A machine learning driven sensor array for rapid identification of multiple heavy metal ions

Abstract

This invention belongs to the field of food safety and environmental monitoring technology, and discloses a method for rapid identification of multiple heavy metal ions using a machine learning-driven sensor array. The main steps include: (1) preparing trimetallic nanozymes PVP-PtCuAu NCs that simultaneously possess oxidase-like activity, peroxidase activity, and laccase activity; (2) constructing a three-channel nanozyme sensor array based on the three enzyme activities of PVP-PtCuAu NCs; (3) using the nanozyme sensor array to achieve rapid identification of multiple typical heavy metal ions; (4) integrating machine learning algorithms and the nanozyme sensor array to achieve rapid identification and accurate prediction of multiple heavy metal ions, and successfully applying it to the detection of actual water samples. This invention constructs a nanozyme sensor array integrating machine learning algorithms, which has advantages such as rapid detection, simple operation, and suitability for on-site analysis, breaking through the bottleneck of simultaneous rapid detection of multiple heavy metal ions and possessing good application prospects.

Description

Technical Field

[0001] This invention belongs to the field of food safety and environmental monitoring technology, and relates to a method for rapid identification of various heavy metal ions using a machine learning-driven sensor array. Background Technology

[0002] Heavy metal ion pollution is characterized by its wide range of sources, high toxicity, and difficulty in degradation. It can enter the human body through water, soil, and the food chain, producing cumulative toxicity. Unlike organic pollutants, metal ions cannot be decomposed through chemical or biological processes and easily accumulate in organisms, posing a continuous and serious threat to ecosystems. Therefore, establishing a rapid, accurate, and portable detection method is of great significance.

[0003] Current methods for heavy metal detection mainly rely on large, precision instruments, resulting in complex, costly, and unsuitable for rapid on-site analysis. Colorimetric analysis has gained attention due to its ease of operation and intuitive results; however, traditional single-signal colorimetric systems suffer from poor selectivity and weak anti-interference capabilities, making simultaneous identification of multiple heavy metal ions difficult. While colorimetric detection of natural enzymes exhibits some selectivity, its tolerance to extreme conditions is weak, hindering practical applications in on-site detection. Nanozymes, with their advantages of structural stability, ease of preparation, tunable catalytic activity, and strong anti-interference capabilities, have become important alternatives to natural enzymes. In particular, multi-metal nanozymes, through synergistic effects between components, can simultaneously exhibit multiple enzyme activities in a single material, providing multiple catalytic response sites, which is beneficial for constructing detection systems with high stability and multidimensional signal output.

[0004] Sensor array technology, by constructing multiple sensing units with differentiated response characteristics to form a cross-reaction pattern, can acquire the characteristic comprehensive color signal of the target analyte, thereby achieving efficient differentiation and identification of substances with similar structures or physicochemical properties. This strategy does not rely on highly specific recognition elements, but rather combines pattern recognition methods to significantly improve the selectivity and reliability of the detection system. Simultaneously, machine learning models can mine potential features from high-dimensional, nonlinear data and achieve efficient analysis of complex signals, thereby improving the identification accuracy and detection efficiency of various heavy metal ions, and enhancing the stability and accuracy of the detection system. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting multiple heavy metal ions in water that is fast, stable, low-cost, and suitable for on-site analysis. It also integrates a smartphone platform to achieve intelligent analysis and classification of the detection results, thereby overcoming the shortcomings of existing detection technologies.

[0006] Specifically, this project constructs a colorimetric sensor array detection system based on a smartphone platform to achieve rapid identification and quantitative analysis of multiple heavy metal ions. The method uses PVP-PtCuAu NCs trimetallic nanozymes as colorimetric probes, constructing three simulated colorimetric channels—peroxidase-like, oxidase-like, and laccase-like—to react with target heavy metal ions. During the catalytic process, the oxidation of the TMB substrate and the coupled oxidation of 2,4-DCP and 4-APP occur, generating a clear colorimetric signal. Subsequently, by combining a colorimetric sensor array, smartphone image acquisition technology, and machine learning algorithms, rapid detection of multiple heavy metal ions is achieved, and the method has been successfully applied to real environmental samples, providing a new technical approach for constructing portable and intelligent methods for detecting multiple heavy metals.

