Machine Learning Accelerated Discovery of Antimicrobial Inorganic Nanomaterials.

Area of Science:

• Materials Science
• Nanotechnology
• Computational Chemistry

Background:

• Rising infectious diseases and antimicrobial resistance necessitate novel therapeutic agents.
• Antimicrobial inorganic nanomaterials offer potential but face challenges in efficient discovery.
• Current methods struggle with the complexity and vastness of nanomaterial data.

Purpose of the Study:

• To apply machine learning for the first time to discover novel antimicrobial inorganic nanomaterials.
• To identify key material features influencing antimicrobial activity.
• To develop a predictive model for efficient and cost-effective antimicrobial agent design.

Main Methods:

• Extracted data on over 2,000 antimicrobial nanomaterials from 8,000+ papers.
• Utilized unsupervised machine learning for data distribution analysis and feature exploration.
• Trained and evaluated multiple machine learning models, identifying five key features from 27 dimensions.
• Employed genetic programming-symbolic classification to generate a predictive mathematical formula.

Main Results:

• Identified five critical features out of 27 for antimicrobial activity prediction.
• Developed a predictive model with 0.83 accuracy for structure-activity relationships.
• Predicted 43 novel antimicrobial inorganic nanomaterials using the derived formula.
• Experimentally synthesized and validated the antibacterial properties of four predicted nanomaterials.

Conclusions:

• Machine learning provides a powerful, next-generation approach for designing antimicrobial inorganic nanomaterials.
• This study demonstrates a novel method for accelerating the discovery of effective and cost-efficient antimicrobial agents.
• The findings open new avenues for integrating machine learning into materials science discovery pipelines.