A fluorescent array sensing system based on biomimetic taste and construction method thereof and application in food safety intelligent detection

Abstract

The application discloses a fluorescent array sensing system based on biomimetic taste and a construction method and application in intelligent detection of food safety. 2 As a biomimetic tongue carrier, MXene nanosheet adsorbs multicolor silver nanoclusters synthesized by oligonucleotide chains as biomimetic taste bud cells, different DNA chains can form different fluorescent emissions, and the system has the characteristic of fluorescent design, which can greatly simplify the signal molecule preparation process. 2 When multiple hazards are added to the sensing array, Ti3C MXene can preferentially capture target objects, while silver nanoclusters with weak binding capacity are released or adsorbed, resulting in changes in different color fluorescent signals and forming a characteristic "fingerprint map". Combined with machine learning algorithms, the system can realize rapid, accurate, qualitative and quantitative intelligent identification of multiple food hazards, providing a new strategy for simultaneous detection of multiple food hazards.

Description

Technical Field

[0001] This invention relates to the field of food hazardous substance detection and control, and in particular to a biomimetic taste-based fluorescence array sensing system, its construction method, and its application in intelligent food safety detection. Background Technology

[0002] Food safety is directly related to public health and social stability. Potential hazards in food include heavy metal ions, pesticide residues, pathogenic microorganisms, antibiotics, and fungal toxins. These hazards are not only complex in origin and diverse in type, but some can also pose long-term health risks to humans even at low concentrations. Therefore, the rapid and accurate detection and control of food hazards is a core component of modern food safety management systems.

[0003] Existing detection technologies, such as chromatography, mass spectrometry, electrochemistry, and immunoassay, while possessing high sensitivity and accuracy, generally suffer from drawbacks including long detection cycles, expensive equipment, complex operation, and the need for specialized personnel and suitable laboratory environments. Furthermore, these methods mostly target single substances, making it difficult to simultaneously distinguish multiple analogues of the same type of hazard in a single detection, such as different types of pathogens, pesticide residues with different structures, or metal ions with different valence states. This limitation is particularly pronounced in complex food matrices, significantly hindering the realization of rapid on-site screening and multi-target monitoring.

[0004] Bionic gustatory sensing arrays mimic the cross-response characteristics of human taste bud cells. Utilizing multiple recognition units with varying selectivity but limited specificity, they generate differentiated signal responses to various analytes, and then achieve pattern recognition through multivariate data analysis or machine learning algorithms. The core principle is that each sensing unit produces different degrees of fluorescence response to similar targets, forming a unique "fingerprint," thereby enabling rapid differentiation and identification of multiple analogues of the same hazardous substance. Unlike traditional specific detection, this cross-response strategy does not rely on a single highly selective probe, but rather on signal combination patterns for identification. It offers advantages such as multi-target parallel analysis, rapid response, and good scalability, providing a new solution for the simultaneous detection of multiple hazardous substances in complex food systems.

[0005] Currently, there are numerous patent applications and papers concerning the detection of hazardous substances in food using sensor arrays. For example, authorized invention patent 202210172775.1 describes the preparation of various quantum dots using different synthesis methods to construct fluorescent arrays for the simultaneous detection of multiple heavy metal ions; authorized patent 202211421323.9 describes the synthesis of B-doped graphene quantum dots using different solvents to form a 3×2 fluorescent sensor array for the simultaneous detection of multiple plasticizers; invention patent application 202311257075.3 proposes the synthesis of fluorescent MOF film arrays using different metal ions and organic ligands for simultaneous response to different types of antibiotics; and other invention patent applications... Please refer to patent application 202411284521.4, which utilizes a fluorescent array composed of three fluorescent polymers and one amphiphilic polymer to achieve rapid detection of multiple bacteria by using a single nanopolymer quantum dot in response to various bacterial metabolites; similar invention patent applications include 202310533617.9, 202411045198.5, 202510417828.5, 202411672200.1, 202510833301.7, 202110200295.7, 202210071097.X, 202210618987.8, 202210697836.6, etc. Similarly, the journal article Angewandte Chemie International Edition, 2024, 63(16), e202318483, developed a dual-channel fluorescence “on” sensor array by electrostatically combining six aggregation-induced emission (AIE) fluorophores to simultaneously identify 20 bacteria. Through machine algorithm optimization, the accuracy of real samples can reach 96.7%. Similar research can be found in Advanced Functional Materials, 2019, 29(4), 1805986. Besides fluorescence sensing arrays, colorimetric sensing arrays have also been widely reported in the detection and control of hazardous substances in food. For example, authorized invention patent 202110166644.8 proposes a colorimetric array sensor based on silver ions and o-phenylenediamine and its derivatives for the identification of various small organic molecules such as aldehydes, ketones, acids, and esters. Invention patent application 202510254111.3 develops a single-atom iron nanozyme system that utilizes the inhibition of oxidation of three chromogenic substrates by different antioxidants and polyphenols to construct a colorimetric sensing array, achieving the identification and detection of five antioxidants and polyphenols. Invention patent application 202510499395.2 synthesizes a hierarchical ZrMOF@Pt@ZrMOF composite material, utilizing this composite material to catalyze the enzymatic reaction between the chromogenic substrate and hydrogen peroxide, achieving the simultaneous detection of seven organophosphorus pesticides.

[0006] While the aforementioned patent applications and papers demonstrate good results in the simultaneous detection of multiple food hazards, it's evident that these arrays require multiple highly sensitive sensing units, each requiring complex and cumbersome synthesis steps. This undoubtedly hinders practical applications. Furthermore, in these detection systems, signal molecules and target analytes often interact directly to generate measurable signals. Although this method is structurally simple and fast-responding, it suffers from insufficient sensitivity, poor anti-interference capabilities, limited detection range, and poor signal stability in practical applications. In the direct binding mode, signal molecules lack a signal amplification mechanism, making them susceptible to background noise at low target analyte concentrations and prone to non-specific interference from structurally similar substances, leading to false positives or false negatives. Based on these two shortcomings, it is necessary to find more suitable signal molecules and sensing strategies to simplify the sensor array fabrication process and significantly improve detection sensitivity.

[0007] DNA-AgNCs are a class of ultra-small silver atom clusters stabilized by interactions such as electrostatics, coordination bonds, and hydrogen bonds, using DNA molecules with specific sequences and spatial configurations as templates. Due to their size being in the metal-molecule transition region (<2 nm), they exhibit unique molecular-state fluorescence emission characteristics, with emission wavelengths adjustable from green to near-infrared by modulating the DNA sequence, base composition, and secondary structure. Furthermore, the DNA backbone provides excellent dispersion and stabilization for silver atoms, resulting in high quantum yield, good photostability, and biocompatibility in aqueous solutions. Invention patent application CN202310063500.9 constructed a DNA-AgNCs-based fluorescent biosensor for the detection of the insecticide thiram. Although it showed good detection results, the method only validated one insecticide and lacked validation against interference from other insecticides. For complex food media, if multiple insecticides are present, the accuracy of this method needs further validation. Similar studies include authorized patents ZL202211154362.7 and ZL201811064348.1, as well as invention patent application CN20241079567.9, all of which lacked validation studies on analogues of the target substance and could not achieve simultaneous detection of multiple food hazards in a single test. In addition, thanks to the programmability of the sequence, DNA-AgNCs can not only serve as multicolor fluorescent signal units, but also achieve specific binding and response to target molecules by introducing specific recognition sequences into the DNA chain. Therefore, it has broad application potential in multiplex detection, molecular imaging, and biosensing.

[0008] MXenes are two-dimensional nanomaterials composed of transition metal carbides, nitrides, or carbonitrides. They possess high specific surface area, excellent metallic conductivity, and abundant tunable surface functional groups (–OH, –O, –F, etc.), endowing them with outstanding fluorescence quenching capabilities and diverse surface chemical modification potential. Thanks to their surface's ability to efficiently bind various molecules through electrostatic interactions, π–π stacking, hydrogen bonding, and coordination interactions, MXenes can not only serve as highly efficient fluorescence quenchers but also as a platform for recognizing food hazards. When signal molecules preferentially bind to the MXene surface, their fluorescence is significantly quenched; conversely, when the target analyte and signal molecule compete for binding on the MXene surface, the signal molecule is released, and fluorescence is restored, thus forming a competitive recognition detection mode. Compared to traditional methods relying on specific modification of signal molecules, this strategy achieves efficient detection without complex chemical modifications to signal molecules, while significantly improving the sensitivity, selectivity, detection range, and signal stability of the sensing array, providing an effective approach for constructing high-performance, multi-target rapid detection platforms. Summary of the Invention

[0009] To address the aforementioned issues, this invention discloses a multicolor fluorescence array detection method based on a biomimetic taste strategy, utilizing the competitive binding of DNA-AgNCs and MXene. This method is suitable for the simultaneous and highly sensitive detection of various food hazards. Using oligonucleotide chains as templates, multicolor DNA-AgNCs are synthesized in one step, simplifying the preparation process of multicolor signal units. MXene serves as a highly efficient fluorescence quencher and recognition platform. Through the competitive binding of food hazards and signal molecules on the MXene surface, signal amplification, enhanced selectivity, and a broadened detection range are achieved. This system possesses multi-channel response, high photostability, biocompatibility, and anti-interference capabilities, enabling rapid and accurate multiplex detection in complex food matrices. It features simple preparation, high sensitivity, strong selectivity, and wide applicability.

