Method for identifying different kinds of protein powder

Abstract

The invention discloses a method for identifying different kinds of protein powder, and belongs to the field of protein powder source detection. The method comprises the following steps: S1, synthesizing a gold nanocluster sensing solution; s2, preparing different types of protein powder into an analyte solution, uniformly mixing the analyte solution with the gold nano-cluster sensing solution synthesized in S1, a buffer solution, a TMB solution and an H2O2 solution, and measuring the absorbance change rate by using a microplate reader; taking different protein powder types and the correspondingly measured absorbance change rate values as a detection training set; s3, carrying out model training; and S4, preparing the protein powder to be predicted into a detection solution, and inputting the absorbance change rate value obtained by corresponding measurement into the machine learning model trained in S3 to obtain a protein powder type prediction result. The method can accurately identify protein powder from different sources, shows great potential, and has the advantages of convenience in operation, low cost, high detection precision and the like in cooperation with a machine learning algorithm.

Description

Technical Field

[0001] The present invention relates to the field of detecting the source of protein powder, and specifically to a method for identifying different types of protein powder. Background Art

[0002] As a protein supplement, protein powder plays a key role in promoting muscle synthesis and recovery, especially suitable for people with high-intensity exercise. Protein powder is derived from various raw materials and usually contains additives, resulting in only subtle differences in its sensory characteristics (such as appearance, texture, smell, and taste). These differences are difficult to accurately distinguish by human senses, but differentiating protein powder based on the raw material source is crucial for food safety and quality control.

[0003] Currently, the Kjeldahl method is mostly used for protein powder detection. Based on the principle that the nitrogen content in protein is about 16%, the Kjeldahl method can only estimate the protein content by measuring the total nitrogen content of the substance. This method is inexpensive, but it can only quantify the protein content, cannot distinguish proteins from different sources, and is easily affected by nitrogen-containing substances such as melamine.

[0004] Existing detection techniques also use enzyme-linked immunosorbent assay (ELISA), liquid chromatography-mass spectrometry (LC-MS), and electrophoresis. However, due to limitations such as time-consuming processes, expensive equipment, and professional operation, it is difficult to meet the requirements of on-site rapid detection. Summary of the Invention

[0005] Based on the above technical problems, the present invention proposes a method for identifying different types of protein powder.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A method for identifying different types of protein powder, comprising the following steps:

[0008] S1. Synthesize a gold nanocluster sensing solution;

[0009] S2. Establish a protein powder detection training set;

[0010] Prepare analyte solutions from different types of protein powder respectively, and mix the analyte solutions with the gold nanocluster sensing solution synthesized in S1, buffer solution, TMB solution, and H 2 O 2 solution, and use a microplate reader to measure the change rate of absorbance; use different protein powder types and the corresponding measured change rate values of absorbance as the protein powder detection training set;

[0011] S3. Perform model training;

[0012] Select a machine learning model and train it using the protein powder detection training set established in S2 to obtain a trained machine learning model;

[0013] S4. Perform protein powder type prediction;

[0014] Prepare the protein powder to be predicted into a detection solution, and mix the detection solution with the gold nanocluster sensing solution, buffer solution, TMB solution and H 2 O 2 solution, and use an enzyme-linked immunosorbent assay (ELISA) reader to measure the rate of change in absorbance; input the measured absorbance change rate value into the machine learning model trained in S3 to obtain the protein powder type prediction result.

[0015] In the above step S1:

[0016] Mix and stir an aqueous solution of bovine serum albumin and a chloroauric acid solution to obtain a mixed solution; then adjust the pH value of the mixed solution with an alkali solution, and continuously stir and react at 70°C - 80°C. After the reaction, purify the obtained AuNCs solution by dialysis, and then centrifuge to remove unreacted substances to finally obtain the gold nanocluster sensing solution.

[0017] Preferably, add Fe 3+ , Cu 2+ , Al 3+ , Ag + four metal ions to the gold nanocluster sensing solution, and together with the gold nanocluster sensing solution without added metal ions, a total of five sensing solutions with different degrees of nanozyme enzyme activity are constructed.

[0018] More preferably, the concentration of the aqueous solution of bovine serum albumin is 40 - 60 mg / mL, the concentration of the chloroauric acid solution is 8 - 15 mM, control the volume ratio of the aqueous solution of bovine serum albumin to the chloroauric acid solution to be 1:1 - 3, and the stirring time is 5 - 10 minutes; then adjust the pH value of the mixed solution to 12 with NaOH, and continuously stir and incubate at 60 - 70°C for 15 - 20 minutes to promote the formation of AuNCs; after the reaction, purify the obtained AuNCs solution by dialysis for 20 - 24 hours, and then centrifuge at 8000 - 10,000 rpm for 10 - 15 minutes to remove unreacted substances; finally, store the purified AuNCs solution at 4°C for standby, which is the gold nanocluster sensing solution;

[0019] Prepare 0.1 mM solutions of ferric chloride, copper chloride, aluminum chloride and silver nitrate respectively, and mix them evenly with the 5 mg / mL AuNCs solution at a volume ratio of 1:1, and incubate for 5 - 10 min to obtain four sensing solutions with different peroxidase-like activities; the 5 mg / mL AuNCs solution is directly used as the 5th sensing solution.

