Integrated Machine-Learning Framework for Balanced Performance-Safety Design of Iron-Based Remediation Materials.

Area of Science:

• Environmental Science
• Materials Science
• Computational Chemistry

Background:

• Iron-based materials are crucial for environmental remediation.
• Optimizing these materials for contaminant removal can increase risks to organisms.
• A balance between performance and safety is needed.

Purpose of the Study:

• To develop an integrated machine learning framework for coassessing the performance and safety of iron-based remediation materials.
• To predict both remediation efficiency and joint toxicity.
• To create a practical prescreening tool for rational material application.

Main Methods:

• A machine learning framework was developed using AutoGluon.
• A quantitative structure-activity relationship (QSAR) model was trained on 1007 literature-derived cases.
• The model incorporated material descriptors, pollutant properties, and experimental conditions.

Main Results:

• The QSAR model achieved high accuracy in predicting remediation performance (0.88) and joint toxicity (0.90).
• Material heterogeneity was identified as a key factor for performance, while organismal attributes influenced toxicity.
• A novel weighting scheme and applicability domain filter were used for material ranking.

Conclusions:

• The developed framework enables quantitative coassessment of performance and safety for iron-based remediation materials.
• It supports the rational design and application of materials by prioritizing candidates.
• This approach addresses the trade-off between contaminant reactivity and ecological risk.