[0010] The objective of this invention is achieved through the following technical solutions.

[0011] A method for constructing a biomimetic taste-based fluorescence array sensing system includes the following steps: (1) The fabrication steps of the bionic tongue recognition element are as follows: Using the MAX phase precursor as raw material, the A element in the MAX phase is removed by chemical etching to obtain multilayer Ti3C2Tx MXene nanosheets with functional groups on the surface; the multilayer Ti3C2Tx MXene is then subjected to intercalation treatment, and the intercalated multilayer Ti3C2Tx MXene is peeled into single-layer or few-layer ultrathin nanosheets by ultrasonic peeling, mechanical shearing, stirring peeling or electrostatic peeling to obtain a biomimetic tongue recognition element.

[0012] Preferably, the chemical formula of the MAX phase in step (1) is: Where M is a transition metal element, A is a group A element, X is carbon or nitrogen, n = 1 ~ 3, preferably Ti3AlC2, Ti2AlC, Ti3SiC2 or Ti2AlN.

[0013] Preferably, the etching agent in step (1) is a fluoride solution, including one or more of hydrofluoric acid, a mixture of lithium fluoride and hydrochloric acid, a mixture of ammonium fluoride and hydrochloric acid, ammonium hydrogen fluoride (NH4HF2) solution, molten salt or alkaline etching system; the etching temperature is 20 ~ 60 ℃, and the etching time is 12 ~ 72 h.

[0014] Preferably, the intercalating agent in step (1) includes one or more of dimethyl sulfoxide, tetrabutylammonium hydroxide, tetramethylammonium hydroxide, lithium chloride, sodium chloride, surfactant or small molecule organic amine to increase the interlayer spacing, the intercalation time is 12 to 48 h, and the intercalation temperature is 20 to 60 °C.

[0015] Preferably, the peeling in step (1) is performed using ultrasonic peeling (power 10 ~ 200 W, 1 ~ 60 min, 0 ~ 40℃), mechanical shear peeling (rotation speed 1000 ~ 20000 rpm, 1 ~ 60 min), stirring intercalation peeling (rotation speed 200 ~ 2000 rpm, 6 ~ 72 h, 0 ~ 60 ℃), or electrostatic / electrochemical peeling (voltage -1.0 ~ +5.0 V or current density 0.1 ~ 10 mA·cm). -2 Monolayers were peeled off using at least one of the following methods (1 min ~ 6 h), and the monolayer products were then collected by centrifugation fractionation.

[0016] (2) The functionalization steps of the bionic tongue recognition element are as follows: The monolayer Ti3C2Tx MXene nanosheets prepared in step (1) are functionalized to achieve specific recognition of various hazardous substances and improve the dispersibility, anti-pollution and stability of the carrier in complex matrices; the hazardous substances are selected from at least one of metal ions, foodborne pathogens, pesticide residues, fungal toxins and antibiotics.

[0017] Preferably, the functionalization modification in step (2) includes grafting a hydrophilic polymer onto the Ti3C2Tx surface, including polyacrylic acid (PAA), polyethylene glycol (PEG), phosphorylated polyethylene glycol (PEG-PO3H2), dextran, sodium alginate, or chitosan, to improve dispersibility and antifouling properties. The concentration of the polyacrylic acid monomer is 0.05~1.0 M, the reaction temperature is 0~80 ℃, and the time is 2~24 h; the molecular weight of the polyethylene glycol is 500~20000 Da, the concentration is 0.1~50 mg / mL, the temperature is 0~80 ℃, and the reaction time is 0.5~48 h; the molecular weight of the phosphonic polyethylene glycol backbone is 500~10000 Da, the concentration is 0.01~20 mM, the pH is 1.5~8.0, and the temperature is 0~80 ℃; the molecular weight of the chitosan is 50000~500000 Da. Da, with a degree of deacetylation of 70-95%, a solution concentration range of 0.01-5.0 wt%, a pH of 4.0-60 ℃, and a reaction time of 0.5-48 h.

[0018] Preferably, the functionalization modification in step (2) includes introducing ligands containing thiol, amino, or phosphate groups onto the surface of Ti3C2Tx MXene nanosheets to chelate Hg... 2+ Pb 2+ Cd 2+ Cu 2+ Fe 2+ Zn 2+ Mn 2+ As 3+ The metal ions; the ligand is at least one of mercaptoacetic acid, cysteine, ethylenediaminetetraacetic acid, and phosphorylated polyethylene glycol, which is modified on the surface of Ti3C2Tx MXene nanosheets by covalent bonding, electrostatic adsorption, or π–π interaction.

[0019] Preferably, the functionalization modification in step (2) includes immobilizing a specific recognition molecule on the surface of Ti3C2Tx MXene nanosheets. The recognition molecule is selected from at least one of monoclonal antibodies, peptides, aptamers, phage display peptides, lectins, and phenylboronic acid, and is used to bind Staphylococcus aureus, Escherichia coli O157:H7, Salmonella, Listeria, or Pseudomonas aeruginosa. The immobilization method is glutaraldehyde crosslinking, EDC / NHS activation coupling, maleimide-thiol coupling, or click chemistry reaction.

[0020] Preferably, the functionalization modification in step (2) includes covalently coupling or physically adsorbing specific enzymes or aptamers on the surface of Ti3C2Tx MXene nanosheets. The enzymes are selected from acetylcholinesterase (AChE), peroxidase, urease, etc., and are used to recognize organophosphates, carbamates, pyrethroids, or glyphosate pesticides. The aptamers are fixed to the Ti3C2Tx surface through π–π interactions or hydrogen bonding.

[0021] Preferably, the functionalization modification in step (2) includes introducing aptamers, antibodies, molecularly imprinted polymers, or cyclodextrin derivatives onto the surface of Ti3C2Tx MXene nanosheets to achieve specific recognition of fungal toxins such as aflatoxin B1 (AFB1), aflatoxin M1 (AFM1), ochratoxin A (OTA), fumonisin B1 (FB1), and zearalenone (ZEN); wherein the aptamer sequence length is 20–80 bases, and may contain stem-loop, G-quadruplex, or hairpin structures; the binding constant Ka between the template molecule of the molecularly imprinted polymer and the target fungal toxin is ≥ 10. 5 M⁻¹, with a coating thickness of 5 ~ 50 nm, to ensure high selectivity and stability in complex food matrices.

[0022] Preferably, the functionalization modification in step (2) includes immobilizing an antibiotic aptamer, molecularly imprinted polymer (MIP) or metal-organic framework (MOF) coating on the surface of Ti3C2Tx MXene nanosheets for specific binding of tetracycline, fluoroquinolone, macrolide or β-lactam antibiotics; the immobilization method is surface in-situ polymerization, sol-gel method or layer-by-layer self-assembly method.

[0023] (3) The preparation steps of the biomimetic taste bud cell signaling element are as follows: Oligonucleotide DNA with different structures is used as a template. The DNA structure includes single-stranded, hairpin, G-quadruplex, or other secondary structures. The DNA modulates the multicolor fluorescence emission of silver nanoclusters by adjusting its base composition, length, and structure. In a suitable pH buffer system, the DNA template is reacted with silver ions (Ag). + The mixture is combined with a borate reducing agent and reduced under mild conditions to generate silver nanoclusters (AgNCs). By adjusting the reaction temperature, pH value, reducing agent concentration and reaction time, the size and fluorescence properties of silver nanoclusters can be precisely controlled in one step, and various fluorescent emission silver nanocluster signal elements can be obtained.

[0024] Preferably, in the preparation of the biomimetic taste bud cell signaling element in step (3), the DNA template has a sequence length of 10 to 40 nucleotides, the DNA structure is a stem-loop structure, the stem part contains a stable double strand formed by complementary base pairing, and the loop part contains a sequence rich in cytosine (C) or rich in guanine (G).

