Method for rapidly detecting and identifying food-borne pathogens by using two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array based on machine learning

Abstract

A method for rapidly detecting and identifying food-borne pathogens by using a two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array based on machine learning comprises the following steps: synthesizing three nitrogen-doped graphene quantum dots through a hydrothermal method, determining the optimal excitation wavelength, concentration, reaction temperature and pH value of the nitrogen-doped graphene quantum dots, and further determining the fluorescence intensity of the nitrogen-doped graphene quantum dots under the optimal condition; constructing a fluorescence sensor array, mixing the three graphene quantum dots with target heavy metal ions, incubating, then measuring fluorescence intensity to obtain fluorescence response data, and normalizing the data to obtain the fluorescence response data. And analyzing by using machine learning methods of a decision tree, a support vector machine, K nearest neighbor, linear discriminant analysis, naive Bayes and a neural network. The prepared fluorescence sensor array can rapidly identify heavy metal ions, is excellent in accuracy, sensitivity and stability, provides an economical and efficient solution for heavy metal ion detection, and has remarkable application potential in environmental monitoring.

Description

Technical Field

[0001] This invention belongs to the field of microbial detection technology, specifically relating to a method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array. Background Technology

[0002] Foodborne pathogens are a key cause of food safety problems and resource waste, placing significant pressure on public health systems and the socio-economic landscape. The core challenge in controlling contamination sources during food distribution lies in the rapid identification and accurate classification of microbial species.

[0003] Current routine methods for microbial diagnosis still rely on culture and isolation techniques, with detection cycles lasting 1-3 days. While polymerase chain reaction (PCR) has improved detection speed and sensitivity, it suffers from insufficient result stability and the risk of misinterpretation. Surface-enhanced Raman scattering (SERS), although capable of microbial identification, is limited by high instrument costs and the need for specialized databases, restricting its practical application. Therefore, developing foodborne pathogen screening technologies with rapid response, high accuracy, and low cost has significant scientific value and application prospects.

[0004] Currently, various microbial detection technologies based on molecular recognition principles have been successfully established. While these technologies exhibit good reliability and stability, they are typically only applicable to the identification of single or specific bacterial species, and struggle to effectively handle complex samples containing multiple coexisting microorganisms. Against this backdrop, sensor arrays are considered a breakthrough solution for achieving precise analysis of mixtures. Current research in this field primarily focuses on two types of strategies: electrochemical detection and fluorescence detection. Among these, fluorescence sensors, leveraging the superior photophysical properties of luminescent molecules, demonstrate a significant advantage in sensitivity compared to electrochemical methods, making them more suitable for the precise analysis of low-concentration microorganisms.

[0005] Graphene quantum dots (GQDs) possess a unique zero-dimensional (0D) structure rich in edge active sites, granting them multifunctional properties such as electrocatalytic activity, chemiluminescence, and photoluminescence (PL). GQDs not only exhibit excellent photostability, efficient electron transfer capabilities, and good biosafety, but also show potential application value in the field of microbial detection. Doping with heteroatoms such as sulfur and nitrogen significantly enhances their luminescence performance and quantum yield. However, in practical applications, these nanomaterials are prone to aggregation due to their high surface energy, leading to a decrease in catalytic activity. Therefore, it is necessary to find suitable materials to provide anchoring sites to ensure their stability and effectiveness in microbial detection.

[0006] MXenes, as emerging two-dimensional transition metal carbide / nitride materials, possess atomic-level thickness and a graphene-like layered structure. Their surface chemistry can be precisely controlled, exhibiting excellent electrical conductivity, good hydrophilicity, and outstanding multifunctional properties. Among them, Ti3C2T... X MXene nanosheets have become the most studied material in this system due to their unique synergistic mechanism of charge transfer and catalytic activity. Compared with traditional carbon-based nanomaterials, Ti3C2T... X MXene surfaces are rich in controllable chemically active sites. These functional groups can not only effectively anchor nanoparticles but also enhance the properties of composite materials through interfacial interactions. Studies by Geng et al. have shown that in monolayer Ti3C2T... X Loading carbon quantum dots onto MXene can significantly improve the efficiency of reactive oxygen species (ROS) generation; the Ti3C2T developed by Yu's team X The MXene / gold nanoparticle composite system enables rapid detection of microorganisms and efficient photothermal sterilization. Although the two-dimensional / zero-dimensional (2D / 0D) heterojunction formed by MXene and graphene quantum dots exhibits a significant interfacial synergistic enhancement effect, its application in fluorescent pathogen sensing is still insufficient.

[0007] To achieve accurate detection of target analytes, standardized preprocessing of sensor data and the establishment of reliable analytical models are crucial. Recent research indicates that machine learning algorithms (especially Naive Bayes and K-nearest neighbors) demonstrate significant advantages in the field of microbial identification. These intelligent analytical methods not only significantly improve the accuracy of pathogen identification in complex samples but also reduce operational errors and analytical costs through automated processes, providing efficient technical support for food microbial identification. Summary of the Invention

[0008] The purpose of this invention is to provide a method for the rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array.

[0009] To achieve the above and other related objectives, the technical solution provided by this invention is: a method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array, characterized by comprising the following steps:

[0010] S1: Nitrogen-sulfur co-doped graphene quantum dots were synthesized using a one-step hydrothermal method;

[0011] S2: Preparation of monolayer Ti3C2T using a hydrochloric acid / lithium fluoride etching system X MXene nanosheets;

[0012] S3: Graphene quantum dots co-doped with nitrogen and sulfur and monolayer Ti3C2T X Using MXene nanosheets as raw materials, heterojunction composite materials were constructed through in-situ hydrothermal assembly;

[0013] S4: Measure the fluorescence emission spectrum of the heterojunction composite material, optimize its excitation wavelength, reaction concentration, and incubation time with bacteria, and measure the fluorescence intensity of the heterojunction composite material under optimal conditions;

[0014] S5: Single colonies of pathogenic bacteria cultured on LB agar plates were washed with PBS and resuspended, and the absorbance at 600 nm was measured using a spectrophotometer; based on the constructed CFU-OD... 600 Standard curves were obtained by diluting the bacterial suspension to a series of target concentrations using PBS buffer.

