The removal of organic micropollutants (MPs) from water by means of adsorption is determined by the properties of the adsorbent and the physicochemical properties of the MPs. However, it is not yet possible to predict the removal of MPs by a specific adsorbent based only on their physicochemical properties. In this research, we developed a novel statistical framework to establish Quantitative Structure–Activity Relationship (QSAR) models between the physicochemical properties of a set of 153 MPs and their distribution coefficients (Kd) measured for a β-cyclodextrin polymer (CDP) and activated carbon (AC) adsorbents. Adsorption isotherms were developed from batch experiments for a mixture of 196 structurally diverse MPs at 2 µg/L with CDP and AC at concentrations ranging between 5 mg/L and 100 mg/L. The batch isotherm data was used to estimate Kd for 153 of the MPs. Computational software (Marvin and PaDel-Descriptor) were used to calculate 1195 molecular descriptors for each of the 153 MPs to describe their physicochemical properties. A combined elastic-net and stepwise regression framework was then constructed to identify the most significant descriptors for CDP from all 1195 molecular descriptors. An all-possible regressions analysis was then employed to exhaust all sub-combinations of the selected 26 predictors to evaluate the significance and performance of each model scenario. The same approach was used to develop multivariate models describing adsorption on AC for the same group of MPs. The cross-validated determination coefficient Q2 for all models is over 0.90, which demonstrates the functional stability and predictive power of the resulting QSARs and confirms the significant relationship between the selected molecular descriptors and Kd. The application domain of the built QSARs was characterized by the leverage approach using a Williams plot, with the leverages of the MPs (diagonal elements of the Hat matrix) versus the studentized residuals of the MPs. These data provide important insights into the mechanisms by which MPs adsorb to CDP and AC and the resulting models have utility as predictive tools for evaluating adsorption-based water treatment processes.

QSARs to predict distribution coefficients of polar organic micropollutants on activated carbon and β-cyclodextrin polymers

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.