[0025] Preferably, in the preparation of the biomimetic taste bud cell signaling element in step (3), the borate reducing agent is selected from sodium borate (NaBH4), sodium borohydride, or boric acid (H3BO3).

[0026] Preferably, the preparation of the biomimetic taste bud cell signaling element in step (3) is carried out in a buffer solution of phosphate, tris(hydroxymethyl)aminomethane, citrate, acetate, etc. at pH 7.0 to 10.0, at a temperature of -20 ℃ to 37 ℃, and for a reaction time of 0.5 to 48 h.

[0027] This invention utilizes oligonucleotide chains as templates to directly synthesize multicolor silver nanoclusters (AgNCs) via a one-step reduction method. Different base sequences and conformations allow for precise control of the nanocluster's particle size and electronic structure, thereby endowing it with multiple distinguishable emission wavelengths and optical properties. This method not only avoids the cumbersome steps of traditional multicolor fluorescent probes, which require separate synthesis, purification, and modification, but also offers advantages such as mild reaction conditions, simple operation, high yield, and mass production capability. It enables the simultaneous acquisition of multicolor sensing units in the same reaction system, significantly improving array construction efficiency and reducing production costs.

[0028] (4) The fabrication steps of the fluorescence sensing array based on the biomimetic taste sensing strategy are as follows: The multicolor DNA-AgNCs prepared in step (3) were diluted with different buffer solutions to prepare working solutions of appropriate concentrations. Different concentrations of functionalized Ti3C2Tx MXene nanosheets modified in step (2) were gradually added to the DNA-AgNCs working solutions. After thorough mixing, the optimal emission peak intensity of each DNA-AgNCs under the optimal excitation wavelength was measured. As the concentration of functionalized Ti3C2Tx nanosheets increased, the fluorescence intensity of DNA-AgNCs was gradually quenched. When the fluorescence intensity of DNA-AgNCs decreased to 25-40% of the initial fluorescence intensity, the biomimetic taste sensing system was successfully prepared. Furthermore, the binding constant between DNA-AgNCs and functionalized Ti3C2Tx nanosheets at this concentration was calculated using the Stern-Volmer equation to evaluate the differences in binding performance of each DNA-AgNCs and to screen suitable DNA templates for the preparation of DNA-AgNCs.

[0029] (1) Where F0 and F represent the fluorescence emission intensity of DNA-AgNCs and DNA-AgNCs / Ti3C2Tx nanosheets at the optimal emission wavelength. K SV is the binding constant, and M is the concentration of Ti3C2Tx nanosheets.

[0030] Preferably, in step (4), the preparation of the fluorescent sensing array based on the biomimetic taste strategy involves the following dilution conditions for the DNA-AgNCs working solution: the initial concentration of DNA-AgNCs is 5-50 mM, and it is diluted with a buffer solution of phosphate, tris(hydroxymethyl)aminomethane, citrate, acetate, etc. with a concentration of 5-20 mM and a pH of 7.0-10.0 at a ratio of 1:3-1:20 (v / v) to prepare a DNA-AgNCs working solution with a final concentration of 0.05-5 μM.

[0031] Preferably, in step (4), the concentration of the functionalized Ti3C2Tx MXene nanosheet dispersion is in the range of 0.01 ~ 1.0 mg / mL, and it is incubated in a shaker at 30 ~ 50℃ and 200 ~ 800 rpm for 5 ~ 60 min to ensure thorough mixing.

[0032] Preferably, in step (4), when the binding constant between DNA-AgNCs and functionalized Ti3C2Tx nanosheets calculated using the Stern-Volmer equation is 10 ~ 40, it indicates that the DNA-AgNCs synthesized from the DNA template are suitable for preparing a fluorescence sensing array system.

[0033] A biomimetic taste-based fluorescent array sensing system prepared by any of the above methods.

[0034] The application of the above-described biomimetic taste-based fluorescence array sensing system in the detection of hazardous substances in food includes the following steps: (1) Preparation of standard solutions for food hazard analytes: Select at least one of heavy metal ions, pesticide residues, mycotoxins, foodborne pathogens and antibiotics as target analytes, and prepare single-system or multi-component mixed-system analyte solutions with different concentration gradients. Preferably, the food hazard analyte in step (1) is a heavy metal ion selected from Hg. 2+ Pb 2+ Cd 2+ Cu 2+ Fe 2 + Zn 2+ Mn 2+Or As 3+ The concentration gradient range is 0 ~ 100 μM; the pesticide residues are selected from organophosphates, carbamates, pyrethroids, or glyphosate, with a concentration gradient range of 0 ~ 100 μM; the mycotoxins are selected from aflatoxin B1, ochratoxin A, fumonisin B1, or vomitoxin, with a concentration gradient range of 0 ~ 100 μM; the foodborne pathogens are selected from Staphylococcus aureus, Escherichia coli O157:H7, Salmonella, Listeria, or Pseudomonas aeruginosa, with a concentration gradient in the range of OD. 600 The value is characterized as 0 to 0.5. The antibiotic is selected from tetracyclines, fluoroquinolones, macrolides, or β-lactams, with a concentration gradient range of 0 to 100 μM.

[0035] Preferably, the two-component mixing system in step (1) is prepared with equal concentrations and in volume ratios of 1:9, 2:8, 3:7, 4:6, 5:5, 6:4, 7:3, 8:2, and 9:1.

[0036] Preferably, the three-component mixture system in step (1) is prepared with equal concentrations and in volume ratios of 4:3:3, 6:2:2, 8:1:1, 3:4:3, 2:6:2, 4:4:2, 3:3:4, 2:2:6, and 1:1:8.

[0037] (2) Construction of the mixed system: The working solutions of each channel of the fluorescence array sensing system prepared in step (4) are dispensed into microplates, the analyte solutions are added to each, and the mixture is incubated under constant temperature to allow it to fully combine. The fluorescence emission intensity at the optimal excitation / emission wavelength of each channel is measured. Preferably, in step (2), the volume of working solution added to each channel is 50 ~ 150 μL, the volume of analyte solution added is 20 ~ 50 μL, the incubation temperature is 20 ~ 40℃, and the rotation speed is 200 ~ 300 rpm.

[0038] (3) Data modeling and pattern recognition: The fluorescence response matrix is ​​used for pattern recognition and classification by machine learning algorithm, the model parameters are optimized by optimization algorithm, and the model performance is analyzed based on evaluation index to verify the recognition ability and stability of the sensing system for different analytes; Preferably, in step (3), the machine learning algorithm is selected from the following: autonomous component analysis (PCA), hierarchical clustering analysis (HCA), linear discriminant analysis (LDA), k-nearest neighbors (KNN), support vector machine (SVM), random forest (RF), logistic regression (LR), convolutional neural network (CNN), recurrent neural network (RNN), Transformer or long short-term memory network (LSTM). Preferably, the parameter optimization in step (3) adopts at least one of cross-validation, grid search, Bayesian optimization, genetic algorithm optimization, adaptive gradient optimization, learning rate scheduling, Dropout regularization or L1 / L2 regularization; Preferably, the model evaluation metrics in step (3) include accuracy, precision, recall, F1-score, and confusion matrix.

[0039] (4) Blind sample and actual sample detection: The sensing system is applied to the detection of blind samples and actual samples, and the machine learning model is used to predict and classify the samples to evaluate the applicability of the sensing system in the detection of food hazards.

[0040] Preferably, the actual sample in step (4) can be fruit juice beverages, milk and dairy products, fruits and vegetables, meat and meat products, aquatic products and their products, grains and grain products, beverages, etc. Preferably, step (4) sample pretreatment includes dilution, homogenization, settling, and filtration; the blind sample and the actual sample are added to the analyte from step (1) at a preset target level (in μM or OD). 600 Characterization), and incubation and measurement are performed according to the conditions of step (2), and the prediction and classification are completed by the model established in step (3).

[0041] Based on a biomimetic taste strategy, this invention proposes a highly sensitive array detection strategy based on the competitive binding between multicolor silver nanoclusters (DNA-AgNCs) and Ti3C2TxMXene and various food hazards. This strategy simplifies the construction process of the sensing unit and significantly improves the sensitivity, selectivity, detection range and signal stability of the sensing array through the competitive mechanism.

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects: (1) Multicolor DNA-AgNCs were prepared based on different DNA templates to realize the construction of multi-channel fluorescence signals, which simplified the dependence of traditional arrays on multiple fluorescent probes or complex modifications, reduced the preparation cost and improved the stability and consistency of the system.

[0043] (2) Using Ti3C2Tx MXene with surface-specific functionalization as a co-recognition substrate, its excellent conductivity and multi-site interaction ability are used to introduce a competitive binding mechanism, which significantly improves the response diversity and discrimination ability of various food hazards such as heavy metal ions, pesticide residues, mycotoxins, foodborne pathogens and antibiotics.

