Predicting the Sorption Capacity of Perfluoroalkyl and Polyfluoroalkyl Substances in Soils: Meta-Analysis and Machine...

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Soil Science

Background:

• Predicting perfluoroalkyl and polyfluoroalkyl substances (PFAS) soil sorption is crucial for environmental risk assessment.
• Traditional experimental methods for determining PFAS sorption are time-consuming and inefficient.

Purpose of the Study:

• To develop robust machine learning models for predicting PFAS soil sorption capacity.
• To identify key factors influencing PFAS sorption in soils.

Main Methods:

• Compiled a comprehensive dataset of 44 PFAS and 405 soils from 35 literature reports.
• Constructed machine learning models using LightGBM with RDKit or PaDEL descriptors.
• Utilized SHapley Additive exPlanation (SHAP) analysis to interpret model predictions.

Main Results:

• Machine learning models achieved high accuracy (R² up to 0.89) in predicting PFAS sorption.
• PFAS properties were the primary drivers of sorption, followed by environmental and soil properties.
• Identified critical pH (around 6) and soil organic carbon (SOC) thresholds influencing PFAS sorption.

Conclusions:

• Developed accurate and reliable computational models for PFAS soil sorption prediction.
• Provided insights into the mechanisms governing PFAS sorption, including electrostatic interactions and hydrophobic effects.
• The models and findings support informed environmental decision-making regarding PFAS contamination.