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.

Molecular image-convolutional neural network (CNN) assisted QSAR models for predicting contaminant reactivity toward OH radicals: Transfer learning, data augmentation and model interpretation

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

Back-Propagation Network Improved by Conjugate Gradient Based on Genetic Algorithm in Qsar Study on Endocrine Disrupting Chemicals

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

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

AI-enhanced chemical paradigm: From molecular graphs to accurate prediction and mechanism

Prediction of molecular-specific mutagenic alerts and related mechanisms of chemicals by a convolutional neural network (CNN) model based on SMILES split

Molecular Graph Encoding Convolutional Neural Networks for Automatic Chemical Feature Extraction.

Predicting Structure‐dependent Properties Directly from the Three Dimensional Molecular Images Via Convolutional Neural Networks

Prediction of the Aquatic Toxicity of Aromatic Compounds to Tetrahymena Pyriformis Through Support Vector Regression

CRNNTL: Convolutional Recurrent Neural Network and Transfer Learning for QSAR Modeling in Organic Drug and Material Discovery

Using Support Vector Regression Coupled with the Genetic Algorithm for Predicting Acute Toxicity to the Fathead Minnow

Inductive transfer learning for molecular activity prediction: Next-Gen QSAR Models with MolPMoFiT

Advancing chronic toxicity risk assessment in freshwater ecology by molecular characterization-based machine learning

A Deep Learning-Based Chemical System for QSAR Prediction

Predicting Chemical Immunotoxicity through Data-Driven QSAR Modeling of Aryl Hydrocarbon Receptor Agonism and Related Toxicity Mechanisms

CRNNTL: convolutional recurrent neural network and transfer learning for QSAR modelling

Quantum chemistry-augmented neural networks for reactivity prediction: Performance, generalizability, and explainability

Linking electron ionization mass spectra of organic chemicals to toxicity endpoints through machine learning and experimentation

Direct Prediction of Physicochemical Properties and Toxicities of Chemicals from Analytical Descriptors by GC–MS

QSAR-QSIIR-based prediction of bioconcentration factor using machine learning and preliminary application