[0044] (3) By combining machine learning and pattern recognition algorithms, the multi-channel fluorescence response matrix is ​​modeled and optimized to achieve high-precision and automated classification and prediction of various hazards, enhance the adaptability and stability of the detection system in complex food matrices, and has broad potential for promotion and application. Attached Figure Description

[0045] Figure 1 The image shows the X-ray diffraction (XRD) patterns of Ti3C2 and Ti3AlC2 prepared in Example 1.

[0046] Figure 2 The dispersion diagram of Ti3C2 modified with CS, PAA, PEG, and PEG-PO3H2 prepared in Example 2 in water.

[0047] Figure 3 The Fourier transform infrared absorption spectra of Ti3C2, PAA-Ti3C2 and APBA-PAA-Ti3C2 prepared in Examples 1 and 2 are shown.

[0048] Figure 4 The images show the fluorescence emission spectra of the four DNA-AgNCs prepared in Example 3, as well as images under UV 365 nm excitation and sunlight.

[0049] Figure 5 The graph shows the fluorescence intensity changes of the four DNA-AgNCs prepared in Example 3 as the concentration of APBA-PAA-Ti3C2 increases (0 ~ 0.6 mg / mL).

[0050] Figure 6 The images show fingerprints (A) and two-dimensional LDA analysis (B) of the fluorescence sensing array prepared in Example 4 for detecting the concentration of single-component pathogenic bacteria, as well as the detection and analysis of two-component (E. coli + S. aureus, Figure C) and three-component (E. coli + S. aureus + S. typhimurium, Figure D) mixed bacteria.

[0051] Figure 7 The figures show a comparison of the accuracy of machine learning and deep learning prediction models for detecting pathogens in tap water constructed in Example 6 (A), a graph of various indicators of the CNNs model (B), and a confusion matrix (C).

[0052] Figure 8 The fluorescence sensing array prepared in Example 4 was used to detect a single component (0.1 OD) in beef. 600Two-dimensional LDA analysis diagrams for the detection of five pathogenic bacteria (Figure A), two-component (E. coli and S. aureus, S. aureus and S. typhimurium, L. monocytogenes and S. typhimurium, Figure B), and three-component (E. coli and S. aureus and S. typhimurium (Group A), S. aureus and S. typhimurium and L. monocytogenes (Group B), L. monocytogenes and S. typhimurium and S. putrefaciens (Group C), E. coli and S. typhimurium and S. putrefaciens (Group D, Figure C) pathogenic bacteria.

[0053] Figure 9 Two-dimensional LDA analysis diagrams for the detection of five pathogenic microorganisms using only four DNA-AgNCs (A), using unfunctionalized Ti3C2 as a vector (B), and using PAA-Ti3C2 as a vector (C) are shown for comparison examples 1, 2, and 3.

[0054] Figure 10 The DNA-AgNCs@Cys-PEG-Ti3C2, DNA-AgNCs@MIP-chitosan-Ti3C2, and DNA-AgNCs@AChE-PEG-PO3H2-Ti3C2 prepared in Example 9 were used to simultaneously detect five 0.1 μM heavy metal ions, fungal toxins, and organophosphorus pesticides.

[0055] Figure 11 This is a schematic diagram illustrating the application of a biomimetic taste-based fluorescence array sensing system in intelligent food safety detection. Detailed Implementation

[0056] The following examples, using foodborne pathogens as a typical representative of food hazards, will be used to further illustrate the present invention in detail. However, the implementation of the present invention is not limited thereto.

[0057] Example 1: Preparation of ultrathin Ti3C2 nanosheets as a biomimetic tongue carrier LiF (3 g) was gradually added to 40 mL of 9 M HCl, and stirred for 30 min to ensure complete dissolution. Then, Ti3AlC2 powder (2 g) was added in five portions over 1 h, and the mixture was stirred continuously at 35 °C for 36 h. The resulting suspension was washed more than five times with ultrapure water and then centrifuged at 4000 rpm for 10 min. The resulting precipitate was redispersed in approximately 200 mL of ultrapure water and sonicated under nitrogen protection for 1 h. After centrifugation at 4000 rpm for 30 min, the supernatant was collected to obtain ultrathin monolayer Ti3C2 nanosheets.

[0058] according to Figure 1 XRD results showed that most of the diffraction peaks of Ti3AlC2 disappeared and the (002) diffraction peak shifted to the lower angle region, which proved that the MAX phase was successfully exfoliated and ultrathin monolayer Ti3C2 nanosheets were prepared.

[0059] Example 2: Functionalization of ultrathin Ti3C2 nanosheets To increase the hydrophilicity of Ti3C2 nanosheets, physical adsorption methods were employed on the Ti3C2Tx surface to enhance hydrophilicity and antifouling capabilities. These methods included the use of chitosan, polyacrylic acid (PAA), polyethylene glycol (PEG), and phosphonated polyethylene glycol (PEG-PO3H2). The relevant preparation process is as follows: Chitosan modification: Chitosan was dissolved in a 0.5% (v / v) aqueous acetic acid solution to prepare a 1 wt% solution. 6 mL of this solution was slowly added to 40 mL of Ti3C2 solution (1 g / L), and the mixture was sonicated for 6 h under nitrogen protection. The mixture was then centrifuged at 5000 rpm for 30 min to remove large particles, and free chitosan and residual acetic acid were removed by repeated centrifugation and washing. After 24 h of lyophilization, Chitosan-Ti3C2 dry powder was obtained.

[0060] Polyacrylic acid (PAA) modification: 6 mL of PAA was added to 40 mL of Ti3C2 solution (1 g / L). The entire system was then sonicated for 6 h under nitrogen protection. The mixture was then centrifuged at 5000 rpm for 30 min to remove large particles. Excess PAA was removed by dialysis for 28 h, followed by lyophilization for 24 h to obtain a dry solid powder, PAA-Ti3C2.

[0061] Polyethylene glycol (PEG) modification: 6 mL of polyethylene glycol (PEG, molecular weight 5000 Da, concentration 20 mg / mL) solution was slowly added to 40 mL of Ti3C2 solution (1 g / L), and the system was sonicated for 6 h under nitrogen protection. Subsequently, the mixture was centrifuged at 5000 rpm for 30 min to remove large particles, and unbound free PEG was removed by dialysis for 24 h. Finally, the resulting mixture was lyophilized for 24 h to obtain a dry solid powder, PEG-Ti3C2.

[0062] Phosphonophosphated polyethylene glycol (PEG-PO3H2) modification: 6 mL of phosphonophosphated polyethylene glycol (PEG-PO3H2, molecular weight 10000 Da, concentration 10 mg / mL) solution was added to 40 mL of Ti3C2 solution (1 g / L) at pH = 4, and the mixture was sonicated for 6 h under nitrogen protection. After the reaction, the mixture was centrifuged at 5000 rpm for 30 min to remove large particles, and unbound free polymer was removed by dialysis for 24 h. Finally, the product was lyophilized for 24 h to obtain a dry powder PEG-PO3H2-Ti3C2.

[0063] according to Figure 2 Compared to Ti3C2, Ti3C2 modified with chitosan (CS), polyacrylic acid (PAA), polyethylene glycol (PEG), and phosphonated polyethylene glycol (PEG-PO3H2) still maintains good dispersibility and hydrophilicity after standing for 1 day.

[0064] Example 3: Specific recognition modification of ultrathin Ti3C2 nanosheets Taking pathogen detection as an example, foodborne pathogens contain abundant polysaccharides (such as lipopolysaccharide LPS, teichoic acid, and saccharide chains) in their cell walls and capsules. These polysaccharides often contain ortho-hydroxyl groups on their monosaccharide units (glucose, galactose, mannose, etc.). When boric acid-modified Ti3C2 comes into contact with the polysaccharides on the bacterial surface, the boric acid groups specifically bind to the cis-diol sites on the sugar chains, capturing or aggregating the bacteria on the nanosheet surface. Specifically, 100 mg of PAA-Ti3C2, 383.4 mg of EDC·HCl, and 230.2 mg of NHS were dissolved in 50 mL of ultrapure water and stirred for 2 h to activate the carboxyl groups. Then, 205.41 mg of 3-aminophenylboronic acid (APBA) was added, and the mixture was stirred at room temperature for 24 h. The product was collected by centrifugation at 10,000 rpm for 8 min and washed three times with ultrapure water to obtain a functionalized APBA-PAA-Ti3C2 biomimetic tongue carrier platform capable of specifically recognizing or capturing bacteria.