[0007] The present invention employs the following steps:

[0008] A machine learning-driven sensor array method for rapid identification of multiple heavy metal ions includes the following steps:

[0009] (1) Preparation of PVP-coated trimetallic nanozymes PVP-PtCuAu NCs:

[0010] At room temperature, aqueous solutions of chloroplatinic acid hexahydrate (H₂PtCl₆·6H₂O), copper sulfate pentahydrate (CuSO₄·5H₂O), chloroauric acid (HAuCl₄), and polyvinylpyrrolidone (PVP) were mixed thoroughly. The mixture was then stirred vigorously at a certain temperature, and ascorbic acid solution was added as a reducing agent. The reaction was continued for a period of time. After the reaction was completed, the product was collected by centrifugation and washed several times with deionized water. Finally, PVP-coated PtCuAu trimetallic nanozymes, namely PVP-PtCuAu NCs, were obtained by freeze-drying.

[0011] (2) Construction of the colorimetric sensor array:

[0012] A multi-channel detection mode with cross-response characteristics was constructed by setting different chromogenic substrate systems; among them...

[0013] A chromogenic substrate system consisting of TMB solution and H2O2 solution was used as a peroxidase-like chromogenic channel. After adding acetate buffer, chromogenic reaction solution A was obtained.

[0014] Using TMB solution as a single chromogenic substrate system as an oxidase-like chromogenic channel, and adding acetate buffer, chromogenic reaction solution B was obtained.

[0015] The chromogenic substrate system of the coupling reaction of 2,4-dichlorophenol solution (2,4-DCP) and 4-aminoantipyrine solution (4-APP) was used as the laccase-like chromogenic channel. After adding MES buffer, the chromogenic reaction solution C was obtained.

[0016] (3) Detection of multiple heavy metal ions based on colorimetric sensor array:

[0017] The PVP-PtCuAu NCs probe obtained in step (1) was added to the test system containing different target heavy metal ions and incubated at room temperature to allow the heavy metal ions to interact with the nanozymes, thereby regulating their enzyme-like catalytic activity.

[0018] Next, after each heavy metal ion test system is incubated, it is dropped into the colorimetric reaction solutions A, B and C prepared in step (2) to react. Different color responses are generated in different colorimetric systems of the sensor array, forming a characteristic comprehensive chromatographic response. After the reaction is completed, the resulting reaction solution is transferred to a 96-well microplate for unified colorimetric reading and recording. With the assistance of a color recognition APP, the corresponding RGB values ​​are extracted to establish a color feature data matrix. Subsequently, the data matrix is ​​input into a machine learning model for training and prediction to realize the identification of target species and quantitative analysis of concentration.

[0019] (4) Determination of heavy metal ions in actual water samples

[0020] After collecting actual environmental water samples, the target heavy metal ions are added to the sample system, and the detection process of step (3) is repeated to obtain the RGB color signal of the actual water sample and input it into the machine learning model to obtain the type and concentration of heavy metal ions in the actual sample.

[0021] (5) Identification of multi-component heavy metal ion mixtures

[0022] Different metal ions were combined into binary, ternary and quaternary mixed systems in different molar ratios to simulate mixed samples under different pollution composition conditions. The detection process of step (3) was repeated to obtain the RGB color signals of different mixed systems. With the help of the powerful multidimensional signal processing capabilities of the machine learning model, the mixing situation of various heavy metal mixtures was obtained.

[0023] In step (1), the ratio of the amounts of H2PtCl6·6H2O, CuSO4·5H2O, HAuCl4, PVP aqueous solution and ascorbic acid aqueous solution is 3 mL: 3 mL: 3 mL: 15 mL: 3 mL; wherein, the concentration of H2PtCl6·6H2O is 5 mM, the concentration of CuSO4·5H2O is 5 mM, the concentration of HAuCl4 is 5 mM, the concentration of PVP aqueous solution is 1 mg / mL, and the concentration of ascorbic acid aqueous solution is 100 mM.

[0024] In step (1), the reaction temperature is 30 °C and the reaction time is 2 h.

[0025] In step (2),

[0026] In the colorimetric reaction solution A, the volume ratio of TMB solution, H2O2 solution, and acetate buffer is 0.1 mL: 6.4 μL: 2 mL.

[0027] In the colorimetric reaction solution C, the volume ratio of 2,4-DCP solution, 4-APP solution and MES buffer is 1 mL: 1 mL: 5 mL.

