Machine Learning Optimizes Carbon Quantum Dot Sensor for Amoxicillin in Milk

A research team has developed a fluorescent sensing platform utilizing carbon quantum dots to detect amoxicillin in milk, combining experimental design and machine learning to optimize detection performance. The analytical system relies on measuring fluorescence quenching to identify the antibiotic in complex matrices.

To verify the physical and chemical properties of the nanomaterials, the authors characterized the carbon quantum dots using multiple analytical instruments. Fourier-transform infrared spectroscopy, transmission electron microscopy, scanning electron microscopy, energy-dispersive X-ray spectroscopy, and fluorescence spectroscopy were employed to examine the particles. These techniques confirmed the formation of the nanoscale carbon quantum dot structure and verified the presence of functional chemical groups on the surface of the dots.

Optimization of the sensing process was conducted through design of experiments, utilizing response surface methodology alongside a central composite design. The investigators systematically evaluated four experimental parameters: solution pH, reaction time, operating temperature, and sensor dosage. The optimization procedure was designed to maximize the fluorescence quenching response, measured as the ratio F₀/F, by evaluating how operational conditions interact during antibiotic detection.

Statistical evaluation of the experimental design demonstrated strong model adequacy, yielding a coefficient of determination of 0.9834, an adjusted coefficient of determination of 0.9680, and a predicted coefficient of determination of 0.9082. Analysis identified pH, reaction time, and sensor dosage as the most influential experimental variables affecting the fluorescence quenching response. In addition, diagnostic plots verified the underlying statistical assumptions of linearity, normality, and homoscedasticity, supporting the predictive reliability of the model.

To further validate and enhance predictive performance, machine learning models were applied to evaluate the multivariate sensing data. An ensemble learning approach using Random Forest achieved superior predictive accuracy compared to quadratic regression, demonstrating a coefficient of determination of approximately 0.96 alongside lower error metrics. The authors reported that the ensemble model effectively captured nonlinear multivariate interactions occurring among the evaluated parameters.

Calibration experiments established two linear response ranges for amoxicillin detection, spanning 10 to 300 micromolar and 300 to 650 micromolar. The corresponding detection limits were 41.47 micromolar and 101.78 micromolar, with quantification limits of 125.66 micromolar and 308.43 micromolar, respectively. Sensor reproducibility was confirmed by a blank precision relative standard deviation of 3.39 percent, while interference studies showed negligible effects from common ions and biomolecules, validating the selectivity of the platform in complex matrices.

The authors concluded that combining structural characterization, design of experiments, and machine learning models confirms the overall robustness of carbon quantum dots as an analytical platform for detecting amoxicillin in complex milk samples.

Subject of Research: Fluorescent detection of amoxicillin in milk using carbon quantum dots

Article Title: Fluorescent sensing of amoxicillin in milk using carbon quantum dots: optimization by design of experiments and machine learning

Article References: Adami, D., Barbod, D., Bijari, S., & Karami, C. (2026). Fluorescent sensing of amoxicillin in milk using carbon quantum dots: optimization by design of experiments and machine learning. BMC Chemistry. https://doi.org/10.1186/s13065-026-01927-6

Image Credits: AI Generated

DOI: 10.1186/s13065-026-01927-6

Keywords: amoxicillin, carbon quantum dots, fluorescent sensing, milk analysis, response surface methodology, machine learning, random forest, food safety testing, Fluorescent, sensing, milk, carbon