[0065] according to Figure 3 Ti3C2 at 3439 cm-1 A broad peak was observed, attributed to the surface -OH groups on the terminal functional group, confirming its pristine MXene properties. After PAA functionalization, peaks of 1720, 1384, and 1168 cm⁻¹ appeared. -1 The characteristic absorption peaks at these locations correspond to the C=O stretching vibration, the COO⁻ symmetric stretching vibration, and the C–O vibration, respectively, indicating the successful introduction of a large number of carboxyl groups onto the MXene surface. Subsequently, APBA grafting introduced peaks at 1617 and 1398 cm⁻¹. -1 The new peak at 1085 cm⁻¹ can be attributed to B–O / B–O–H and aromatic ring vibrations, while at 1085 cm⁻¹ -1 The absorption peaks observed at these locations, originating from C–N or B–O–C bonds, confirm the boric acid modification achieved through EDC / NHS-mediated amide coupling. These results collectively confirm the successful functionalization of PAA-Ti3C2 with APBA.

[0066] Example 4: Preparation of biomimetic taste bud signaling molecules DNA-AgNCs Specifically, ammonium acetate buffer (pH = 6.9) and AgNO3 were precisely added to a test tube containing the oligonucleotides. The final concentrations were: 100 mM ammonium acetate, 500 μM AgNO3, and 50 μM DNA oligonucleotides, respectively. After a brief vortexing, the mixture was incubated at 95 °C for 2 min to dissolve the DNA secondary structure. Subsequently, the sample was directly transferred to ice and frozen for 20 min. NaBH4 at a final concentration of 500 μM was added to reduce the AgNO3. + Finally, after gently stirring the mixture for a period of time, it was placed in a refrigerator for 24 hours, during which time fluorescence was allowed to gradually appear.

[0067] The base sequences of the four DNA strands are as follows: S1: 5'-TATCCGTCCCCCCCCCCACGGATA-3'; S2: 5'-TATCCGTCCCCCCCACGGATA-3'; S3: 5'-TATCCGTCCCCCCCCCCCCACGGATA-3'; S4: 5'-CCCCCCCCCCCC-3'.

[0068] Excitation and emission spectra of four DNA-AgNCs are as follows Figure 4 As shown, under ultraviolet light irradiation, the four DNA-AgNCs emitted green, yellow, orange-red, and dark red fluorescence, respectively, indicating that DNA sequences with different base compositions can be used to prepare DNA-AgNCs with good fluorescence performance, which are suitable as signal molecules for fluorescence sensing arrays.

[0069] Example 5: Fabrication of a fluorescence sensor array based on a biomimetic taste sensing strategy Four DNA-AgNCs were diluted 1:9 with Tris-HNO3 buffer (10 mM, pH = 7.4) to prepare working solutions. Subsequently, different concentrations of APBA-PAA-Ti3C2 were gradually added to the DNA-AgNCs solutions. The four fluorescence emissions of the resulting DNA-AgNCs / APBA-PAA-Ti3C2 mixtures at the optimal excitation wavelengths (412, 471, 564, and 585 nm) were recorded. A biomimetic taste sensing system was constructed when the four fluorescence intensities of the DNA-AgNCs were quenched by approximately 30%.

[0070] according to Figure 5 After adding APBA-PAA-Ti3C2, the fluorescence emission intensity of the four DNA-AgNCs was significantly reduced, indicating that the fluorescence quenching effect was effective. When the concentration was 0.1 mg / mL, the fluorescence intensity of the four DNA-AgNCs was quenched by about 30%, and the fluorescence sensing array of the biomimetic taste system was successfully constructed.

[0071] Example 6: Application of a fluorescence sensing array based on a biomimetic taste sensing strategy in the detection of various pathogenic bacteria Single colonies of *Escherichia coli*, *Staphylococcus aureus*, *Listeria monocytogenes*, *Salmonella typhimurium*, and *Salmonella putrefaciens* grown on solid LB agar plates were selected, inoculated into LB broth, and then cultured overnight at 37 °C. The five pathogenic strains were obtained by centrifugation at 8000 rpm for 10 min. The cultures were then washed three times with Tris-HNO3 (10 mM, pH = 7.4) and resuspended in Tris-HNO3 (10 mM, pH = 7.4) to achieve the concentrations required for subsequent experiments. Binary bacterial mixtures were prepared by uniformly mixing two strains at volume ratios of 1:9, 2:8, 3:7, 4:6, 5:5, 6:4, 7:3, 8:2, and 9:1, with a final concentration of 0.1 OD. 600 Furthermore, by uniformly mixing the samples at volume ratios of 4:3:3, 6:2:2, 8:1:1, 3:4:3, 2:6:2, 4:4:2, 3:3:4, 2:2:6, and 1:1:8, the final concentration was also adjusted to 0.1 OD. 600The 50 μL bacterial suspensions (single-component, two-component, and three-component) prepared above were thoroughly mixed with 150 μL of the 0.1 mg / mL DNA-AgNCs / APBA-PAA-Ti3C2 mixture constructed in Example 5 in black 96-well plates and incubated at 37 °C for 60 min. Subsequently, the fluorescence emission of the DNA-AgNCs / APBA-PAA-Ti3C2 complex was measured before and after the addition of the bacterial suspension, with excitation / emission wavelengths of 412 / 530, 474 / 560, 564 / 630, and 585 / 660 nm, respectively. In subsequent experiments, five independent replicates were performed for each pathogenic bacterium experimental group.

[0072] according to Figure 6 -A, 0.1 OD 600 Different pathogenic bacteria at different concentrations can generate unique fingerprint patterns after responding to the fluorescence sensing array. After dimensionality reduction by LDA (… Figure 6 -B), different types of pathogenic bacteria can form different clusters, thus proving that the prepared fluorescent sensing array can effectively distinguish different types of pathogenic bacteria. Furthermore, not only can it distinguish single-component pathogenic bacteria, but the fluorescent sensing array based on a biomimetic taste sensing strategy can also distinguish two-component ( Figure 6 -C) and three components ( Figure 6 Mixed bacterial solution of -D).

[0073] Example 7: Detection of multiple foodborne pathogens in tap water using a fluorescence sensing array based on a biomimetic taste sensing strategy. Tap water was selected as the actual sample to test the performance of the biomimetic taste sensing system and to evaluate the accuracy of the standard addition method for detecting foodborne pathogens. Specifically, 9 mL of tap water was collected directly from the laboratory pipeline without any pretreatment. Subsequently, different types of bacteria (all at a concentration of 0.1 OD) were added. 600 The bacteria were identified by repeating each species 55 times, thus establishing a training matrix containing 4 fluorescence channels × 5 bacterial species × 55 replicates. KNN, SVM, RF, LR, and CNNs were used to model the collected dataset. The KNN, SVM, RF, and LR models were optimized using 5-fold cross-validation. The CNN models were trained by converting the training matrix (1 × 4) into a fingerprint (2 × 2). The Adam optimizer and class cross-entropy loss function were used to achieve efficient convergence. Accuracy, precision, recall, and F1-score were used to evaluate the performance of the models.

[0074] according to Figure 7As shown in -A, all five machine learning models demonstrated good predictive performance, with the SVM model achieving an accuracy of 97.10% in distinguishing different pathogenic bacteria in tap water. From... Figure 7 -B shows that converting the training matrix (1 × 4) into a fingerprint map (2 × 2) results in a higher accuracy of 98.55% and a precision of 98.33% for CNNs. Furthermore, from... Figure 7 The confusion matrix of -C shows that only one S. typhimurium sample was misclassified as S. aureus. The above results prove that the introduction of deep learning models helps to improve the ability of array sensors to distinguish similar targets.

[0075] Example 8: Detection of multiple foodborne pathogens in beef using a fluorescence sensing array based on a biomimetic taste sensing strategy. For raw beef samples, 25 g of raw beef was mixed with 20 mL of bacterial suspension (OD200). 600 Different concentrations (OD values ​​of 0.5, 1.0, and 2.0) were mixed in 225 mL of Tris-HNO3 buffer (10 mM, pH = 7.4) and homogenized for 5 min. After standing for 30 min to promote precipitation of aggregates, the supernatant was collected using a 0.45 µm sterile filter membrane to remove impurities and large microbial aggregates. Different concentrations (OD values ​​of 0.5, 1.0, and 2.0) were then added to the diluted beef solution. 600 Five bacterial suspensions (E. coli, S. aureus, S. typhimurium, L. monocytogenes, and S. putrefaciens) with concentrations of 0.5, 1.0, and 2.0 were used. Each concentration of each bacterium was tested in five replicates. After the fluorescence sensing array responded, the resulting dataset consisted of 4 fluorescence channels × 5 bacterial suspensions × 3 concentrations × 5 replicates, which were used for LDA analysis. In addition, two blind test groups were prepared for each experimental group, for a total of 30 blind test samples, to evaluate the performance of the fluorescence sensing array in detecting multiple foodborne pathogens in beef.