[0015] S6: Construct a fluorescence sensor array, mix the heterojunction composite material solution with bacterial solutions of different concentrations at a set volume ratio, incubate, and measure the fluorescence intensity at the optimized excitation wavelength to optimize signal detection. Use a fluorescence spectrophotometer to record the fluorescence changes before and after the interaction between the fluorescent material and bacteria.

[0016] S7: Fluorescence response data were normalized, and a fluorescence feature recognition model for foodborne pathogens was constructed based on five machine learning algorithms. The dataset was divided into training and test sets in an 8:2 ratio using stratified random sampling, and five-fold cross-validation was used to evaluate model performance. Principal component analysis was introduced to process the raw fluorescence data to improve feature extraction efficiency. The machine learning process was implemented using Matlab, and data visualization analysis was performed using Origin 2024 and Matlab. Model performance was comprehensively evaluated using four metrics: precision, recall, F1 score, and accuracy.

[0017] The preferred technical solution is as follows: In step S1, the method for synthesizing nitrogen-sulfur co-doped graphene quantum dots includes:

[0018] (1) Synthesis of graphene quantum dot 1

[0019] Add anhydrous citric acid and thiourea to water and stir. Then transfer the mixture to an autoclave with a Teflon liner and heat at 150-170°C for 5-8 hours.

[0020] (2) Synthesis of graphene quantum dot 2

[0021] Anhydrous citric acid and L-cysteine ​​are added to water and stirred. Then the mixture is heated at 170-190°C for 8-12 hours in a reaction vessel with a Teflon liner.

[0022] (3) Synthesis of graphene quantum dot 3

[0023] Add anhydrous citric acid, sulfur powder, and ammonia to water, stir, and then heat at 170-190°C for 10-14 hours in a Teflon-lined autoclave.

[0024] The preferred technical solution is: in step S2, a single layer of Ti3C2T X The methods for synthesizing MXene nanosheets include:

[0025] The aluminum layer in Ti3AlC2 was selectively etched using LiF / HCl solution. After acid washing and purification, multiple water washings and centrifugation were performed to control the stratification. When the supernatant became turbid and separated into layers, high-speed centrifugation was used to enrich the precipitate. Continuous washing until neutral was performed to obtain multilayer Ti3C2. X Monolayer Ti3C2T was finally prepared by ultrasonic exfoliation, liquid phase separation and freeze drying under nitrogen protection. X MXene nanosheets.

[0026] The preferred technical solution is as follows: Add Ti3AlC2 powder to LiF / HCl solution and stir continuously at 40-50°C for 20-30 hours; centrifuge the resulting reaction solution at 3000-4000 r / min for 3-8 minutes and collect the bottom precipitate; redissolve the precipitate with hydrochloric acid, wash with deionized water, repeat 2-3 times; then wash with deionized water and centrifuge at 3000-4000 r / min for 3-8 minutes; continue to centrifuge and wash with deionized water 3-5 times until the supernatant is turbid and separated into layers; use high-speed centrifugation at 9000-10000 r / min to collect the bottom precipitate; repeat the deionized water washing-centrifugation until the pH of the supernatant is ≥ 6 to obtain the precipitate; disperse the precipitate in deionized water and sonicate in an ice-water bath under nitrogen protection for 1.5-2.5 hours; centrifuge the ultrasonic dispersion at 3000-4000 r / min for 20-40 minutes, collect the upper liquid phase, and finally freeze-dry to obtain a monolayer Ti3C2T. X MXene nanosheets.

[0027] The preferred technical solution is: in step S3, a single layer of Ti3C2T is applied... X MXene nanosheets and nitrogen-sulfur co-doped graphene quantum dots were added to deionized water and stirred to form a uniform suspension. The suspension was then transferred to a reactor for hydrothermal reaction. The product was centrifuged, washed multiple times with deionized water, and finally freeze-dried to obtain a heterojunction composite material.

[0028] The preferred technical solution is as follows: In step S4, the optimal excitation wavelengths of the three heterojunction composite materials used in the test are 350 nm, 360 nm and 360 nm, respectively; the optimal reaction concentration is 0.3 mg / mL; and the optimal incubation time with bacteria is 3 minutes.

[0029] The preferred technical solution is as follows: In step S5, the pathogens include Salmonella Typhimurium, Shigella, Escherichia coli, Staphylococcus aureus, Pseudomonas aeruginosa, and Listeria monocytogenes.

[0030] The preferred technical solution is as follows: In step S7, the formula for normalizing the fluorescence response data is: (F0-F) / F0; where F0 is the fluorescence intensity before the addition of pathogens, and F is the fluorescence intensity after the addition of pathogens; the five machine learning algorithms are: support vector machine, decision tree, K nearest neighbor, Naive Bayes and linear discriminant analysis.

[0031] The preferred technical solution is as follows: Three heterojunction composite materials are selected, and experiments are conducted against six pathogens. Each experiment is repeated eighteen times to establish a training dataset. The calculation method involves multiplying the three heterojunction composite materials, the six pathogens, and the eighteen repeated experiments together.

[0032] By employing the above-described technical solution, the advantages of this invention compared to the prior art are:

[0033] 1. The sensor array of the present invention has the advantage of low cost. Its preparation method is simple and inexpensive, which is conducive to its widespread application in the food industry and reduces detection costs.

[0034] 2. The sensor array of the present invention has high sensitivity and can detect low concentrations of pathogens, with a detection limit as low as 1.0×10³ CFU / mL, which can effectively identify small amounts of pathogenic microorganisms in the sample.

[0035] 3. The sensor array of the present invention has strong anti-interference ability and can maintain high sensitivity in complex matrices such as pork, milk and tap water, without being affected by matrix components, thus ensuring accurate and reliable detection results.

[0036] 4. The sensor array of the present invention has good stability, exhibits good stability under different environmental conditions, and the results of multiple measurements are consistent, which can ensure long-term stable detection in practical applications.