[0028] The concentrations of the TMB solution were 2 mg / mL, the H2O2 solution was 0.75%, the acetate buffer was 0.2 mol / L, the 2,4-DCP solution was 6 mM, the 4-APP solution was 6 mM, and the MES buffer was 60 mM. The pH of the acetate buffer was adjusted to 4, and the pH of the MES buffer was adjusted to 7.

[0029] In step (3), the color recognition program can accurately acquire images and convert optical signals into values ​​for the Red, Green and Blue channels.

[0030] In step (3), the machine learning model can process the color feature data matrix generated by the color recognition program and predict the category and concentration of the target object according to the qualitative model and the quantitative model, respectively.

[0031] In step (4), the actual environmental water sample is taken from the water sample of the lower reaches of the Yangtze River or the tap water of the main campus of Jiangsu University as the representative actual sample matrix.

[0032] In step (5), the selected proportions of the multi-component heavy metal ion mixtures with different molar ratios need to cover a series of relative concentration changes from single components to multiple proportions coexisting, so as to comprehensively reflect the compositional differences of the multi-component mixture system.

[0033] Compared with existing technologies, the advantages of this invention are:

[0034] (1) This invention utilizes multi-metal nanozymes to replace natural enzymes as catalytic materials, overcoming the problems of easy inactivation, harsh storage conditions, and short service life of natural enzymes. This allows the detection system to maintain stable catalytic performance over a wide temperature and pH range, making it suitable for rapid analysis of samples in complex environments. It has higher practicality and durability.

[0035] (2) The present invention uses PtCuAu trimetallic nanozymes as a single catalytic material. Through the synergistic effect of multiple metals, oxidase-like, peroxidase-like and laccase-like activities are achieved simultaneously in the same nanostructure. This avoids the complicated process of preparing multiple functional materials separately in traditional sensor arrays, significantly improves the simplicity, consistency and repeatability of system construction, and enhances the catalytic stability and environmental adaptability of the material.

[0036] (3) The present invention constructs a three-channel colorimetric sensor array, which generates cross-response to heavy metal ions through different color development systems, forming characteristic color fingerprint information. Compared with the single signal detection method, it can effectively improve the ability to distinguish multiple target ions, reduce the impact of background interference on the detection results, and thus significantly improve the selectivity and reliability of detection.

[0037] (4) This invention introduces a smartphone as a signal acquisition terminal, directly converting the colorimetric results into RGB digital information, enabling portable reading of the detection signal. Combined with Origin's built-in LDA and HCA machine learning algorithms, it analyzes multidimensional RGB data, extracting effective features from complex nonlinear response signals. This significantly improves the ability to distinguish and identify multiple target heavy metal ions, and establishes a nonlinear mapping relationship between the response signal and the target concentration, achieving high-precision quantitative prediction of the target analyte. Compared with traditional univariate analysis methods, this invention does not rely on human experience for complex data interpretation. It has advantages such as simple structure, high stability, rich detection dimensions, strong identification ability, high portability, higher automation level and data processing efficiency, effectively shortening detection time and reducing human error. It has good application prospects in environmental monitoring and food safety fields. Attached Figure Description

[0038] Figure 1 A schematic diagram of a rapid detection method for multiple heavy metal ions integrated into a smartphone platform;

[0039] Figure 2 (A) SEM image of PVP-PtCuAu NCs; (B) TEM image of PVP-PtCuAu NCs;

[0040] Figure 3(A) XPS full spectrum of PVP-PtCuAu NCs; (B) XRD image of PVP-PtCuAu NCs; (C) Fourier transform infrared spectrum of PVP-PtCuAu NCs.

[0041] Figure 4 Validation of the catalytic performance of three enzymes in PVP-PtCuAu NCs: (A) peroxidase; (B) laccase; (C) oxidase.

[0042] Figure 5 (A) The effect of different pH values ​​on the catalytic performance of the three enzymes; (B) The effect of reaction time on the catalytic performance of the three enzymes;

[0043] Figure 6 Steady-state kinetic analysis of the activities of three enzymes in PVP-PtCuAu NCs, where: (AC) are the Michaelis-Menten kinetic curves of the peroxidase-like, oxidase-like, and laccase-like catalytic systems with H2O2, TMB, and 2,4-DCP as substrates, respectively; (DF) are the Lineweaver-Burk double reciprocal fitting curves of the corresponding systems, respectively.