[0076] In addition to single-component samples, mixed groups of E. coli and S. aureus, S. aureus and S. typhimurium, and L. monocytogenes and S. typhimurium were prepared (mixing ratios of 1:3, 1:1, and 3:1, respectively, with a final concentration of 0.1 OD). 600 After the fluorescence sensing array responded, the resulting dataset consisted of 4 fluorescence channels × 3 groups of mixed bacterial suspensions × 3 mixing ratios × 5 replicates, which were used for LDA analysis. In addition, 2 blind test groups were prepared for each experimental group, for a total of 18 blind test samples, to evaluate the performance of the fluorescence sensing array in detecting multiple foodborne pathogens in beef.

[0077] In addition to the samples mentioned above, three-component mixtures were prepared: E. coli and S. aureus and S. typhimurium (Group A); S. aureus and S. typhimurium and L. monocytogenes (Group B); L. monocytogenes and S. typhimurium and S. putrefaciens (Group C); and E. coli and S. typhimurium and S. putrefaciens (Group D). The mixing ratios were 1:1:8, 3:3:4, and 4:4:2, respectively, with a final concentration of 0.1 OD. 600 After the fluorescence sensing array responded, the resulting dataset consisted of 4 fluorescence channels × 4 groups of three-component mixed bacterial suspensions × 3 mixing ratios × 5 replicates, which were used for LDA analysis. In addition, 2 blind test groups were prepared for each experimental group, for a total of 24 blind test samples, to evaluate the performance of the fluorescence sensing array in detecting multiple foodborne pathogens in beef.

[0078] according to Figure 8 LDA analysis showed that the blind test samples (red dots) had a recognition rate of 100%, and all were located near each category. Therefore, the fluorescence sensing array prepared based on the biomimetic taste sensing strategy can effectively distinguish five pathogenic bacteria at different concentrations in beef samples, as well as detect two-component and three-component mixed bacteria, and has the capability for practical application.

[0079] Comparative Example 1: Detection of 5 pathogenic microorganisms using only four DNA-AgNCs cluster molecules (without Ti3C2 vector) Five 0.1 OD strains were cultured according to the method in Example 6. 600 Suspensions of E. coli, S. aureus, L. monocytogenes, S. typhimurium, and S. putrefaciens were prepared at various concentrations, and four DNA-AgNCs working solutions were prepared according to Example 4. In black 96-well plates, 150 μL of DNA-AgNCs working solution was added to each well, followed by 50 μL of bacterial suspension. The mixture was thoroughly mixed and incubated at 37 °C for 60 min (in the dark). Each experimental group was performed in five independent replicates. After incubation, fluorescence intensity was measured at four channels at excitation / emission wavelengths of 412 / 530, 474 / 560, 564 / 630, and 585 / 660 nm. A dataset of four DNA-AgNCs × five bacterial species × five replicates was constructed and analyzed using LDA to evaluate the ability of DNA-AgNCs to respond to multicolor pathogenic bacteria without competition from two-dimensional nanocarriers.

[0080] according to Figure 9As indicated by -A, when using only four DNA-AgNCs cluster molecules to detect five pathogenic microorganisms, LDA analysis revealed overlap between clusters, making them indistinguishable (0.1 OD). 600 Five pathogenic bacteria at specific concentrations.

[0081] Comparative Example 2: Detection of five pathogenic microorganisms using unfunctionalized Ti3C2 as a carrier To evaluate the effect of using pristine Ti3C2 (unmodified PAA and APBA) as a carrier on DNA-AgNCs signal modulation and bacterial recognition performance, a biomimetic sensor array was prepared by replacing the APBA-PAA-Ti3C2 step described in Example 5 with Ti3C2. The detection steps were performed in Comparative Example 1, except that the working solution was DNA-AgNCs / Ti3C2.

[0082] according to Figure 9 -B, using unfunctionalized Ti3C2 as a carrier to detect five pathogenic microorganisms, LDA analysis revealed overlap between clusters, making them indistinguishable (0.1 OD). 600 Five pathogenic bacteria at specific concentrations.

[0083] Comparative Example 3: Detection of five pathogenic microorganisms using PAA-Ti3C2 as a carrier. To evaluate the impact of polyacrylic acid (PAA) modification of Ti3C2 on detection performance in terms of improved dispersibility and non-specific interactions, relative to the necessity of APBA (boronic acid) specific recognition, a biomimetic sensor array was prepared by replacing the APBA-PAA-Ti3C2 step described in Example 5 with PAA-Ti3C2. The detection steps were performed as in Comparative Example 1, except that the working solution was DNA-AgNCs / PAA-Ti3C2.

[0084] according to Figure 9 -C, using PAA-Ti3C2 as a carrier to detect five pathogenic microorganisms, after LDA analysis, there was overlap between clusters, making them indistinguishable (0.1 OD). 600 Five pathogenic bacteria at specific concentrations.

[0085] Example 9: Application of a fluorescence array based on a biomimetic taste sensing strategy to the detection of other food hazard analytes Heavy metal ion detection: PEG-Ti3C2 obtained in Example 2 was modified with cysteine ​​to specifically recognize Hg. 2+ Cd 2+ Pb 2+ Cu 2+ Zn 2+The method for grafting heavy metal ions is as follows: 0.5 g of L-cysteine ​​was dissolved in 10 mL of deionized water, and the pH was adjusted to approximately 5.5 to maintain adequate ionization of carboxyl / amine groups and prevent oxidation of thiol groups. 6 mL of the cysteine ​​solution was added to 40 mL of a pre-prepared PEG-Ti3C2 dispersion (1.0 mg / mL). The mixture was magnetically stirred at room temperature for 12 h under nitrogen protection to allow cysteine ​​to be stably grafted onto PEG-Ti3C2 via electrostatic adsorption. After the reaction, the mixture was centrifuged at 5000 rpm for 30 min, followed by repeated washing with deionized water to remove free amine groups and small molecules. Finally, the mixture was lyophilized to obtain Cys-PEG-Ti3C2, and a 10 μM Hg solution was prepared as described in Example 6. 2+ Cd 2+ Pb 2+ Cu 2+ Zn 2+ The heavy metal ion solution was prepared, and the sensing and data analysis steps were consistent with those for pathogenic microorganism detection, thereby verifying the possibility of the prepared DNA-AgNCs@Cys-PEG-Ti3C2 simultaneously detecting multiple metal ions.

[0086] Figure 10 -A, LDA analysis showed that the biomimetic taste fluorescence sensing array constructed based on DNA-AgNCs@Cys-PEG-Ti3C2 can effectively distinguish 10 μM Hg. 2+ Cd 2+ Pb 2+ Cu 2+ Zn 2+ The method was validated by the detection of heavy metal ion solutions, demonstrating its scalability to the detection of many other food hazard analytes.

[0087] Mycotoxin Detection: The Chitosan-Ti3C2 obtained in Example 2 was modified with template molecules of molecularly imprinted polymers to specifically recognize mycotoxins such as aflatoxin B1 (AFB1), aflatoxin M1 (AFM1), ochratoxin A (OTA), fumonisin B1 (FB1), and zearalenone (ZEN). The specific method is as follows: 40 ml of 1.0 mg / mL Chitosan-Ti3C2 was prepared as a substrate using 0.1 M pH=7.4 PBS phosphate buffer solution. The target template molecule was dissolved in a small amount of methanol to prepare a 0.5 mM template solution. The final template to Ti3C2 surface molar ratio was approximately 1:50 as needed. Subsequently, the functional monomer methacrylic acid (molar ratio 1:4) and the crosslinking agent ethylene glycol dimethacrylate (molar ratio 10:1) were added to the template solution. The prepared template / functional monomer / crosslinking agent mixture was slowly added to 40 mL of chitosan-Ti3C2 dispersion, and the mixture was gently stirred to allow the template to fully undergo non-covalent interaction with the functional monomer at the chitosan layer interface (and allowed to stand for 60 min to allow pre-binding to occur). Under nitrogen protection, the free radical initiator azobisisobutyronitrile (AIBN, or ammonium persulfate / tetramethylethylenediamine APS / TEMED in the aqueous phase) was added to the system to bring the final initiator concentration to 2.0 wt% (relative to the total monomer amount). After purging with nitrogen for 30 min to replace oxygen, the reaction system was placed at 60 °C for 12 h to polymerize and form a thin molecularly imprinted polymer film in situ on the Chitosan-Ti3C2 surface. After polymerization, the precipitate was collected by centrifugation at 5000 rpm for 20 min. The supernatant was discarded, and the sample was repeatedly washed with methanol / acetic acid (v / v = 9:1) solution (shaking for 10 min each time, at least 5 times) to completely remove the template molecules, thus obtaining molecularly imprinted chitosan-Ti3C2, i.e., MIP-chitosan-Ti3C2. A 10 μM solution of mycotoxins AFB1, AFM1, OTA, FB1, and ZEN was prepared as in Example 6, while the sensing and data analysis procedures were consistent with those for pathogenic microorganism detection, thereby verifying the possibility of simultaneously detecting multiple mycotoxins using the prepared DNA-AgNCs@MIP-chitosan-Ti3C2.