[0037] 5. The sensor array of the present invention has a fast response speed, which can complete the detection and give the results in a short time. The incubation time is only 3 minutes, which is suitable for the rapid monitoring needs of the food industry.

[0038] 6. The sensor array of this invention has a wide quantitative analysis range, from 1.0×10³ to 1.0×10³. 7 The method allows for quantitative analysis of six bacteria within a concentration range of CFU / mL, covering the common concentration range of pathogens in food and offering broad applicability. Attached Figure Description

[0039] Figure 1TEM image (A) of GQDs1 in Example 2; histogram of particle size distribution of GQDs1 (B); multilayer Ti3C2T X TEM characterization of the layered structure of MXene (C); after peeling, the monolayer Ti3C2T X TEM image of MXene (D); TEM image of MXene@GQDs1 composite (E); HRTEM image of MXene@GQDs1 (F); XPS spectra of the three MXene@GQDs composites (GI); Fourier transform infrared (FT-IR) spectra of the three MXene@GQDs composites (J); Zeta potential distribution of the three MXene@GQDs (K); 30-day fluorescence storage stability of the three MXene@GQDs composites (L).

[0040] Figure 2 Optimization of the reaction conditions in Example 3: excitation wavelength (AC), concentration (D), and incubation time (E).

[0041] Figure 3 The image shows the fluorescence spectra of the three MXene@GQDs from Example 4 after reacting with different concentrations of six single bacteria (Salmonella Typhimurium, Shigella, Escherichia coli, Staphylococcus aureus, Pseudomonas aeruginosa, and Listeria monocytogenes). The concentration range of the six bacteria was 1.0 × 10⁻⁶. 3 CFU / mL - 1.0×10 7 CFU / mL (excitation wavelength: 360 nm).

[0042] Figure 4 The images show the fluorescence response spectra of the fluorescence sensor array for six types of bacteria in Example 5 (A); a performance comparison of five machine learning algorithms (B); a confusion matrix based on the Naive Bayes algorithm (C); and a parallel coordinate plot based on the Naive Bayes algorithm (D). All bacteria are within a range of 1.0 × 10⁻⁶. 3 CFU / mL.

[0043] Figure 5 The images shown are: (A) a three-dimensional principal component space projection map of the six bacterial samples classified using machine learning in Example 5; (B) a two-dimensional PCA score map of the six bacterial samples; (C) a fluorescence response heatmap of the six bacteria based on three MXene / GQDs; and (D) a hierarchical clustering ring dendrite diagram of the six bacterial groups. All bacteria are within a 1.0 × 10⁻⁶ range. 3 CFU / mL.

[0044] Figure 6 shows the results of E. coli concentration gradient identification and classification based on a fluorescence sensor array in Example 6 (1.0 × 10³ to 1.0 × 10³). 7CFU / mL: Includes a two-dimensional principal component analysis (PCA) score plot (A), a linear correlation plot of principal component factor 1 and E. coli concentration (B), a fluorescence response circular clustering heatmap (C), and a fluorescence response circular stacked bar chart (D).

[0045] Figure 7 This refers to the identification and classification of Staphylococcus aureus concentration gradients based on a fluorescence sensor array in Example 6 (1.0 × 10³ - 1.0 × 10³). 7 CFU / mL): Two-dimensional PCA score plot (A); Linear correlation plot of principal component factor 1 and Staphylococcus aureus concentration (B); Fluorescence response circular clustering heatmap (C); Fluorescence response circular stacked bar chart (D).

[0046] Figure 8 This refers to the identification and classification of Salmonella typhimurium concentration gradients based on a fluorescence sensor array in Example 6 (1.0 × 10³ - 1.0 × 10⁻⁶). 7 CFU / mL): Two-dimensional PCA score plot (A); Linear correlation plot of principal component factor 1 and Salmonella typhimurium concentration (B); Fluorescence response circular clustering heatmap (C); Fluorescence response circular stacked bar chart (D).

[0047] Figure 9 This refers to the identification and classification of Shigella based on a fluorescence sensor array (1.0 × 10³ - 1.0 × 10³) in Example 6. 7 CFU / mL): Two-dimensional PCA score plot (A); Linear correlation plot of principal component factor 1 and Shigella concentration (B); Fluorescence response circular clustering heatmap (C); Fluorescence response circular stacked bar chart (D).

[0048] Figure 10 This is for the identification and classification of *Pseudomonas aeruginosa* concentration gradients based on a fluorescence sensor array in Example 6 (1.0 × 10³ - 1.0 × 10³). 7 CFU / mL): Two-dimensional PCA score plot (A); Linear correlation plot of principal component factor 1 and Pseudomonas aeruginosa concentration (B); Fluorescence response circular clustering heatmap (C); Fluorescence response circular stacked bar chart (D).

[0049] Figure 11 This refers to the identification and classification of Listeria monocytogenes concentration gradients (1.0 × 10³ - 1.0 × 10³) based on a fluorescence sensor array in Example 6. 7 CFU / mL): Two-dimensional PCA score plot (A); Linear correlation plot of principal component factor 1 and Listeria monocytogenes concentration (B); Fluorescence response ring clustering heatmap (C); Fluorescence response ring stacked bar chart (D).

[0050] Figure 12The images shown in Example 7 are: PCA score diagram (A), corresponding confusion matrix diagram (B), and parallel coordinate diagram (C) for identifying two mixed bacteria, *Escherichia coli* and *Shigella*, using fluorescence array method; and PCA score diagram (D), corresponding confusion matrix diagram (E), and parallel coordinate diagram (F) for three mixed bacteria, *Escherichia coli*, *Shigella*, and *Pseudomonas aeruginosa*. All bacterial concentrations were 1.0 × 10⁻⁶. 3 CFU / mL.

