TAGRA 2025

Aromatic micropollutants (AMPs) are a group of organic chemical compounds that have an aromatic ring structure and may be substituted with various functional groups such as hydroxyl, amino, nitro or halogen. These pollutants are found in very low concentrations in surface water, ground water and waste water. Investigating the relationship between the chemical structure of Aromatic micropollutants and their reactivity in photocatalytic processes is crucial to improve the efficiency of removal methods and predict the environmental behavior of these pollutants. Developing models based on the structural features of AMPs can help predict the rate of degradation and understand their ultimate fate in nature. In this study, quantitative structure-activity relationship (QSAR) models were developed to predict the photooxidation reaction of aromatic micro-pollutants (AMPs) using the multiple linear regression (MLR) and support vector machine (SVM). The dataset consisted of 30 compounds, were divided into two training and test subsets by hierarchical clustering method. Genetic algorithm (GA) was used as a feature selection tool to identify the most relevant molecular descriptors. Model validation was performed using Y- randomization test, cross-validation, and external test set methods. The genetic algorithm- multiple linear regression (GA-MLR) model with three selected descriptors showed favorable statistical parameters (R2train=0.822, R2test=0.920). Comparison of the models' performance shows that, GA-SVM provided accurate results with strong statistical parameters for the training and test data sets (R2train=0.939, R2test=0.922). Analysis of the results showed that spheroid, atomic masses, and also difference between partial positively- and negatively-charged surface areas, of molecules play a decisive role in their activity. The developed models can be used as efficient tools in the targeted design of high-performance AMPs and understanding its photooxidation reaction behavior. The comparison between different models, allowed us to examine the advantages and limitations of linear and nonlinear models in analyzing the structure-reactivity relationship of Aromatic micropollutants.