[0088] Figure 10 -B, LDA analysis showed that the biomimetic taste fluorescence sensor array constructed based on DNA-AgNCs@MIP-chitosan-Ti3C2 could effectively distinguish 10 μM mycotoxin solutions of AFB1, AFM1, OTA, FB1, and ZEN, verifying that the method can be extended to the detection of more other food hazard analytes.

[0089] Pesticide residue detection: The PEG-PO3H2-Ti3C2 obtained in Example 2 was modified with acetylcholinesterase (AChE) to specifically recognize organophosphorus pesticides such as trichlorfon, dichlorvos, dimethoate, malathion, and chlorpyrifos. The specific method is as follows: 10 ml of 1.0 mg / mL PEG-PO3H2-Ti3C2 was prepared using 0.1M pH=7.4 PBS phosphate buffer solution. Under ice bath conditions, 0.5 mL of AChE containing 1000 U / mL of active enzyme was slowly added dropwise until the total enzyme activity of the system was approximately 500 U. After addition, the mixture was gently incubated on a shaker at 4 °C for 12 h to facilitate physical adsorption and coordination fixation of the enzyme. To enhance fixation stability, after incubation, glutaraldehyde, an acetaldehyde condensation crosslinking agent, was slowly added to the system to a final concentration of approximately 0.05% (v / v). The reaction was continued at 4 °C for 30 min, followed by blocking the residual aldehyde groups with 0.1 M glycine for 1 h. After the reaction, the precipitate was collected by centrifugation at 5000 rpm for 10 min, the supernatant was discarded, and the mixture was washed three times with PBS buffer to remove unadsorbed free enzyme and crosslinking agent residue. The surface was then blocked with 1% (w / v) BSA at room temperature for 30 min (which can improve resistance to non-specific adsorption and preserve enzyme activity). Excess BSA was removed by centrifugation and the mixture was resuspended in PBS. Finally, the resulting AChE-PEG-PO3H2-Ti3C2 suspension was aliquoted and stored at 4 °C. The organophosphorus pesticide solutions of trichlorfon, dichlorvos, dimethoate, malathion, and chlorpyrifos were prepared according to Example 6, and the sensing and data analysis steps were consistent with those for pathogenic microorganism detection, thereby verifying the possibility of the prepared DNA-AgNCs@AChE-PEG-PO3H2-Ti3C2 to simultaneously detect multiple organophosphorus pesticides.

[0090] Figure 10 -C, according to LDA analysis, the biomimetic taste fluorescence sensing array constructed based on DNA-AgNCs@AChE-PEG-PO3H2-Ti3C2 can effectively distinguish 10 μM organophosphorus pesticide solutions of trichlorfon, dichlorvos, omethoate, malathion, and chlorpyrifos, verifying that the method can be extended to the detection of more other food hazard analytes.

[0091] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the embodiments described above. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for constructing a fluorescence array sensing system based on biomimetic taste, characterized in that, Includes the following steps: (1) Fabrication of bionic tongue recognition element: Using MAX phase precursor as raw material, the A element in MAX phase is removed by chemical etching to obtain multilayer Ti3C2Tx MXene nanosheets with functional groups on the surface; the multilayer Ti3C2Tx MXene nanosheets are intercalated and the intercalated multilayer Ti3C2Tx MXene nanosheets are peeled into single-layer or few-layer ultrathin nanosheets by ultrasonic peeling, mechanical shearing, stirring peeling or electrostatic peeling to obtain bionic tongue recognition element; (2) Functionalization of bionic tongue recognition element: The monolayer Ti3C2Tx MXene nanosheets prepared in step (1) are functionalized to achieve specific recognition of various hazards and improve the dispersibility, anti-pollution and stability of the carrier in complex matrices; the hazards are selected from at least one of metal ions, foodborne pathogens, pesticide residues, fungal toxins and antibiotics; (3) Preparation of biomimetic taste bud cell signaling elements: Oligonucleotide DNA with different structures is used as a template. The DNA structure includes single strand, stem-loop, quadruplex or other secondary structures. The DNA regulates the multicolor fluorescence emission of silver nanoclusters by adjusting the base composition, length and structure. In a pH buffer system, the DNA template is mixed with silver ions, and borate reducing agent is added. Silver nanoclusters are generated under mild conditions to obtain DNA-AgNCs. (4) Construction of a biomimetic taste-based fluorescence array sensing system: The DNA-AgNCs prepared in step (3) were diluted to prepare a working solution. The functionalized Ti3C2Tx MXene nanosheets prepared in step (2) were added to the DNA-AgNCs working solution and mixed. The optimal emission peak intensity of DNA-AgNCs was measured under the optimal excitation wavelength. When the fluorescence intensity decreased to 25% ~ 40% from the initial value, the construction of the biomimetic taste-based fluorescence array sensing system was completed. The binding constant between the two was calculated using the Stern-Volmer equation to evaluate the difference in binding performance of different DNA templates. The functionalization of the identification element in step (2), wherein the functionalization modification includes: (a) Modification of hydrophilicity and antifouling properties: The hydrophilicity and antifouling ability of Ti3C2Tx MXene nanosheets are improved by physical adsorption, including one or more of polyacrylic acid, polyethylene glycol, phosphonated polyethylene glycol, and chitosan; the concentration of the polyacrylic acid monomer is 0.05 ~ 1.0 M, the reaction temperature is 0 ~ 80 ℃, and the reaction time is 2 ~ 24 h; the molecular weight of the polyethylene glycol is 500 ~ 20000 Da, the concentration is 0.1 ~ 50 mg / mL, the temperature is 0 ~ 80 ℃, and the reaction time is 0.5 ~ 48 h; the molecular weight of the phosphonated polyethylene glycol backbone is 500 ~ 10000 Da, the concentration is 0.01 ~ 20 mM, the pH is 1.5 ~ 8.0, and the temperature is 0 ~ 80 ℃; the molecular weight of the chitosan is 50000 ~ 500000. Da, with a degree of deacetylation of 70-95%, a solution concentration range of 0.01-5.0 wt%, a pH of 4.0-60 ℃, and a reaction time of 0.5-48 h; (b) Metal ion recognition modification: Ligands containing thiol, amino, or phosphate groups are introduced onto the surface of Ti3C2Tx MXene nanosheets to chelate Hg-selective ions. 2+ Pb 2+ Cd 2+ Cu 2+ Fe 2+ Zn 2+ Mn 2+ As 3+ The metal ions; the ligand is at least one of mercaptoacetic acid, cysteine, ethylenediaminetetraacetic acid, and phosphorylated polyethylene glycol, which is modified on the surface of Ti3C2Tx MXene nanosheets by covalent bonding, electrostatic adsorption or π–π interaction; (c) Foodborne pathogen recognition modification: Specific recognition molecules are immobilized on the surface of Ti3C2Tx MXene nanosheets. The recognition molecules are selected from at least one of monoclonal antibodies, peptides, aptamers, phage display peptides, lectins, and phenylboronic acid, and are used to bind to Staphylococcus aureus, Escherichia coli O157:H7, Salmonella, Listeria, or Pseudomonas aeruginosa. The immobilization method is glutaraldehyde crosslinking, EDC / NHS activation coupling, maleimide-thiol coupling, or click chemistry reaction. (d) Pesticide residue recognition modification: Specific enzymes or aptamers are covalently coupled or physically adsorbed onto the surface of Ti3C2Tx MXene nanosheets. The enzymes are selected from acetylcholinesterase, peroxidase, or urease, and are used to recognize organophosphates, carbamates, pyrethroids, or glyphosate pesticides. The aptamers are fixed to the surface of Ti3C2Tx MXene nanosheets through π–π interactions or hydrogen bonding. (e) Mycotoxin Recognition Modification: Aptamers, antibodies, molecularly imprinted polymers, or cyclodextrin derivatives are introduced onto the surface of Ti3C2Tx MXene nanosheets to achieve specific recognition of aflatoxin B1, aflatoxin M1, ochratoxin A, fumonisin B1, and zearalenone; wherein the aptamer sequence length is 20-80 bases, and may contain stem-loop, G-quadruplex, or hairpin structures; the binding constant Ka between the template molecule of the molecularly imprinted polymer and the target mycotoxin is ≥ 10. 5 M -1 Furthermore, the coating thickness is 5 to 50 nm to ensure high selectivity and stability in complex food matrices; (f) Antibiotic recognition modification: Antibiotic aptamers, molecularly imprinted polymers or metal-organic framework coatings are immobilized on the surface of Ti3C2Tx MXene nanosheets for specific binding of tetracyclines, fluoroquinolones, macrolides or β-lactams; the immobilization methods are surface in-situ polymerization, sol-gel method or layer-by-layer self-assembly method.