[0051] Figure 13 The fluorescence response spectra of the fluorescence sensor array in Example 8 to 13 typical interfering substances (including Na⁺, K⁺, Ca²⁺, F⁻, NO⁻, SO₄²⁻, Cl⁻, HPO₄²⁻, glucose, fructose, L-cysteine, arginine, and lysine) (A); the correlation between fluorescence signal intensity and the number of measurement repetitions (B); and the response characteristics of fluorescence signal to temperature changes (C).

[0052] Figure 14 The images shown in Example 9 are PCA score charts for a single bacterial species in a blind sample from PBS using the fluorescence array method: (A); (B); and (C); (C); (B) for a binary mixture of *E. coli* and *Shigella*; and (C) for a ternary mixture of *E. coli*, *Shigella*, and *Pseudomonas aeruginosa*. The bacterial concentration for all samples was 1.0 × 10⁻⁶. 3 CFU / m.

[0053] Figure 15 The PCA score charts for six foodborne pathogens in the actual samples from Example 10 are as follows: tap water (A); milk (B); pork (C). The bacterial concentration for all samples was 1.0 × 10⁻⁶. 3 CFU / m.

[0054] Figure 16 A flowchart illustrating how machine learning algorithms are used to process data.

[0055] Figure 17 The flowcharts for the synthesis of the composite material in Example 1 (A) and the flowcharts for the use of the invention (B) are shown. Detailed Implementation

[0056] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in these embodiments.

[0057] Please see Figure 1-17It should be noted that the structures, proportions, sizes, etc., illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size are not permitted. The following embodiments are provided to better understand the invention, but are not intended to limit it. Unless otherwise specified, the experimental materials used in the following embodiments were purchased from conventional consumable and biochemical reagent stores.

[0058] Example 1: Preparation of three types of MXene@GQDs

[0059] 1. Preparation of three graphene quantum dots (GQDs1, GQDs2, and GQDs3): For the synthesis of GQDs1, 0.84 g of anhydrous citric acid and 1.24 g of thiourea were dissolved in 20 mL of ultrapure water and stirred for 5 min using a magnetic stirrer. The mixture was then transferred to a Teflon-lined autoclave and heated at 160 °C for 6 h. For GQDs2, 4.2 g of anhydrous citric acid and 2 g of L-cysteine ​​were dissolved in 20 mL of ultrapure water and stirred for 5 min. The mixture was then heated at 180 °C for 10 h in a Teflon-lined autoclave. For GQDs3, 0.42 g of anhydrous citric acid, 0.3 g of sulfur powder, and 7.5 mL of 28% ammonia solution were dissolved in 7.5 mL of ultrapure water and stirred for 5 min. The mixture was then heated at 180 °C for 12 h in a Teflon-lined autoclave. After the reaction was complete, the products were cooled to room temperature in a fume hood. Each product was double-filtered through a 0.22 μm polyethersulfone membrane to remove large particulate impurities. The filtrate was transferred to a dialysis bag with a molecular weight cutoff of 3500 Da and dialyzed at room temperature for 48 hours, with ultrapure water replaced every 6 hours. The final solution was freeze-dried for 24 hours to obtain three graphene quantum dot powders (GQDs1, GQDs2, and GQDs3).

[0060] 2. Single-layer Ti3C2T XSynthesis of MXene: Initially, 1.6 g of LiF was added to 20 mL of 9 M HCl solution and stirred in an ice-water bath for 15 minutes to obtain a LiF / HCl solution. Then, 1 g of Ti3AlC2 powder was slowly added, and the mixture was stirred continuously at 45°C for 24 hours. The reaction solution was centrifuged at 3500 r / min for 5 minutes, and the bottom precipitate was collected. The precipitate was redissolved in 30 mL of 1 M HCl, washed with deionized water, and this process was repeated 2-3 times. Then, it was washed with 35 mL of deionized water and centrifuged at 3500 r / min for 5 minutes. This was repeated 3-5 times with deionized water until the supernatant became turbid and separated into layers (indicating that MXene had begun to separate). Due to the low liquid concentration, a high-speed centrifugation at 10,000 r / min was used to collect the bottom precipitate. The washing and centrifugation process was repeated until the pH of the supernatant was ≥ 6, yielding a viscous, multilayered Ti3C2T solution. X The precipitate was dispersed in 60 mL of deionized water and sonicated in an ice-water bath for 2 hours (240 W) under nitrogen protection to prevent oxidation. The ultrasonic dispersion was centrifuged at 3500 r / min for 30 minutes, and the supernatant was collected. Finally, it was freeze-dried to obtain a monolayer MXene powder.

[0061] 3. Synthesis of MXene@GQDs Composite Material: Two-dimensional / zero-dimensional MXene@GQDs heterostructures were prepared using a hydrothermal method. Monolayer MXene and GQDs were mixed at a 1:1 mass ratio (total mass 100 mg), and 30 mL of deionized water was added. The mixture was vigorously stirred at room temperature for 30 minutes to form a homogeneous suspension. The suspension was then transferred to a 50 mL polytetrafluoroethylene-lined reactor and hydrothermally reacted at 100°C for 4 hours. After the reaction, the product was centrifuged, washed repeatedly with deionized water, and freeze-dried for 24 hours to obtain MXene@GQDs powder.

[0062] Example 2: Characterization of MXene@GQDs

[0063] This study comprehensively characterized the morphology, structure, and surface charge properties of MXene@GQDs using transmission electron microscopy (TEM), Fourier transform infrared spectroscopy (FT-IR), X-ray photoelectron spectroscopy (XPS), and Zeta potential analysis. Simultaneously, the fluorescence stability of MXene@GQDs over 30 days was measured using a fluorescence spectrophotometer. The results are as follows: TEM analysis showed that MXene@GQDs exhibited a dispersed spherical morphology in aqueous solution; demonstrating multilayered MXene (Ti3C2T) X TEM images of the nanosheets show them as opaque, highly detailed sheet-like structures that contrast sharply with the background, reflecting their substantial thickness; while under TEM, the monolayer Ti3C2T... XThe nanosheets exhibit ultrathin and near-transparent properties; XPS analysis confirmed that three types of graphene quantum dots have been successfully loaded onto Ti3C2T. X Above; FT-IR spectroscopy analysis showed that the surfaces of the three MXene@GQDs were rich in functional groups; Zeta potential tests indicated that all three MXene@GQDs carried a negative charge. Furthermore, all three MXene@GQDs composites exhibited good fluorescence stability during a 30-day storage period.