[0044] Figure 7 The response analysis results of the colorimetric sensor array to five metal ions (5 μM) include: (A) heatmap; (B) fingerprint spectrum; (C) radar chart; (D) typical LDA score map; (E) HCA cluster map; (F) three-dimensional typical LDA score map;

[0045] Figure 8 Colorimetric sensor array for Pb 2+ Ag + As 3+ The quantitative analysis results include: typical LDA score diagrams (A, D, G) for target ions at different concentrations, HCA cluster diagrams (B, E, H) and relationship diagrams between LDA factor 1 and metal ion concentration (C, F, I).

[0046] Figure 9 The colorimetric sensor array identifies metal ion mixtures and real water samples, including typical LDA score maps of target metal ions in binary (A and D), ternary (B), quaternary mixture (C), river water (E), and tap water samples (F). Detailed Implementation

[0047] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. These embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and processes to facilitate understanding of the technical features of the present invention. They do not constitute any limitation on the scope of protection of the present invention. All technical solutions formed by equivalent transformations or substitutions fall within the scope of protection of the present invention.

[0048] Example 1: A colorimetric sensor array integrated with a smartphone platform for rapid detection of various heavy metal ions

[0049] (1) Preparation of PVP-coated PtCuAu trimetallic nanozymes

[0050] At room temperature, 3 mL (5 mM) of aqueous solution of chloroplatinic acid hexahydrate (H₂PtCl₆·6H₂O), 3 mL (5 mM) of aqueous solution of copper sulfate pentahydrate (CuSO₄·5H₂O), 3 mL (5 mM) of aqueous solution of chloroauric acid (HAuCl₄), and 15 mL (1 mg / mL) of polyvinylpyrrolidone (PVP) were mixed thoroughly. The mixture was then stirred vigorously at 30 °C, and ascorbic acid solution (3 mL, 100 mM) was added as a reducing agent. The reaction was continued with vigorous stirring for 2 h. After the reaction was complete, the product was collected by centrifugation and washed several times with deionized water to remove unreacted substances and impurities. Finally, the product was freeze-dried to obtain PVP-coated PtCuAu trimetallic nanozyme, which was stored at 4 °C for later use.

[0051] Results analysis:

[0052] from Figure 2 (A) SEM and Figure 2 The TEM image in (B) shows that the prepared trimetallic nanomaterial is assembled from multiple small-sized nanoparticles, exhibiting a mulberry-like cluster structure with a small particle size of about 100 nm, surrounded by irregular shadows, which may be related to the coating of PVP.

[0053] also, Figure 3 The XPS full spectrum in (A) shows that the two sets of peaks at 75.3 eV / 72.0 eV and 74.5 eV / 71.1 eV correspond to Pt 4f, respectively. 5 / 2 and Pt 4f 7 / 2 The spin-orbit splitting signal of the orbit indicates that Pt in the surface material is not in a single valence state, but rather contains metallic Pt. 0 and partially oxidized Pt 2+ A pair of typical double peaks appear at 87.7 eV and 84.0 eV, corresponding to Au 4f, respectively. 5 / 2 and Au 4f 7 / 2Energy levels, which are consistent with the metallic Au state in the literature. 0 The characteristic positions are basically consistent, indicating that gold mainly exists in the zero-valent form. The characteristic peaks appearing near 951.4 eV and 931.5 eV belong to Cu 2p, respectively. 1 / 2 and Cu 2p 3 / 2 The orbitals indicate that copper exists primarily in a low valence state. These characterizations provide evidence for the successful preparation of PVP-PtCuAu NCs. Figure 3 XRD analysis in (B) reveals a broad diffraction peak at approximately 38°, which can be attributed to the (111) crystal plane of the face-centered cubic (fcc) noble metal structure. Another weaker diffraction peak appears at approximately 64°, corresponding to the (220) crystal plane. These diffraction peaks fall within the standard spectra range for Au (ICDD PDF#98-000-0230), Cu (ICDD PDF#98-000-0172), and Pt (COD ID#1011111), indicating the successful fabrication of a multi-metal nanoparticle cluster. Figure 3 (C) Fourier transform infrared (FTIR) spectroscopy analysis shows that at approximately 3675 cm⁻¹ -1 The absorption peak appearing at 2902 cm⁻¹ can be attributed to the O–H stretching vibration; -1 The absorption peaks in the vicinity mainly originate from C–H stretching vibrations; at approximately 1230 cm⁻¹ -1 The absorption peak at 1060 cm⁻¹ is usually associated with C–N or C–O vibrations; while the absorption peak at 1060 cm⁻¹ is associated with C–N or C–O vibrations. -1 The nearby characteristic peaks can be attributed to C–O stretching vibrations. Compared with the infrared spectrum of pure PVP, the intensity or position of some characteristic peaks have changed, indicating that PVP molecules may interact with the surface of metal nanoparticles through coordination or electrostatic interactions, thereby achieving effective stabilization and coating of the nanoparticles.

