Molecular Similarity Analysis as Tool to Predict Environmental Properties and Prioritize Research among Emerging Contaminants in the Environment

The very large number of emerging contaminants entering the water supply makes it desirable to assess their environmental fate and behavior without direct measurements. Structure-property prediction models are promising in this regard. Although traditional, regression-based structure-property relationships have been proven to be accurate for prediction of some parameters, these regression models are either too narrowly-defined to be of practical benefit for more than one small class of chemicals, or so broad that data scarcity compromises their predictive accuracy. Quantitative molecular similarity assessment (QMSA) is one particularly appealing alternative to traditional, regression-based models. QMSA models are based on the assumption that “similar” molecules behave “similarly”, such that parameters of interest for a target chemical can be computed based on parameter values for structurally similar chemicals. This is an evolving technique, which tends to be particularly appealing for applications in which predictive accuracy may be hampered by limited data availability.
This research has three main objectives: 1) demonstrating that QMSA models can accurately predict environmental parameters of interest for highly diverse chemical classes, then measuring fundamental fate and transport parameters for several key emerging contaminants, to externally validate QMSA hypotheses; 2) applying measured fate parameters to predict the fate of emerging contaminants in WWTPs; and 3) assessing the extent to which QMSA can help prioritize among unmeasured chemicals and determine which additional measurements will result in maximally increased model accuracy.
The results of this research are promising. Virtual experiments showed that valid QMSA models have been created for accurate prediction of three distinct environmental parameters: in vitro estrogenicity, sorption distribution coefficient Kd, and pseudo first-order biodegradation rate constant kb. Laboratory experiments for this research have focused on the measurement of Kd and kb for three highly-prescribed pharmaceutical (i.e., metformin, fluconazole, and benazepril), in wastewater obtained from a municipal wastewater treatment plant. Measured Kd and kb values are consistent with QMSA model predictions; furthermore, incorporation of these two parameters into simple mass balance models accurately predicts the effluent concentrations of studied emerging contaminants in WWTPs, demonstrating the usefulness of both QMSA and simple mass balance models. Finally we showed that three proposed QMSA-based prioritization approaches affords better improvement in QMSA estimation of property values among the remaining unmeasured compounds than random selection. This criterion provides additional information on selection of unmeasured emerging contaminants in terms of improving QMSA models’ accuracy.