Machine Learning Tools to Assist the Synthesis of Antibacterial Carbon Dots

Area of Science:

• Nanotechnology
• Materials Science
• Computational Biology

Background:

• Multidrug-resistant bacteria (MRB) pose a significant global health threat, necessitating novel antimicrobial strategies.
• Nanoparticles, particularly carbon dots (CDs), offer promising alternatives to antibiotics due to their unique antimicrobial mechanisms.
• CDs combat MRB by disrupting bacterial cell walls, inhibiting DNA/enzymes, inducing hyperthermia, or generating reactive oxygen species.

Purpose of the Study:

• To investigate the relationship between the physicochemical properties of carbon dots (CDs) and their antimicrobial efficacy against multidrug-resistant bacteria (MRB).
• To utilize machine learning (ML) tools for predicting and decoding this structure-activity relationship.
• To accelerate the development and clinical translation of high-performance antimicrobial nanoparticles.

Main Methods:

• A dataset of 121 carbon dot (CD) samples was compiled, including synthetic conditions and intrinsic properties, with Minimum Inhibitory Concentration (MIC) as the output.
• Four classification algorithms (KNN, SVM, RF, XGBoost) were trained and validated.
• Ensemble learning methods demonstrated superior performance, and ε-poly(L-lysine) CDs (PL-CDs) were synthesized to validate the ML models' practical application.

Main Results:

• Machine learning models, particularly ensemble methods, accurately predicted the antimicrobial activity of carbon dots (CDs).
• The study identified key physicochemical features influencing CD antimicrobial capacity.
• Synthesized PL-CDs confirmed the predictive power and practical utility of the developed ML models.

Conclusions:

• Machine learning-based high-throughput theoretical calculations can effectively predict the antibacterial effects of carbon dots (CDs).
• This approach accelerates the discovery of novel nanoparticles for combating multidrug-resistant bacteria (MRB).
• The findings support the potential clinical translation of ML-guided nanoparticle development.