[0054] (2) Validation and optimization of enzyme-like catalytic and colorimetric properties of PVP-PtCuAu NCs

[0055] Results analysis:

[0056] The enzyme-like catalytic activity of PVP-PtCuAuNCs is crucial for the detection of target analytes. We evaluated the peroxidase, oxidase, and laccase-like activities of the material in detail. Upon addition of this material to a TMB and H2O2 coexisting system, the initially colorless solution rapidly transformed into a distinct blue solution because PVP-PtCuAu NCs catalyzed the decomposition of H2O2 to generate reactive oxygen species, which further oxidized TMB. Even without the addition of H2O2, the addition of PVP-PtCuAu NCs to the TMB system still resulted in a gradual blue solution with corresponding absorption peaks, indicating that the material can directly activate dissolved oxygen and catalyze the oxidation of the substrate TMB, exhibiting oxidase-like activity. Further verification of its laccase-like activity was conducted using a 2,4-DCP and 4-APP coupled colorimetric system. After the addition of PVP-PtCuAu NCs, the system gradually generated a red quinone imine product, which produced a characteristic absorption signal in the visible light region, while the blank system showed no obvious color development. This demonstrates that the nanozyme can catalyze the single-electron oxidation and coupling reactions of phenolic substrates, exhibiting laccase-like catalytic function. Figure 4 (A), (B), and (C) show the absorbance curves of PVP-PtCuAu NCs added to the three colorimetric systems, respectively. The maximum absorption peaks are generated at 652 nm and 510 nm, respectively, indicating that they can induce colorimetric signals. Figure 5 (A) and (B) respectively tested the absorbance at different pH and reaction conditions to select the optimal reaction environment and reaction time.

[0057] (3) Steady-state dynamic analysis of PVP-PtCuAu NCs

[0058] Results analysis:

[0059] To further evaluate the catalytic performance of nanozymes, the steady-state kinetics of the activities of three enzymes in their respective substrate systems were analyzed using the Michaelis-Menten equation and the Lineweaver-Burk double reciprocal plot method. For peroxidase activity, its kinetic parameters were determined using different concentrations of H₂O₂ as substrates, such as... Figure 6 As shown in (A, D), the Michaelis-Menten constant (K) is calculated. m ) and maximum initial reaction rate (V max The concentrations were 3.299 mM and 0.163 μM, respectively. -1 ·s. Higher K m This indicates that the material exhibits a relatively low affinity for H2O2, but its high V... maxThis indicates that once H2O2 is activated, it can rapidly participate in the reaction to generate reactive oxygen species, thereby accelerating the oxidation process of TMB. For oxidase-like activity, kinetic parameters were determined using TMB as a substrate, such as... Figure 6 As shown in (B, E), K is calculated. m and V max The values ​​were 0.7021 mM and 0.0224 μM, respectively. -1 Compared to the peroxidase-like system, this system exhibits a lower catalytic rate, indicating that under conditions without added H2O2, the oxidation process of TMB is mainly limited by the activation efficiency of dissolved oxygen, leading to a decrease in overall catalytic efficiency. However, it still exhibits a high Vo. max This indicates that the material possesses strong oxygen molecule activation ability and exhibits good oxidase-like catalytic performance. For laccase-like activity, kinetic parameters were determined using 2,4-DCP as a substrate, such as... Figure 6 As shown in (C, F), K m and V max The concentrations were 0.07249 mM and 0.00369 μM, respectively. -1 ·s. This system has the lowest K m The value indicates that the material has a high substrate affinity for 2,4-DCP, but its V... max The significantly lower values ​​compared to the two aforementioned enzyme-like systems indicate that while the substrates are easily adsorbed and recognized, the subsequent catalytic conversion rate is relatively slow. PVP-PtCuAu NCs exhibited significantly differentiated catalytic response strengths and kinetic behaviors across the three substrate systems, which is beneficial for constructing sensor arrays with differentiated output characteristics. This can provide rich and distinguishable response information for the subsequent identification and quantitative analysis of various metal ions, thereby improving the identification accuracy and analytical reliability of the detection system.