2. The method for constructing a biomimetic taste-based fluorescence array sensing system according to claim 1, characterized in that, The chemical formula of the MAX phase mentioned in step (1) is M. n+1 AX n Where M is a transition metal element, A is a group A element, X is carbon or nitrogen, and n = 1 ~ 3; the etchant is a fluoride-containing solution, including one or more of hydrofluoric acid, a mixture of lithium fluoride and hydrochloric acid, a mixture of ammonium fluoride and hydrochloric acid, ammonium bifluoride solution, molten salt or alkaline etching system, with an etching temperature of 20 ~ 60℃ and an etching time of 12 ~ 72 h; the intercalating agent includes one or more of dimethyl sulfoxide, tetrabutylammonium hydroxide, tetramethylammonium hydroxide, lithium chloride, sodium chloride, surfactant or small molecule organic amine, to increase the interlayer spacing, with an intercalation time of 12 ~ 48 h and an intercalation temperature of 20 ~ 60 ℃; the ultrasonic stripping conditions are power 10 ~ 200 W, time 1 ~ 60 min, and temperature 0 ~ 40 ℃; the mechanical shearing stripping speed is 1000 ~ 20000 rpm and time is 1 ~ 60 min; the stirring intercalation stripping speed is 200 ~ 2000 rpm. rpm, time 6 ~ 72 h, temperature 0 ~ 60 ℃; electrostatic stripping voltage -1.0 ~ +5.0V or current density 0.1 ~ 10 mA·cm -2 The time ranges from 1 minute to 6 hours.

3. The method for constructing a biomimetic taste-based fluorescence array sensing system according to claim 1, characterized in that, The MAX phase mentioned in step (1) is Ti3AlC2, Ti2AlC, Ti3SiC2 or Ti2AlN.

4. The method for constructing a biomimetic taste-based fluorescence array sensing system according to claim 1, characterized in that, In step (3), the preparation of the biomimetic taste bud cell signaling element involves a DNA template with a sequence length of 10 to 40 nucleotides and a stem-loop structure. The stem contains a base pairing to form a stable double strand, and the loop contains a sequence rich in cytosine or guanine. The borate reducing agent is selected from sodium borate, sodium borohydride, or boric acid. The reduction reaction is carried out in a phosphate, tris(hydroxymethyl)aminomethane, citrate, and acetate buffer solution at pH 7.0 to 10.0, at a temperature of -20°C to 37°C, for a reaction time of 0.5 to 48 h. In step (4), the biomimetic taste fluorescence array sensing system is constructed as follows: the initial concentration of DNA-AgNCs is 5-50 mM; the dilution buffer is 5-20 mM, pH 7.0-10.0, consisting of phosphate buffer, tris(hydroxymethyl)aminomethane buffer, citrate buffer, or acetate buffer, with a dilution volume ratio of 1:3-1:20, resulting in a final working solution concentration of 0.05-5 μM; the concentration of the functionalized Ti3C2Tx MXene nanosheet dispersion is 0.01-1.0 mg / mL; the volume ratio of the DNA-AgNCs working solution to the functionalized Ti3C2Tx nanosheet dispersion is 1:1-10:1; the mixing is performed by incubation on a shaker at 30-50 °C, 200-800 rpm, and 5-60 °C. min to ensure thorough mixing; when the binding constant between DNA-AgNCs and functionalized Ti3C2Tx nanosheets is calculated to be 10~40 using the Stern-Volmer equation, it indicates that the DNA-AgNCs synthesized from this DNA template are suitable for preparing fluorescent array sensing systems.

5. A biomimetic taste-based fluorescent array sensing system constructed by the method described in any one of claims 1-4.

6. The application of the biomimetic taste-based fluorescence array sensing system as described in claim 5 in the detection of hazardous substances in food, characterized in that, Includes the following steps: (1) Preparation of standard solutions for food hazard analytes: Select at least one of heavy metal ions, pesticide residues, mycotoxins, foodborne pathogens and antibiotics as target analytes, and prepare single-system or multi-component mixed-system analyte solutions with different concentration gradients. (2) Construction of the mixed system: The working solution of each channel of the biomimetic taste-based fluorescence array sensing system described in step (4) of claim 1 is dispensed into a microplate, and the analyte solution described in step (1) is added to each. The system is incubated under constant temperature conditions to allow it to fully combine, and the fluorescence emission intensity at the optimal excitation / emission wavelength of each channel is measured. (3) Data modeling and pattern recognition: The fluorescence response matrix is ​​used for pattern recognition and classification by machine learning algorithm, the model parameters are optimized by optimization algorithm, and the model performance is analyzed based on evaluation index to verify the sensor system’s ability to identify different analytes and its stability. (4) Blind sample and actual sample detection: The above sensor system is applied to the detection of blind samples and actual samples, and the above machine learning model is used to predict and classify the samples to evaluate the applicability of the sensor system in the detection of food hazards.

7. The application of a biomimetic taste-based fluorescence array sensing system according to claim 6 in the detection of hazardous substances in food, characterized in that, In step (1), the target analyte is a heavy metal ion selected from Hg. 2+ Pb 2+ Cd 2+ Cu 2+ Fe 2 + Zn 2+ Mn 2+ Or As 3+ The concentration gradient range is 0–100 μM; the pesticide residues are selected from organophosphates, carbamates, pyrethroids, or glyphosate, with a concentration gradient range of 0–100 μM; the mycotoxins are selected from aflatoxin B1, ochratoxin A, fumonisin B1, or vomitoxin, with a concentration gradient range of 0–100 μM; the foodborne pathogens are selected from Staphylococcus aureus, Escherichia coli O157:H7, Salmonella, Listeria, or Pseudomonas aeruginosa, with a concentration gradient in the range of OD. 600 The characterization is 0 ~ 0.5; the antibiotics are selected from tetracyclines, fluoroquinolones, macrolides or β-lactams, with a concentration gradient range of 0 ~ 100 μM; wherein the two-component mixing system is prepared at equal concentrations and in volume ratios of 1:9, 2:8, 3:7, 4:6, 5:5, 6:4, 7:3, 8:2, 9:1; the three-component mixing system is prepared at equal concentrations and in volume ratios of 4:3:3, 6:2:2, 8:1:1, 3:4:3, 2:6:2, 4:4:2, 3:3:4, 2:2:6, 1:1:

8.

8. The application of a biomimetic taste-based fluorescence array sensing system according to claim 6 in the detection of hazardous substances in food, characterized in that, In step (2), the volume of working solution added to each fluorescence channel is 50-150 μL, the volume of analyte solution added is 20-50 μL, the incubation temperature is 20-40℃, and the rotation speed is 200-300 rpm.

9. The application of a biomimetic taste-based fluorescence array sensing system according to claim 6 in the detection of hazardous substances in food, characterized in that, In step (3), the machine learning algorithm selected is autonomous component analysis, hierarchical clustering analysis, linear discriminant analysis, K-nearest neighbors, support vector machine, random forest, logistic regression, convolutional neural network, recurrent neural network, Transformer or long short-term memory network; parameter optimization adopts at least one of cross-validation, grid search, Bayesian optimization, genetic algorithm optimization, adaptive gradient optimization, learning rate scheduling, Dropout regularization or L1 / L2 regularization; model evaluation metrics include accuracy, precision, recall, F1-score and confusion matrix.

10. The application of a biomimetic taste-based fluorescence array sensing system according to claim 6 in the detection of hazardous substances in food, characterized in that, In step (4), the actual samples can be fruit juice beverages, milk and dairy products, fruits and vegetables, meat and meat products, aquatic products and their products, grains and grain products, or beverages; sample pretreatment includes dilution, homogenization, settling, and filtration; the blind sample and the actual sample are added to the analyte from step (1) at a preset target level, in μM or OD. 600 The characterization is performed, and the incubation and determination are carried out under the conditions of step (2). The prediction and classification are completed by the model established in step (3).

Images