📄 chemistry

Selective Fluorescence Detection of Metronidazole in Traditional Dairy Products Using Green-Synthesized Carbon Quantum Dots Enhanced by Experimental Design and Machine Learning

This study presents a green-synthesized carbon quantum dot sensor, optimized via Design of Experiments and machine learning, for the sensitive and selective fluorescence detection of metronidazole in traditional dairy products.

Original authors: Sepideh Gharehyakheh, changiz karami, Sadaf Pirouzi

Published 2026-07-08

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Sepideh Gharehyakheh, changiz karami, Sadaf Pirouzi

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a cup of milk, and you're worried it might have a tiny, invisible amount of a medicine called Metronidazole in it. This medicine is used to fight infections in animals, but if it ends up in our food, it can be bad for our health. The problem is that finding this tiny amount is usually like trying to find a specific grain of sand on a beach using expensive, complicated machines that take a long time.

This research paper introduces a new, simpler, and "greener" way to find that medicine. Here is how they did it, explained in everyday terms:

1. The "Magic Glitter" (Green Synthesis)

Instead of using harsh chemicals to make a sensor, the researchers used a plant called Rosa Canina (wild rose).

The Analogy: Think of the rose plant as a kitchen ingredient. They took the plant, boiled it in water, and then cooked it under high pressure (like a pressure cooker) for a day.
The Result: This process turned the plant juice into tiny, glowing specks called Carbon Quantum Dots (CQDs). You can imagine these as microscopic, glowing glitter. They are special because they naturally shine brightly under a specific type of light.

2. The "Shy Light" (How Detection Works)

These glowing specks are the heart of the sensor.

The Analogy: Imagine the glowing glitter is a shy child who loves to shine. When you put them in a clean room (pure water), they shine very brightly.
The Interaction: However, if you introduce the "bad guy" (Metronidazole) into the room, the glitter gets distracted and stops shining as brightly. The more Metronidazole there is, the dimmer the light gets.
The Claim: By measuring exactly how much the light dims, the researchers can calculate exactly how much medicine is in the milk.

3. The "Recipe Master" (Experimental Design)

Making the sensor work perfectly isn't just about guessing. The researchers used a method called Design of Experiments (DoE).

The Analogy: Imagine you are baking a cake. You don't just throw ingredients in randomly; you need the right amount of heat, the right time in the oven, and the right pH (acidity).
The Process: They tested dozens of different combinations of temperature, time, and acidity to find the "perfect recipe" where the sensor reacts most clearly to the medicine. They used math to map out the best conditions, ensuring the sensor is as sensitive as possible.

4. The "Smart Predictor" (Machine Learning)

To make sure their math was perfect, they taught a computer to learn from their data using Machine Learning.

The Analogy: Imagine you have a student who is taking a test. You give them many practice questions (the data). Some students guess randomly, but the best student (the Random Forest model they chose) learns the patterns and gets almost every answer right.
The Result: This computer model helped them predict the amount of medicine with very high accuracy, confirming that their sensor works better than older, simpler methods.

5. The "Real-World Test" (Dairy Products)

Finally, they didn't just test this in a lab beaker; they tested it in real traditional dairy products (like raw milk).

The Challenge: Milk is a "messy" mix of fats, proteins, and sugars. Usually, these things get in the way and confuse sensors.
The Success: The researchers cleaned up the milk sample just enough and added their glowing sensor. It worked perfectly! It found the medicine even in the "messy" milk without getting confused by the other ingredients. They found that the sensor could detect the medicine accurately about 99% of the time.

Summary of the Claim

The paper claims they have built a fast, cheap, and eco-friendly flashlight made from wild roses. When this flashlight shines on milk, it dims if Metronidazole is present. They used smart math and computer learning to make sure this flashlight is super accurate, and they proved it works on real milk without needing expensive, complicated equipment.

What they didn't claim:

They did not claim this is a device you can buy at a store yet.
They did not claim this cures diseases or changes how doctors treat patients.
They did not claim it works on every type of food, only the dairy products they tested.

Their main goal was simply to prove that this new "rose-based glowing sensor" is a reliable way to check if milk is safe from this specific antibiotic residue.