[0060] (4) Construction of qualitative and quantitative models based on machine learning algorithms

[0061] This implementation utilizes the built-in Linear Discriminant Analysis (LDA) and Hierarchical Cluster Analysis (HCA) in Origin to perform qualitative classification and quantitative regression analysis of target metal ions. This compensates for the shortcomings of sensor arrays in analyzing high-dimensional, nonlinear response signals and the problem of slow detection, thereby improving the ability to quickly detect the types and concentrations of target substances.

[0062] Results analysis:

[0063] The multidimensional optical signals from the sensor array, collected by a mobile app, are imported into Origin. LDA and HCA algorithms are used for rapid data processing, and the output results enable effective identification of complex target substances and concentration prediction. This significantly improves the sensor array's efficiency in resolving complex response signals, providing strong technical support for rapid on-site detection and intelligent analysis.

[0064] (5) Detection of multiple heavy metal ions based on colorimetric sensor array

[0065] Three colorimetric reaction systems were constructed: in the peroxidase-like colorimetric system, TMB (0.1 mL) and H2O2 (6.4 μL) were added sequentially to 2 mL of acetate buffer; in the oxidase-like colorimetric system, TMB (0.1 mL) was added to 2 mL of acetate buffer; and in the laccase-like colorimetric system, 2,4-DCP (1 mL) and 4-APP (1 mL) were added sequentially to 5 mL of MES buffer to prepare the colorimetric reaction solutions corresponding to each response channel. Subsequently, solutions of different target heavy metal ions (20 μL) were mixed with PVP-PtCuAu NCs probe (10 μL) and incubated at room temperature to allow the target heavy metal ions to interact with the nanozyme, thereby regulating the enzyme-like catalytic activity of the nanozyme. After incubation, pre-prepared colorimetric reaction solution (150 μL) was added to each reaction system to generate differentiated colorimetric responses in different colorimetric channels of the sensor array, thereby forming characteristic comprehensive chromatographic signals corresponding to different target heavy metal ions. After the reaction, the resulting reaction solutions were transferred to 96-well microplates for unified colorimetric reading and recording. RGB values ​​were extracted with the support of a color recognition app to establish a color feature data matrix. The sensor array data was input into a machine learning model for training and prediction to achieve target analyte identification and concentration determination. This implementation used Pb... 2+ Ag + As 3+ Cu 2+ Hg 2+ The subject of this study.

[0066] Results analysis:

[0067] After adding different heavy metal ions, the constructed sensor array exhibited significantly differentiated colorimetric changes, enabling the visual identification of the target analyte. For example... Figure 7 As shown in (AC), the heatmap, fingerprint spectrum, and radar image all demonstrate that the three colorimetric channels of the sensor array produce differentiated responses to various heavy metal ions, forming characteristic response fingerprints that can be used to distinguish various target objects. Furthermore, Figure 7(DF) used LDA and HCA algorithms to perform pattern recognition on the multidimensional sensor array data. All target metal ions were accurately classified, and the cluster boundaries of each category of samples were clear with small intra-group dispersion. This indicates that the sensor array constructed in this invention has excellent classification and recognition capabilities, good repeatability, and high data stability.

[0068] (6) Quantitative analysis of multiple heavy metal ions based on colorimetric sensor array

[0069] After constructing each response channel of the sensor array according to the method described in Example (4), different concentrations of target heavy metal ion solutions (20 μL) were mixed with PVP-PtCuAu NCs probes (10 μL) and incubated at room temperature. After incubation, chromogenic reaction solution (150 μL) was added to each reaction system for chromogenic reaction. After the reaction was completed, the resulting reaction solution was transferred to a 96-well microplate for uniform chromogenic reading and recording to obtain the corresponding sensor array response signal for subsequent quantitative analysis.