[0064] Example 3: Optimization of detection conditions for fluorescence sensor array

[0065] This implementation optimized the detection parameters of three MXene@GQDs, determining their optimal excitation wavelength, experimental concentration, and reaction incubation time. The results are as follows: Figure 2 As shown, the optimal excitation wavelengths for MXene@GQDs1, MXene@GQDs2, and MXene@GQDs3 were 350 nm, 360 nm, and 360 nm, respectively. Considering that the fluorescence intensity of MXene@GQDs1 showed little change at excitation wavelengths of 350 and 360 nm, 360 nm was ultimately chosen as the universal excitation wavelength to standardize the experimental conditions. At concentrations of 0.1–0.5 mg / mL, the fluorescence intensity of the three materials tended to stabilize, possibly because the heterojunction structure of MXene@GQD effectively suppressed GQD aggregation and the repackaging effect of MXene. The optimal experimental concentration was determined to be 0.3 mg / mL. During the initial incubation phase (1–3 minutes), the gradual interaction between the materials and bacteria led to a significant increase in the fluorescence quenching rate. Fluorescence quenching reached equilibrium after 3 minutes; therefore, 3 minutes was determined to be the optimal reaction time.

[0066] Example 4: Fluorescence response of three MXene@GQDs to bacteria

[0067] This study investigated *Salmonella typhimurium*, *Shigella*, *Escherichia coli*, *Staphylococcus aureus*, *Pseudomonas aeruginosa*, and *Listeria monocytogenes*. First, single colonies of the six bacteria grown on Luria-Bertani (LB) agar plates were washed twice with phosphate-buffered saline (PBS; 3 mL). Subsequently, the resuspended bacteria were quantitatively analyzed using a spectrophotometer, measuring their optical density (OD600) at 600 nm. Based on a pre-established standard curve of colony forming units (CFU) versus OD600, the bacterial suspension was diluted with PBS to a series of specific concentrations (1.0 × 10³ CFU / mL, ... 4 CFU / mL, 1.0×10 5 CFU / mL, 1.0×10 6CFU / mL, 1.0×10 7 (CFU / mL) was prepared for subsequent experiments. Next, bacterial suspensions of different concentrations were incubated with three types of MXene@GQDs for 3 minutes, and fluorescence changes were detected using a fluorescence spectrophotometer at an excitation wavelength of 360 nm. The experimental results are shown in Figure 3. The three types of MXene@GQDs exhibited distinctly different response characteristics to the six bacteria, strongly demonstrating their excellent selective recognition ability.

[0068] Example 5: Performance test of fluorescence sensor array in distinguishing six types of bacteria

[0069] This study selected *Salmonella typhimurium*, *Shigella*, *Escherichia coli*, *Staphylococcus aureus*, *Pseudomonas aeruginosa*, and *Listeria monocytogenes* as research subjects. Single colonies of pathogens cultured on LB agar plates were washed with PBS and resuspended, and the absorbance (OD value) at 600 nm was measured using a spectrophotometer. The study also utilized a pre-constructed CFU-OD model. 600 Standard curve was obtained by diluting the bacterial suspension to 1.0 × 10⁻⁶ using PBS buffer. 3 CFU / mL was prepared for use. Next, the six bacteria were mixed with three types of MXene@GQDs and incubated for 3 min. Fluorescence changes were detected using a fluorescence spectrophotometer at an excitation wavelength of 360 nm. The fluorescence response data were normalized, and a fluorescence feature recognition model for foodborne pathogens was constructed based on five machine learning algorithms. The dataset was divided into training and test sets in an 8:2 ratio using stratified random sampling, and five-fold cross-validation was used to evaluate the model performance. Principal component analysis (PCA) was introduced to process the raw fluorescence data to improve feature extraction efficiency.

[0070] As shown in Figure 4, the three MXene@GQDs sensing units exhibit specific fluorescence response patterns to the target bacteria, enabling efficient differentiation. All six machine learning algorithms achieved 100% performance; the true positive rate (TPR) and positive predicted value (PPV) in the confusion matrix both reached 100%, demonstrating extremely high recognition accuracy; the parallel coordinate plot clearly reveals the quantitative relationship between the three key recognition features. As shown in Figure 5, in the three-dimensional feature space, the six bacteria form mutually independent clustering regions with extremely significant spatial separation; the scoring plot based on principal component analysis (PCA) further confirms the significant differences between bacteria; the annular grouping heatmap clearly demonstrates the ability of the three MXene@GQDs to distinguish different bacterial species; the dendrogram based on hierarchical clustering analysis presents six different clusters, each corresponding to a unique bacterial type. In summary, these results fully demonstrate the high reliability of the developed sensor array in efficiently differentiating six bacteria.

[0071] Example 6: Performance test of fluorescence sensor array in distinguishing single bacteria at different concentrations

[0072] This study investigated *Salmonella typhimurium*, *Shigella*, *Escherichia coli*, *Staphylococcus aureus*, *Pseudomonas aeruginosa*, and *Listeria monocytogenes*. Single colonies of pathogens cultured on LB agar plates were washed with PBS and resuspended. The absorbance (OD value) at 600 nm was then measured using a spectrophotometer. Based on a pre-constructed CFU-OD600 standard curve, the bacterial suspensions were diluted with PBS buffer to concentrations of 1.0 × 10³ CFU / mL and 1.0 × 10³ CFU / mL, respectively. 4 CFU / mL, 1.0×10 5 CFU / mL, 1.0×10 6 CFU / mL, 1.0×10 7 A concentration of CFU / mL was prepared for use. Next, six different concentrations of bacteria were mixed with three types of MXene@GQDs and incubated for 3 minutes, and fluorescence changes were measured using a fluorescence spectrophotometer. The fluorescence response data were normalized, and a fluorescence feature recognition model for foodborne pathogens was constructed based on the Naive Bayes algorithm. The dataset was divided into training and test sets in an 8:2 ratio using stratified random sampling, and five-fold cross-validation was used to evaluate the model performance. To improve feature extraction efficiency, principal component analysis (PCA) was introduced to reduce the dimensionality of the original fluorescence data.

