed to reveal the effect of environmentaltemperature on the partition process.The obtained models have good predictive ability, robustness, and nonsystematic residuals,which can be used for estimating K_p of chemicals within applicability domain. Intermoleculardispersion interactions, charge-transfer interactions and hindrance effects of molecular sizedetermine the values of K_p, with the hindrance effects playing a negative role in the partitionto particulate phase. The result of temperature dependence model of PCDD/Fs indicates thatenvironmental temperature is an important factor controlling the partition of organicchemicals between gas and particle phases, and included in the form of square in the model.The comparison of models developed by PLS and SVM indicates that SVM methord canextract the information of relationship between molecular structural parameters and K_p moreadequately than PLS, resulted in the model with better goofness-of-fitting and predictivepower. This phenomenon reveals that there may be some nonlinear relationship betweenmolecular structures which affect the partition and K_p.The values of logK_OA of 246 organic chemicals were obtained by thermodynamicequilibrium equation basing solvation free energy calculated using OMNISOL SM5.0. Though the calculated logK_OA values are not very exact because of the shortcoming ofsoftware, there is good correlationship between calculated and observed values of logK_OA. Sowe can adjust the calculated values by this correlationship to obtain more accurate logK_OAvalues.