[0070] Results analysis:

[0071] Pb 2+ Ag + As 3+ The result is as follows Figure 8 As shown, with Pb 2+ For example (corresponding) Figure 8 (A, B, C), for different concentrations (1–40 μM) of Pb 2+ The induced array response signal was analyzed. For example... Figure 8 Figure A and the typical LDA score plot show that this sensor array can effectively distinguish target ions at different concentration levels. Meanwhile, as shown in Figure A... Figure 8 Figure B, and the HCA analysis results further validated this conclusion, showing different concentrations of Pb. 2+ All samples were accurately clustered, and no obvious cluster confusion was observed, indicating that the array is effective for Pb. 2+ Concentration changes exhibited good response consistency and resolution. Since Factor 1 carries the most significant differential information between samples, we established a relationship between the LDA factor (Factor 1) and concentration, such as... Figure 8The results, as shown in Figure C, indicate that within the low concentration range (1-15 μM), Factor 1 exhibits a good linear relationship with concentration, demonstrating high sensitivity and strong quantitative response capability. However, in the high concentration range (15-40 μM), this relationship gradually deviates from linearity, with the curve slope gradually decreasing, exhibiting a significant nonlinear trend. This may be related to the gradual saturation of the sensing sites and the desynchronization of response gains in different channels. The same method was used for quantitative analysis of other heavy metal ions, and the results also showed that the sensor array can effectively distinguish different heavy metal ions and obtained linear or nonlinear relationships between Factor 1 and the concentration of each metal ion, demonstrating its quantitative detection capability within a certain range and further highlighting its application potential in complex analytical systems.

[0072] (7) Detection of mixed heavy metal ions and multiple heavy metal ions in actual water samples based on colorimetric sensor array

[0073] In real-world monitoring environments, metal ion pollution is often more complex, typically manifesting as a complex mixture of multiple ions. Therefore, this study further evaluates the ability of the developed sensor array to identify metal ion mixtures. Simultaneously, to test the applicability of the sensor array, we further explored its detection capability in actual water samples. In this experiment, Pb was prepared in different molar ratios. 2+ / Ag + and Pb 2+ / Hg 2+ Binary mixtures, ternary mixtures, and quaternary mixtures with different molar ratios were also included. Water samples from the lower reaches of the Yangtze River and tap water were collected as representative actual water sample matrices. Five metal ions were introduced into the matrices, and spiked samples with a concentration of approximately 10 μM were prepared. Subsequently, the actual samples were detected using a colorimetric sensor array.

[0074] Results analysis:

[0075] like Figure 9The LDA analysis results show that the sensor array can successfully distinguish between binary (A and D), ternary (B), and quaternary (C) mixtures with different molar ratios, and samples corresponding to different component ratios exhibit good distribution differences in the discrimination space. These results confirm that even with an increase in the types of metal ions and the increased complexity of the mixed system, the sensor array can still effectively distinguish samples with different mixing ratios, achieving a classification accuracy of 100%. This demonstrates the excellent recognition capability of the constructed sensor array for complex multi-component metal ion systems, providing strong support for its application in complex real-world water samples. Furthermore, regardless of whether the samples are river water or tap water, different target metal ion samples can form good separation in the discrimination space, indicating that the sensor array has strong resistance to matrix interference and good environmental adaptability, verifying its promising application prospects in the rapid identification of metal ion pollutants in physical water samples.

[0076] The above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exemplify all embodiments here. However, obvious variations or modifications derived from this solution are still within the scope of protection of the present invention.

