PREDICTING THE DIFFUSION POTENTIAL OF DIVERSE ORGANIC CONTAMINANTS IN WATER AND ETHANOL PHASES BY FORMULATING NEW 2- PARAMETER MODELS

Abstract

Information about diffusivity of diverse organic contaminants in water and ethanol phases provides insight about their chemo dynamics in aquatic and mucosal phases. The experimental techniques to measure diffusivities are expensive, time- consuming, and not readily available. On the other hand, the existing computational methods are parameter intensive, computationally expensive, and have a limited domain of applicability. Here, I present a simple estimation approach based on solvation properties of contaminants using widely available partition coefficients data for the octanol-water (logK ow ) and air-water (logK aw ) systems. For this purpose, the experimental data of diffusion coefficients for 133 polar and non-polar compounds in water and 58 compounds in ethanol — taken from literature — were used to train and validate the models. Diffusion coefficients in ethanol showed an excellent correlation with logK aw and other intermolecular interaction parameters. The model based on linear combination of logK ow and logK aw was developed with an R 2 of 0.80 and an RMSE of 0.060. In this model, logK aw exhibited a very influential role in describing the variation in diffusivity data for ethanol phase. For water phase the variance in the diffusivities depicted an R 2 of 0.54 and an RMSE of 0.101 based on linear combination of logK ow and logK aw . The performance of our models was similar to widely used Abraham Solvation Model (ASM). However, the Abraham models are limited by the scarcity of experimental ASDs (<8000 chemicals). I evaluated the theoretical rigor and robustness by applying supervised and unsupervised machine learning techniques such as principal component analysis, Pearson correlation analysis, hold-out, leave one out, k-fold, and bootstrap approaches. The results of these analyses showed that the logK ow and logK aw respectively correspond to non-specific and specific intermolecular interactions. Furthermore, there was a good agreement between the train and test datasets for both models. Lastly, I parametrized a mass transfer model to evaluate the transport resistance of organic contaminants in water and ethanol using basic principles of Fick’s Law. Taken together this study sheds light on the kinetics of bioavailability of diverse organic contaminants.