[0020] Preferably, in step S2:

[0021] Dissolve various commercially available common protein powder standard samples in hot water at 50 - 55 °C respectively to prepare protein powder solutions with a concentration of 0.05 - 0.07 mg / mL as the analyte solutions for the identification experiment; sequentially add the sensing solution, different types of protein powder solutions, acetic acid - sodium acetate buffer solution, TMB solution and H 2 O 2 solution into a 96 - well plate, and mix well at room temperature; immediately use an enzyme - linked immunosorbent assay (ELISA) reader to record the absorbance of oxidized TMB at 652 nm after mixing, and repeat each experiment multiple times;

[0022] According to the formula ΔA / A 0 =(A - A 0 ) / A 0 calculate the absorbance change rate, where A 0 is the absorbance value of the blank control.

[0023] For further optimization, dissolve the protein powder in hot water at 50 °C to prepare a protein powder solution with a concentration of 0.05 mg / mL; sequentially add 10 μL of the sensing solution, 10 μL of different types of protein powder solutions, 140 μL of acetic acid - sodium acetate buffer solution, 20 μL of TMB solution and 20 μL of H 2 O 2 solution into a 96 - well plate, and mix well at room temperature; immediately use an ELISA reader to record the absorbance of oxidized TMB at 652 nm after mixing, and repeat each experiment 6 times;

[0024] The concentration of the acetic acid - sodium acetate buffer solution is 0.2 M and the pH is 4.0; the concentration of the TMB solution is 5 mM, and the concentration of the H 2 O 2 solution is 200 mM.

[0025] Preferably, in step S3:

[0026] The machine - learning models used are models among k - nearest neighbor, random forest, decision tree, support vector machine, multi - layer perceptron and linear discriminant analysis algorithms. More preferably, it is the model in the linear discriminant analysis algorithm.

[0027] Preferably, in step S4:

[0028] Dissolve the protein powder to be predicted in hot water at 50 °C to prepare a protein powder solution with a concentration of 0.05 mg / mL; sequentially add 10 μL of the sensing solution, 10 μL of the protein powder solution, 140 μL of acetic acid - sodium acetate buffer solution with a concentration of 0.2 M, 20 μL of TMB solution with a concentration of 5 mM and 20 μL of H 2 O 2The solution was thoroughly mixed at room temperature; immediately after mixing, the absorbance of oxidized TMB at 652 nm was recorded using a microplate reader; according to the formula ΔA / A 0 =(A - A 0 ) / A 0 the change rate of absorbance was calculated, where A 0 was the absorbance value of the blank control.

[0029] The principle and beneficial technical effects of the present invention are as follows:

[0030] The present invention utilizes four metal ions (Fe 3+ , Cu 2+ , Al 3+ , Ag + ) to regulate the peroxidase-like activity of gold nanoclusters (AuNCs) nanozymes, constructing five sensing solutions with nanozyme activities of different degrees. Different protein powders will interact with the five sensing solutions to different extents, that is, different types of protein powders will complex with metal ions to different degrees. This competition for metal ions changes the peroxidase-like activity of the sensor solution. In addition, different protein powders can also bind to AuNCs to different degrees, further affecting their peroxidase-like activity. The change in the peroxidase-like activity of the sensing solution affects the generation of hydroxyl radicals, thereby affecting the oxidation degree of TMB. Combining with machine learning algorithms, different protein powders from different sources, brands, and origins can be accurately discriminated.

[0031] Specifically,

[0032] (1) The present invention uses BSA as a template to simply and rapidly synthesize AuNCs nanozymes with excellent peroxidase-like activity; and uses metal ions to regulate the enzyme activity of AuNCs, constructing five sensing elements with different enzyme activities to improve the detection performance.

[0033] (2) The present invention does not rely on large-scale instruments, is easy to operate, and has low costs.

