Enhanced QSAR Modeling for Predicting Organic Chemical Adsorption onto Soil

Abstract

The adsorption of organic chemicals onto soil plays a critical role in determining their environmental fate, mobility, and potential risks to ecosystems. Understanding and predicting soil adsorption behavior is essential for environmental risk assessment, pollution control, and regulatory decisions. In this study, we develop an enhanced Quantitative Structure-Activity Relationship (QSAR) model to predict the adsorption coefficients (Koc) of organic chemicals based on their physicochemical properties and molecular descriptors. The model integrates multiple machine learning approaches, including Random Forest, Neural Networks, and Hybrid QSAR techniques, to improve prediction accuracy. Key descriptors such as Log P, molecular weight, hydrogen bonding capacity, dipole moment, and soil properties (pH, organic matter content, and texture) were analyzed to determine their influence on adsorption behavior. The developed QSAR model was validated using experimental data from diverse chemical classes, including pesticides, pharmaceuticals, industrial chemicals, and surfactants. Comparative performance evaluation demonstrated that the Hybrid QSAR approach outperformed traditional methods, achieving a higher coefficient of determination (R²) across training, validation, and testing datasets. The results highlight the potential of advanced QSAR modeling as a reliable tool for predicting chemical-soil interactions, supporting environmental risk assessment and sustainable chemical management.