Estimation Models to Understand the Chemodynamics of Diverse Organic Pollutants for Organisms and Passive Sampler

Abstract

This thesis sheds light on the mechanistic aspects of chemodynamics of organic pollutants in three environmentally-important phases: fish, human skin and polydimethylsiloxane (PDMS) passive samplers. Thesis starts with the probing of variability embedded in depuration half-life data for a diverse set of chemicals using different sets of inter-molecular interactions. I started with Abraham solvation parameters and found that together with fish weight information (𝑊𝑓), molecular polarizability ( 𝐸), hydrogen donating capability (𝐵) and the size ( 𝑉) of molecule play important role in describing more than 85% of variability in half-lives of a diverse set of chemicals. Models based on three types of descriptors-equilibrium coefficients (for air-water, octanol-water, organic carbon-water, and organism-water partitioning systems), rate-related coefficient (diffusion in water and ethanol, biotransformation rate), and allometric coefficient (𝑊𝑓)- were used to estimate the depuration half-lives. The model using descriptors of bioconcentration factor (𝐵𝐶𝐹), biotransformation rate constant (𝑘𝑀), and 𝑊𝑓-referred to as BIOCEF model in this study-outperformed other models with R2 of 0.87 and root-mean-square error (RMSE) of 0.44 log unit, when compared to experimental data. Finally, we showed that the estimates of depuration half-life can be directly applied to nonpolar chemicals detected on GC×GC chromatograms with an RMSE of 0.64 log units. In the second part of this thesis, I developed a model to estimate skin – permeability coefficients (𝐾𝑝), by modifying the previous approach. The training dataset was diverse and comprised of representatives of different chemical families. The new two-parameter (2p) model comprising of octanol – water ( 𝐾𝑜−𝑤) and air – water (𝐾𝑎−𝑤) partition coefficients. The 2p- model were able to explain 77% of variability in the dataset with RMSE of 0.53 log unit. The performance of my new model was compared with previously used model, DERMWIN, which focus only on one partitioning property 𝐾𝑜−𝑤 and 𝑀𝑊 and Abraham solvation model have limitation of data accessibility for billions of chemicals. In third phase, two models were developed to estimate the time for chemicals to reach 95 % of equilibrium state by using PDMS - log τ 95-PDMS Model based on inter molecular interaction parameters - (𝐵, 𝑉) was best performed, while other model having partitioning descriptors- 𝐾𝑜−𝑤,𝐾𝑎−𝑤,𝐾𝑜−𝑐, was also showing good predictive efficiency. All studied aspects of chemodynamics explained that partitioning coefficients, diffusion coefficients and intermolecular interaction parameters are important descriptors to understand the transport of chemicals across or near different interfaces.