[0073] like Figure 6 As shown, the detection results of Escherichia coli in this study are as follows: Principal component analysis (PCA) scoring plots indicate that the developed sensor array can effectively distinguish different concentrations (1.0 × 10³ CFU / mL, ... 4 CFU / mL, 1.0×10 5 CFU / mL, 1.0×10 6 CFU / mL, 1.0×10 7 Escherichia coli (CFU / mL) were detected. Furthermore, a significant linear relationship was observed between the fluorescence intensity of the sensor array and the concentration of E. coli. Circular clustering heatmaps clearly demonstrated the distinguishing ability of the three MXene@GQDs for different concentrations of E. coli, while circular bar stacking plots provided quantitative analysis for each dataset, visually illustrating the interaction between E. coli and the three MXene@GQDs at different concentrations. Overall, the three MXene@GQDs exhibited good distinguishing performance for single bacteria at different concentrations. Results for Staphylococcus aureus are shown below. Figure 7 The results for Salmonella typhimurium are shown in [link to results]. Figure 8The results for Shigella can be found in [link to results]. Figure 9 The results for Pseudomonas aeruginosa are shown in [the original text]. Figure 10 The results for Listeria monocytogenes are shown in [the original text]. Figure 11 The results showed that the sensor array was highly effective in distinguishing single bacteria at five different concentrations.

[0074] Example 7: Performance test of fluorescent sensor array in identifying mixed bacteria

[0075] Bacterial suspensions containing two and three different bacteria were prepared to test identification performance. ① Escherichia coli and Shigella were selected as research subjects, and bacterial suspensions containing these two bacteria were prepared. Individual colonies of both bacteria grown on Luria-Bertani (LB) agar plates were washed twice with phosphate-buffered saline (PBS; 3 mL). Then, E. coli and Shigella were mixed in different ratios (1.0 + 0, 0.8 + 0.2, 0.6 + 0.4, 0.5 + 0.5, 0.4 + 0.6, 0.2 + 0.8, 0 + 1.0). The resuspended bacteria were quantified using a spectrophotometer, and the OD value at 600 nm was measured. The results were then analyzed based on a pre-established CFU-OD ratio. 600 Standard curve was obtained by diluting the final suspensions of the two bacteria to 1.0 × 10⁻⁶ with PBS. 3 The bacterial suspension was incubated with three types of graphene quantum dots at CFU / mL for 3 min, followed by fluorescence change measurement using a fluorescence spectrophotometer. ② *Escherichia coli*, *Shigella*, and *Pseudomonas aeruginosa* were selected as research subjects, and a mixed bacterial suspension of these three bacteria was prepared. The procedure was the same as in ①, and the final suspensions of the two bacteria were diluted to 1.0 × 10⁻⁶ with PBS. 3 CFU / mL. The acquired fluorescence response data were normalized, and a mixed bacterial fluorescence feature recognition model was constructed based on the Naive Bayes algorithm.

[0076] like Figure 12 As shown, analysis of the test dataset using the Naive Bayes algorithm demonstrates its superior performance in identifying binary and ternary bacterial mixtures. Principal component analysis (PCA) score plots reveal significant discriminative power. Further confusion matrix and parallel coordinate plot analyses show that the Naive Bayes algorithm achieves 100% perfect recognition accuracy. These results fully demonstrate the excellent ability of the MXene@GQDs-based fluorescent sensor array to identify single bacteria and complex bacterial mixtures.

[0077] Example 8: Analysis of the anti-interference capability and stability of the fluorescence sensor array

[0078] In the anti-interference experiment, three MXene@GQDs sensor units were used to detect thirteen potential interfering substances (Na⁺, K⁺, Ca²⁺, F⁻, NO⁻, SO₄²⁻, Cl⁻, HPO₄²⁻, glucose, fructose, L-cysteine, arginine, and lysine) at a uniform concentration of 500 nM. The volume of both the interference solution and the bacterial suspension was 0.5 mL. *Escherichia coli* was selected as the research subject. First, single bacterial colonies grown on Luria-Bertani (LB) agar plates were washed twice with phosphate-buffered saline (PBS; 3 mL). Subsequently, the resuspended bacteria were quantitatively analyzed using a spectrophotometer, and their optical density (OD600) at 600 nm was measured. Based on a pre-established standard curve of colony forming units (CFU) versus OD600, the bacterial suspension was diluted with PBS to 1.0 × 10³ CFU / mL. Next, the prepared bacterial suspension and thirteen interference solutions were mixed with three types of MXene@GQDs solutions and incubated for 3 minutes, and then the fluorescence changes were detected using a fluorescence spectrophotometer.

[0079] like Figure 13 As shown, the fluorescence signal remained stable in the absence of target bacteria, indicating that environmental matrix components do not interfere with the intrinsic fluorescence of the sensor material. However, when 1.0 × 10⁻⁶ ppm was added... 3 At CFU / mL E. coli, all three sensor units exhibited significant fluorescence responses, confirming the system's excellent anti-interference performance. Repeatability studies showed that the fluorescence intensity remained highly stable across 15 consecutive measurements. Furthermore, from... Figure 13 As can be seen, the fluorescence signal fluctuation is minimal within the temperature range of 20℃-40℃, demonstrating the platform's wide temperature adaptability. These results collectively validate the sensor system's high selectivity, repeatability, and environmental robustness, highlighting its application potential in food safety testing and related fields.

[0080] Example 9: Detection of unknown bacterial samples by a fluorescent sensor array.

