Predicting soil-water partition coefficients of PFAS using machine learning: Model development, interpretation, and validation.

Extreme gradient boosting (XGBoost) achieved high accuracy with R2 values of 0.83 and 0.86 in predicting PFAS soil-water partition coefficients using 10 descriptors.

• Why it matters: Accurate measurement of PFAS soil-water partition coefficients is costly and time-consuming, limiting environmental risk assessments and contaminated-site management.
• What they did: Researchers developed five machine-learning regression models with 2057 literature-derived data points, integrating descriptors like net charge, structural features, and soil properties, and validated the best model with independent soil samples.
• The result: The model's predictions were within one order of magnitude for all samples, offering an efficient, interpretable tool to support environmental risk assessment and contaminated-site management.