Claims

1. A method for rapid identification of multiple heavy metal ions using a machine learning-driven sensor array, characterized in that, Includes the following steps: (1) Preparation of PVP-coated trimetallic nanozymes PVP-PtCuAu NCs: At room temperature, aqueous solutions of chloroplatinic acid hexahydrate (H₂PtCl₆·6H₂O), copper sulfate pentahydrate (CuSO₄·5H₂O), chloroauric acid (HAuCl₄), and polyvinylpyrrolidone (PVP) were mixed thoroughly. The mixture was then stirred vigorously at a certain temperature, and ascorbic acid solution was added as a reducing agent. The reaction was continued for a period of time. After the reaction was completed, the product was collected by centrifugation and washed several times with deionized water. Finally, PVP-coated PtCuAu trimetallic nanozymes, namely PVP-PtCuAu NCs, were obtained by freeze-drying. (2) Construction of the colorimetric sensor array: A multi-channel detection mode with cross-response characteristics was constructed by setting different chromogenic substrate systems; among them... A colorimetric reaction solution A was obtained by using a chromogenic substrate system consisting of TMB solution and H2O2 solution as a peroxidase-like chromogenic channel and adding acetate buffer. Using TMB solution as a single chromogenic substrate system as an oxidase-like chromogenic channel, and adding acetate buffer, chromogenic reaction solution B was obtained. The chromogenic substrate system of the coupling reaction of 2,4-dichlorophenol solution (2,4-DCP) and 4-aminoantipyrine solution (4-APP) was used as the laccase-like chromogenic channel. After adding MES buffer, the chromogenic reaction solution C was obtained. (3) Detection of multiple heavy metal ions based on colorimetric sensor array: The PVP-PtCuAu NCs probe obtained in step (1) was added to the test system containing different target heavy metal ions and incubated at room temperature to allow the heavy metal ions to interact with the nanozymes, thereby regulating their enzyme-like catalytic activity. Next, after each heavy metal ion test system is incubated, it is dropped into the colorimetric reaction solutions A, B and C prepared in step (2) to react. Different color responses are generated in different colorimetric systems of the sensor array, forming a characteristic comprehensive chromatographic response. After the reaction is completed, the resulting reaction solution is transferred to a 96-well microplate for unified colorimetric reading and recording. With the assistance of a color recognition APP, the corresponding RGB values ​​are extracted to establish a color feature data matrix. Subsequently, the data matrix is ​​input into a machine learning model for training and prediction to realize the identification of target species and quantitative analysis of concentration. (4) Determination of heavy metal ions in actual water samples After collecting actual environmental water samples, the target heavy metal ions are added to the sample system, and the detection process of step (3) is repeated to obtain the RGB color signal of the actual water sample and input it into the machine learning model to obtain the type and concentration of heavy metal ions in the actual sample. (5) Identification of multi-component heavy metal ion mixtures Different metal ions were combined into binary, ternary and quaternary mixed systems in different molar ratios to simulate mixed samples under different pollution composition conditions. The detection process of step (3) was repeated to obtain the RGB color signals of different mixed systems. With the help of the powerful multidimensional signal processing capabilities of the machine learning model, the mixing situation of various heavy metal mixtures was obtained.

2. The method as described in claim 1, characterized in that, In step (1), the ratio of the amounts of H2PtCl6·6H2O, CuSO4·5H2O, HAuCl4, PVP aqueous solution and ascorbic acid aqueous solution is 3 mL: 3 mL: 3 mL: 15 mL: 3 mL; wherein, the concentration of H2PtCl6·6H2O is 5 mM, the concentration of CuSO4·5H2O is 5 mM, the concentration of HAuCl4 is 5 mM, the concentration of PVP aqueous solution is 1 mg / mL, and the concentration of ascorbic acid aqueous solution is 100 mM.

3. The method as described in claim 1, characterized in that, In step (1), the reaction temperature is 30 °C and the reaction time is 2 h.

4. The method as described in claim 1, characterized in that, In step (2), In the colorimetric reaction solution A, the volume ratio of TMB solution, H2O2 solution, and acetate buffer is 0.1 mL: 6.4 μL: 2 mL. In the colorimetric reaction solution C, the volume ratio of 2,4-DCP solution, 4-APP solution and MES buffer is 1 mL: 1 mL: 5 mL. The concentrations of the TMB solution were 2 mg / mL, the H2O2 solution was 0.75%, the acetate buffer was 0.2 mol / L, the 2,4-DCP solution was 6 mM, the 4-APP solution was 6 mM, and the MES buffer was 60 mM. The pH of the acetate buffer was adjusted to 4, and the pH of the MES buffer was adjusted to 7.

5. The method as described in claim 1, characterized in that, In step (3), the color recognition program can accurately acquire images and convert optical signals into values ​​for the Red, Green and Blue channels.

6. The method as described in claim 1, characterized in that, In step (3), the machine learning model can process the color feature data matrix generated by the color recognition program and predict the category and concentration of the target object according to the qualitative model and the quantitative model, respectively.

7. The method as described in claim 1, characterized in that, In step (4), the actual environmental water sample is taken from the water sample of the lower reaches of the Yangtze River or the tap water of the main campus of Jiangsu University as the representative actual sample matrix.

8. The method as described in claim 1, characterized in that, In step (5), the selected proportions of the multi-component heavy metal ion mixtures with different molar ratios need to cover a range of relative concentration changes from single components to multiple proportions coexisting.

Images