Assessing biodegradability potential of organic chemicals in aquatic and soil environment through...

Information on the biodegradation potential of organic chemicals in the ecosystem helps us analyze their persistence, bioaccumulation, and toxicity (PBT) behaviour. The environment is exposed to many chemicals from various sources, both intentionally and unintentionally. A preliminary assessment of chemical biodegradation prospects allows for early screening of their persistence and further analysis of their bioaccumulation potential and toxicity hazards. In this work, the authors demonstrate the application of the read-across structure-property relationship (RASPR) methodology and the development of classification-based read-across structure-biodegradability relationship (c-RASBR) models using chemicals collected according to the Organisation for Economic Co-operation and Development (OECD) Test Guidelines (TGs) 302 (Dataset 1) and 307 (Dataset 2). Data balancing was also performed (for Dataset 2) to equalize the number of compounds in each class (i.e., biodegradable or non-biodegradable) using various over-sampling techniques to eliminate bias in the model's prediction towards the majority class. Different machine learning algorithms were employed for model development, and cross-validation (CV) was used to check the robustness of the models. The final c-RASBR models developed with support vector classifier (SVC) appeared to outperform other machine learning models for both datasets, with accuracy ranging from 0.73 to 0.98 for the training set and 0.77 to 0.84 for the test set. SHapley Additive exPlanation (SHAP) analysis was also conducted to evaluate the importance of similarity and error-based RASPR features in the model prediction. The c-RASBR models with fewer features showed a significant improvement in prediction accuracy compared to the Qualitative Structure-Biodegradability Relationship (QSBR) models with a higher number of features. The developed models provide highly reliable predictions of chemicals' aerobic inherent biodegradability in water and biotransformation potential in soil. Additionally, a Python-based tool, "BiodegPred" has been developed to predict whether a given organic chemical is potentially biodegradable or non-biodegradable in water and soil. These models can be used as effective tools to assess the biodegradation potential of organic chemicals in water and soil environments, supporting regulatory decisions regarding chemical use and environmental exposure.