Abstract:Developing quantitative structure-activity relationships (QSARs) is an important approach to predicting the reactivity of HO radicals toward newly emerged organic compounds. As compared with molecular descriptors-based and the group contribution method-based QSARs, a combined molecular fingerprint-machine learning (ML) method can more quickly and accurately develop such models for a growing number of contaminants. However, it is yet unknown whether this method makes predictions by choosing meaningful structural features rather than spurious ones, which is vital for trusting the models. In this study, we developed QSAR models for the logkHO values of 1089 organic compounds in the aqueous phase by two ML algorithms—deep neural networks (DNN) and eXtreme Gradient Boosting (XGBoost), and interpreted the built models by the SHapley Additive exPlanations (SHAP) method. The results showed that for the contribution of a given structural feature to logkHO for different compounds, DNN and XGBoost treated it as a fixed and variable value, respectively. We then developed an ensemble model combining the DNN with XGBoost, which achieved satisfactory predictive performance for all three datasets: Training dataset: R-square (R2) 0.89–0.91, root-mean-squared-error (RMSE) 0.21–0.23, and mean absolute error (MAE) 0.15–0.17; Validation dataset: R2 0.63–0.78, RMSE 0.29–0.32, and MAE 0.21–0.25; and Test dataset: R2 0.60–0.71, RMSE 0.30–0.35, and MAE 0.23–0.25. The SHAP method was further used to unveil that this ensemble model made predictions on logkHO based on a correct 'understanding' of the impact of electron-withdrawing and -donating groups and of the reactive sites in the compounds that can be attacked by HO. This study offered some much-needed mechanistic insights into a ML-assisted environmental task, which are important for evaluating the trustworthiness of the ML-based models, further improving the models for specific applications, and leveraging the implicit knowledge the models carry.

Unveiling similarities and differences in oxidation processes of oxidants and derived reactive oxygen species through machine learning interpretation of oxidation rate constants

An innovative method for predicting oxidation reaction rate constants by extracting vital information of organic contaminants (OCs) based on diverse molecular representations

Predictive models for the aqueous phase reactivity of inorganic radicals with organic micropollutants

Prediction of hydroxyl radical exposure during ozonation using different machine learning methods with ozone decay kinetic parameters

A new tool to predict the advanced oxidation process efficiency: using machine learning methods to predict the degradation of organic pollutants with Fe-carbon catalyst as a sample

Integrated model of ozone mass transfer and oxidation kinetic: Construction, solving and analysis

Modeling the aqueous phase reactivity of hydroxyl radical towards diverse organic micropollutants: An aid to water decontamination processes

Prediction of Free Radical Reactions Toward Organic Pollutants with Easily Accessible Molecular Descriptors.

Utilizing Machine Learning Models with Molecular Fingerprints and Chemical Structures to Predict the Sulfate Radical Rate Constants of Water Contaminants

Modeling the pH and temperature dependence of aqueousphase hydroxyl radical reaction rate constants of organic micropollutants using QSPR approach

Predicting Reaction Rate Constants of Organic Compounds with Oxidants in the Atmospheric Aqueous-Phase Through Multi-Task Learning

Machine learning approaches to predict the apparent rate constants for aqueous organic compounds by ferrate

The structure-activity relationship of aromatic compounds in advanced oxidation processes:a review

Kinematics Modeling of Peroxy Free Radicals: A Deep Reinforcement Learning Approach

Oxidation kinetics of water contaminants: New insights from artificial intelligence

Machine Learning-Assisted QSAR Models on Contaminant Reactivity Toward Four Oxidants: Combining Small Data Sets and Knowledge Transfer

Different Reaction Mechanisms of SO4•− and •OH with Organic Compound Interpreted at Molecular Orbital Level in Co(II)/peroxymonosulfate Catalytic Activation System

Shedding light on “Black Box” machine learning models for predicting the reactivity of HO radicals toward organic compounds

Machine learning for predicting halogen radical reactivity toward aqueous organic chemicals

Assessing the Use of Probes and Quenchers for Understanding the Reactive Species in Advanced Oxidation Processes

Quantitative structure-activity relationship models for predicting singlet oxygen reaction rate constants of dissociating organic compounds