Machine Learning-Assisted Prediction of Photothermal Metal-Phenolic Networks.

Area of Science:

• Materials Science
• Biomedical Engineering
• Computational Chemistry

Background:

• Photothermal therapy (PTT) relies on effective photothermal agents (PTAs).
• Metal-phenolic networks (MPNs) are promising PTAs due to their properties.
• Screening MPNs for optimal photothermal performance is challenging due to vast chemical space.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting the photothermal performance of MPNs.
• To efficiently identify high-performance MPNs for PTT and other biomedical applications.

Main Methods:

• Constructed a database of photothermal properties for 80 modular MPNs.
• Employed feature engineering and model training to optimize ML prediction.
• Utilized an extreme gradient boosting (XGBoost) model for material screening.

Main Results:

• Identified 1,654 high photothermal MPNs from a virtual library of 44,438 candidates.
• Achieved a 70% success rate in experimental validation of predicted high-performance MPNs.
• Discovered novel MPNs with significant potential for photothermal antibacterial applications.

Conclusions:

• An innovative ML-driven approach enables efficient screening of MPN materials for PTT.
• This methodology provides a robust foundation for designing advanced PTAs.
• The study highlights the potential of ML in accelerating materials discovery for biomedical applications.