Technical Summary: Selective Fluorescence Detection of Metronidazole in Traditional Dairy Products

Problem Statement
The excessive and improper use of antibiotics in livestock farming has led to the accumulation of residues, such as metronidazole (MNZ), in food supplies like milk and meat. These residues pose significant public health risks, including allergic reactions, disruption of gut microbiota, and the emergence of antibiotic-resistant bacterial strains. While methods like HPLC and UV-Vis spectroscopy exist for detecting MNZ, they are often limited by high costs, complex sample preparation, and matrix interference. There is a critical need for rapid, sensitive, and cost-effective analytical methods to monitor antibiotic residues in traditional dairy products to ensure food safety.

Methodology
This study developed a green fluorescence sensor based on Carbon Quantum Dots (CQDs) synthesized from Rosa canina (dog rose) via a hydrothermal method. The research integrated experimental design and machine learning to optimize and validate the sensor's performance.

Synthesis and Characterization: CQDs were synthesized by heating Rosa canina extract in a Teflon autoclave at 200°C for 24 hours. The resulting nanomaterials were characterized using Fluorescence Spectroscopy, FTIR, SEM, EDX, and elemental mapping. The CQDs were found to be nearly spherical with an average diameter of ~~2 nm, rich in carbon (~~50.9%) and oxygen (~26%), and functionalized with carboxyl and hydroxyl groups.
Experimental Design (DoE): A Central Composite Design (CCD) within Response Surface Methodology (RSM) was employed to optimize sensor parameters. Three independent variables were tested: pH (4.0–10.0), reaction time (0.5–5.0 min), and temperature (34–60°C). A quadratic model was used to predict the fluorescence quenching response ( $F_0/F$ ).
Machine Learning Integration: To enhance predictive accuracy, six supervised machine learning regression models (SVR, Random Forest, ANN, KNN, GBR, DTR) were trained using Python and Scikit-learn. The models were evaluated based on $R^2$ , RMSE, and MAE to determine the best algorithm for predicting MNZ concentrations.
Sample Preparation: Raw milk samples were processed by precipitating proteins with trichloroacetic acid, neutralizing with NaOH, and centrifuging to obtain a clear supernatant for analysis.

Key Results

Optimization: The quadratic model demonstrated a high statistical fit ( $R^2 = 0.9958$ , Adjusted $R^2 = 0.9919$ ). ANOVA confirmed that pH and reaction time were the most significant factors influencing the fluorescence response, while temperature had a negligible effect under the tested conditions.
Machine Learning Performance: Among the tested algorithms, the Random Forest model outperformed others, achieving the highest $R^2$ (~0.96) and the lowest error rates (RMSE and MAE). This indicated its superior ability to model the non-linear relationships in the sensor data compared to Linear Regression or SVR.
Analytical Performance: Under optimized conditions, the sensor exhibited a linear detection range of 5.0–400.0 µM for metronidazole. The Limit of Detection (LOD) was calculated at 2.24 µM, and the Limit of Quantification (LOQ) at 7.39 µM. The method showed excellent repeatability (RSD = 3.38%) and reproducibility (RSD = 3.47%).
Selectivity and Interference: The sensor displayed high selectivity for metronidazole. Common co-existing substances in dairy, including metal ions (Ca²⁺, K⁺, Na⁺), proteins (casein), sugars (lactose), and other antibiotics (e.g., tetracycline, amoxicillin), did not significantly interfere with the fluorescence signal.
Real-World Application: When applied to spiked traditional dairy samples, the sensor achieved recovery rates between 100.5% and 101.7% with low relative standard deviations, confirming its reliability in complex food matrices.

Significance and Claims
The authors claim that this work presents a sustainable, "green" approach to antibiotic monitoring by utilizing agricultural waste (Rosa canina) to create high-performance sensors. The study highlights the successful integration of Design of Experiments (DoE) and Machine Learning to refine sensor performance and predictive modeling, moving beyond traditional trial-and-error optimization.

The paper asserts that the developed CQD-based sensor offers a viable alternative to expensive and complex chromatographic methods. It provides a sensitive, selective, and rapid tool for detecting metronidazole in traditional dairy products, contributing to food safety surveillance and the mitigation of antibiotic resistance risks. The authors position the method as particularly suitable for routine food safety assessments in artisanal and traditional dairy contexts.