[0081] To evaluate the performance of the fluorescence sensor array, this study employed a blind sample test method to analyze single bacteria, binary mixtures, and ternary mixtures. The bacterial concentrations of all samples were standardized to 1.0 × 10⁻⁶. 3CFU / mL. ① The experimental methods for single bacterial samples of six bacteria are as follows. Six bacteria were selected as research subjects: *Salmonella typhimurium*, *Shigella*, *Escherichia coli*, *Staphylococcus aureus*, *Pseudomonas aeruginosa*, and *Listeria monocytogenes*. Single colonies of the six bacteria grown on Luria-Bertani (LB) agar plates were washed twice with phosphate-buffered saline (PBS; 3 mL). Then, the resuspended bacteria were quantified using a spectrophotometer, and the OD value at 600 nm was measured. Based on the pre-established CFU-OD... 600 Standard curve, obtained by diluting six bacterial suspensions to 1.0 × 10⁻⁶ with PBS. 3 CFU / mL, six different bacterial suspensions were bound to three types of MXene@GQDs for 3 minutes, and fluorescence changes were measured using a fluorescence spectrophotometer. ② The experimental method for mixed bacterial samples is as follows. *Escherichia coli* and *Shigella* were selected as the research objects, and bacterial suspensions of these two bacteria were prepared. Individual colonies of both bacteria grown on Luria-Bertani (LB) agar plates were washed twice with phosphate-buffered saline (PBS; 3 mL). Then, *E. coli* and *Shigella* were mixed in different ratios (1.0 + 0, 0.5 + 0.5, 0 + 1.0), and the resuspended bacteria were quantified using a spectrophotometer. The OD value at 600 nm was measured, and the results were determined based on the pre-established CFU-OD ratio. 600 Standard curve was obtained by diluting the final suspensions of the two bacteria to 1.0 × 10⁻⁶ with PBS. 3 CFU / mL, the bacterial suspension was incubated with three types of MXene@GQDs for 3 min, and then the fluorescence change was measured using a fluorescence spectrophotometer. ③ Escherichia coli, Shigella, and Pseudomonas aeruginosa were selected as research subjects, and a mixed bacterial suspension of these three bacteria was prepared. The operation method was the same as in ②, and Escherichia coli, Shigella, and Pseudomonas aeruginosa were mixed in different ratios (1.0 + 0 + 0, 0.2 + 0.2 + 0.6, 0.4 + 0.4 + 0.2, 0 + 1.0 + 0), and the final suspension of the three bacteria was diluted to 1.0 × 10⁻⁶ with PBS. 3 CFU / mL. Figure 14 As shown, the fluorescence sensor array achieved 100% accurate identification of blind samples.

[0082] Example 10: Detection of bacteria in tap water, milk, and pork using a fluorescence sensor array

[0083] Suspended solids in tap water samples were retained using a 0.22 μm pore size filter membrane. Pork and milk were collected from a local supermarket. In the experiment, 25 g of animal-derived food samples were mixed with 225 mL of sterile PBS buffer solution (pH 7) at a mass-to-volume ratio of 1:10 and homogenized for 5 minutes. The supernatant after centrifugation was collected, serially diluted 10-fold, and then inoculated with six validation strains and cultured overnight. After overnight culture, the bacterial resuspension was diluted to 1.0 × 10³ CFU / mL for later use. 0.5 mL of each of the six diluted bacterial suspensions was mixed with 0.5 mL of MXene@GQDs and incubated for 3 minutes. The fluorescence intensity was then measured using a fluorescence spectrophotometer. The fluorescence response data were normalized, and a fluorescence feature recognition model for foodborne pathogens was constructed based on the Naive Bayes algorithm. As shown in Figure 15, the PCA scores of the six foodborne pathogens in the actual samples—tap water, milk, and pork—showed clear differentiation.

[0084] This invention synthesizes nitrogen-sulfur co-doped graphene quantum dots (GQDs) via a one-step hydrothermal method and prepares monolayer Ti3C2T using a hydrochloric acid / lithium fluoride etching system. X MXene nanosheets. A heterojunction composite material (MXene@GQDs) was constructed by in-situ hydrothermal assembly of GQDs onto the MXene surface. Upon interaction with six foodborne pathogens, photoinduced electron transfer triggered differential fluorescence quenching in the MXene@GQDs. A 3×6 multichannel fluorescence sensor array was constructed by optimizing the probe concentration (0.3 mg / mL), excitation wavelength (360 nm), and incubation time (3 min). Combining principal component analysis and Naive Bayes algorithm, the system achieved 100% classification accuracy for both single and mixed bacterial samples, with a resolution ranging from 1.0×10³ to 1.0×10⁻⁶. 7 The sensor system successfully performed quantitative analysis of six bacteria within a CFU / mL concentration range. It maintained high sensitivity in complex matrices such as pork, milk, and tap water, and exhibited significant environmental adaptability.

[0085] The machine learning model used in this invention was implemented using Matlab (R2023b), and data visualization and analysis were performed using Origin 2024 and Matlab (R2023b). Model performance was comprehensively evaluated using four metrics: precision (PPV), recall (TPR), F1 score, and accuracy.

[0086] The algorithm database construction process is as follows:

[0087] (1) Mix 0.5 mL of six bacterial solutions with 0.5 mL of composite material (MXene@GQDs), incubate for 3 minutes, and then measure the luminescence intensity using a fluorescence spectrophotometer.

[0088] (2) A training matrix was established using the (F0-F) / F0 value (F0 represents the initial fluorescence intensity of MXene@GQDs, and F represents the fluorescence intensity of MXene@GQDs after mixing with bacteria). Eighteen rounds of repeated tests were conducted for each combination of bacteria and MXene@GQDs to form a complete training set (3 types of MXene@GQDs × 6 types of bacteria × 18 experiments).

[0089] (3) The obtained data is processed through a customized machine learning architecture, integrating five machine learning algorithms: K-nearest neighbors, Naive Bayes classifier, decision tree model, linear discriminant analysis (LDA), and support vector machine (SVM). The process of using machine learning algorithms to process the data is described in [link to documentation]. Figure 16 .

[0090] The above description is merely a preferred embodiment for explaining the present invention and is not intended to limit the present invention in any way. Therefore, any modifications or changes made to the present invention under the same inventive spirit should still be included within the scope of protection intended by the present invention.

Claims

1. A method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array, characterized in that: Includes the following steps: S1: Nitrogen-sulfur co-doped graphene quantum dots were synthesized using a one-step hydrothermal method; S2: Preparation of monolayer Ti3C2T using a hydrochloric acid / lithium fluoride etching system X MXene nanosheets; S3: Graphene quantum dots co-doped with nitrogen and sulfur and monolayer Ti3C2T X Using MXene nanosheets as raw materials, heterojunction composite materials were constructed through in-situ hydrothermal assembly; S4: Measure the fluorescence emission spectrum of the heterojunction composite material, optimize its excitation wavelength, reaction concentration, and incubation time with bacteria, and measure the fluorescence intensity of the heterojunction composite material under optimal conditions; S5: Single colonies of pathogenic bacteria cultured on LB agar plates were washed with PBS and resuspended, and the absorbance at 600 nm was measured using a spectrophotometer; based on the constructed CFU-OD... 600 Standard curves were obtained by diluting the bacterial suspension to a series of target concentrations using PBS buffer. S6: Construct a fluorescence sensor array, mix the heterojunction composite material solution with bacterial solutions of different concentrations at a set volume ratio, incubate, and measure the fluorescence intensity at the optimized excitation wavelength to optimize signal detection. Use a fluorescence spectrophotometer to record the fluorescence changes before and after the interaction between the fluorescent material and bacteria. S7: Normalize the fluorescence response data and construct a fluorescence feature recognition model for foodborne pathogens based on five machine learning algorithms; The dataset was divided into training and test sets in an 8:2 ratio using stratified random sampling, and the model performance was evaluated using five-fold cross-validation. To improve feature extraction efficiency, principal component analysis was introduced to process the raw fluorescence data. The machine learning process was implemented using Matlab, and data visualization and analysis were performed using Origin 2024 and Matlab. Model performance was evaluated using four metrics: precision, recall, F1 score, and accuracy.

2. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: In step S1, the method for synthesizing nitrogen-sulfur co-doped graphene quantum dots includes: (1) Synthesis of graphene quantum dot 1 Add anhydrous citric acid and thiourea to water and stir. Then transfer the mixture to an autoclave with a Teflon liner and heat at 150-170°C for 5-8 hours. (2) Synthesis of graphene quantum dot 2 Anhydrous citric acid and L-cysteine ​​are added to water and stirred. Then the mixture is heated at 170-190°C for 8-12 hours in a reaction vessel with a Teflon liner. (3) Synthesis of graphene quantum dot 3 Add anhydrous citric acid, sulfur powder, and ammonia to water, stir, and then heat at 170-190°C for 10-14 hours in a Teflon-lined autoclave.

3. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: In step S2, a single layer of Ti3C2T X The methods for synthesizing MXene nanosheets include: The aluminum layer in Ti3AlC2 was selectively etched using LiF / HCl solution. After acid washing and purification, multiple water washings and centrifugation were performed to control the stratification. When the supernatant became turbid and separated into layers, high-speed centrifugation was used to enrich the precipitate. Continuous washing until neutral was performed to obtain multilayer Ti3C2. X Monolayer Ti3C2T was finally prepared by ultrasonic exfoliation, liquid phase separation and freeze drying under nitrogen protection. X MXene nanosheets.

4. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 3, characterized in that: Add Ti3AlC2 powder to LiF / HCl solution and stir continuously at 40-50°C for 20-30 hours. Centrifuge the resulting reaction solution at 3000-4000 rpm for 3-8 minutes and collect the bottom precipitate. Redissolve the precipitate in hydrochloric acid and wash with deionized water, repeating 2-3 times. Then wash with deionized water and centrifuge at 3000-4000 rpm for 3-8 minutes. Continue centrifuging and washing with deionized water 3-5 times until the supernatant becomes turbid and separates into layers. Centrifuge at 9000-10000 rpm to collect the bottom precipitate. Repeat the deionized water washing-centrifugation until the pH of the supernatant is ≥ 6 to obtain the precipitate. Disperse the precipitate in deionized water and sonicate in an ice-water bath under nitrogen protection for 1.5-2.5 hours. Centrifuge the ultrasonic dispersion at 3000-4000 rpm for 20-40 minutes, collect the upper liquid phase, and finally freeze-dry to obtain a monolayer Ti3C2T. X MXene nanosheets.

5. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: In step S3, the monolayer Ti3C2T X MXene nanosheets and nitrogen-sulfur co-doped graphene quantum dots were added to deionized water and stirred to form a uniform suspension. The suspension was then transferred to a reactor for hydrothermal reaction. The product was centrifuged, washed multiple times with deionized water, and finally freeze-dried to obtain a heterojunction composite material.

6. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: In step S4, the optimal excitation wavelengths for the three heterojunction composite materials used in the test were 350 nm, 360 nm and 360 nm, respectively; the optimal reaction concentration was 0.3 mg / mL; and the optimal incubation time with bacteria was 3 minutes.

7. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: In step S5, the pathogens include Salmonella Typhimurium, Shigella, Escherichia coli, Staphylococcus aureus, Pseudomonas aeruginosa, and Listeria monocytogenes.

8. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: In step S7, the formula for normalizing the fluorescence response data is: (F0-F) / F0; where F0 is the fluorescence intensity before the addition of pathogens, and F is the fluorescence intensity after the addition of pathogens; the five machine learning algorithms are: support vector machine, decision tree, K nearest neighbor, Naive Bayes and linear discriminant analysis.

9. The method for rapid detection and identification of foodborne pathogens using a machine learning-based two-dimensional / zero-dimensional MXene / GQDs Schottky heterojunction fluorescence sensor array according to claim 1, characterized in that: Three heterojunction composite materials were selected, and experiments were conducted against six pathogens. Each experiment was repeated eighteen times to establish a training dataset. The calculation method was to multiply the three heterojunction composite materials, the six pathogens, and the eighteen repeated experiments together.

Images