[0034] (3) The present invention has excellent performance in the detection of protein powders, can accurately identify protein powders from different sources, shows great potential, and when combined with machine learning algorithms, has the advantages of easy operation, low cost, not relying on large and expensive instruments, and high detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic diagram of the method for the present invention to identify different types of protein powders;

[0036] Figure 2Characterization diagrams of the synthesized gold nanocluster sensing solution; among them, (A) is the transmission electron microscope image (TEM), (B) is the high-resolution transmission electron microscope image (HRTEM), (C) is the particle size distribution diagram, (D) is the fluorescence excitation spectrum and emission spectrum diagram, (E) is the CIE chromaticity diagram, (F) is the Zeta potential diagram, (G) is the Fourier transform infrared spectrum diagram (FT-IR) of BSA and AuNCs, (H) is the X-ray photoelectron spectroscopy diagram, and (I) is the Au 4f analysis diagram;

[0037] Figure 3 is Ag + -AuNCs, Al 3+ -AuNCs, AuNCs, Cu 2+ -AuNCs, Fe 3+ -AuNCs in TMB-H 2 O 2 Typical absorption spectra in the system;

[0038] Figure 4 is the characterization diagram of the enzymatic reaction kinetics experiment of AuNCs on H 2 O 2 ; among them, (A) and (B) are the steady-state kinetic curves and Lineweaver-Burk plots of AuNCs on TMB respectively, and (C) and (D) are Cu 2+ -AuNCs on the steady-state kinetic curves and Lineweaver-Burk plots of TMB; (E) and (F) are Fe 3+ -AuNCs on the steady-state kinetic curves and Lineweaver-Burk plots of TMB; (G) and (H) are Ag + -AuNCs on the steady-state kinetic curves and Lineweaver-Burk plots of TMB; (I) and (J) are Al 3+ -AuNCs on the steady-state kinetic curves and Lineweaver-Burk plots of TMB;

[0039] Figure 5 is the characterization diagram of the enzymatic reaction kinetics experiment of AuNCs on TMB; among them, (A) and (B) are the steady-state kinetic curves and Lineweaver-Burk plots of AuNCs on TMB respectively; (C) and (D) are Cu 2+ -AuNCs on the steady-state kinetic curves and Lineweaver-Burk plots of TMB; (E) and (F) are Fe 3+ -AuNCs on the steady-state kinetic curves and Lineweaver-Burk plots of TMB; (G) and (H) are Ag +- Steady-state kinetic curves and Lineweaver-Burk plots of TMB by AuNCs; (I) and (J) are Al 3+ - Steady-state kinetic curves and Lineweaver-Burk plots of TMB by AuNCs;

[0040] Figure 6 Pattern recognition results of 11 protein powders by the AuNCs-based nanozyme colorimetric sensor array; among them, (A) is the colorimetric response pattern of the protein powder by the array sensor (A-A 0 ) radar chart; (B) is the linear discriminant analysis (LDA) chart drawn based on the first two Factors of the colorimetric response pattern. Detailed implementation manners

[0041] The present invention constructs a nanozyme-based colorimetric sensor array using AuNCs for identifying and differentiating protein powders. AuNCs are synthesized by a one-pot reaction using BSA and HAuCl 4 . To adjust its peroxidase-like activity, four metal ions (Fe 3+ , Cu 2+ , Al 3+ and Ag + ) are introduced, generating five different sensor elements.

[0042] Hydrogen peroxide (H 2 O 2 ) can generate hydroxyl radicals (*OH), oxidizing colorless TMB into blue oxTMB. However, under natural conditions, the rate of H 2 O 2 oxidizing TMB to generate *OH is extremely slow. AuNCs can catalyze H 2 O 2 to generate *OH and significantly accelerate the formation of oxTMB due to its peroxidase-like activity. In addition, metal ions regulate the peroxidase-like activity of AuNCs. Fe 3+ , Cu 2+ and Al 3+ enhance the peroxidase-like activity of AuNCs to varying degrees, while Ag + inhibits this activity. Therefore, the combination of AuNCs with four different metal ions, together with AuNCs alone, forms five different sensing elements with different levels of peroxidase-like activity, and these elements are used in sensing experiments.

[0043] When protein powder is added to the system, different types of protein powder will complex with metal ions to varying degrees. This competition for metal ions changes the peroxidase-like activity of the sensor element. In addition, different protein powders can also bind to AuNCs to varying degrees, further affecting their peroxidase-like activity. The change in the peroxidase-like activity of the sensor element affects the generation of hydroxyl radicals, thereby affecting the degree of oxidation of TMB.

[0044] Based on the above principle, as Figure 1 shown, the present invention proposes a method for identifying different types of protein powder assisted by machine learning, including the following steps:

[0045] S1. Synthesize the gold nanocluster sensing solution;

[0046] First, mix an aqueous solution of bovine serum albumin (BSA) (10 mL, 50 mg / mL) with a solution of chloroauric acid (HAuCl 4 ) (10 mL, 10 mM), and stir for 5 minutes; then adjust the pH of the mixture to 12 with NaOH, and continuously stir and incubate at 70 °C for 15 minutes to promote the formation of AuNCs. After the reaction, the obtained AuNCs solution is purified by dialysis for 24 hours, and then centrifuged at 10,000 rpm for 15 minutes to remove unreacted substances. The finally purified AuNCs solution is stored at 4 °C for standby.

[0047] Prepare 0.1 mM solutions of ferric chloride (FeCl 3 ), copper chloride (CuCl 2 ), aluminum chloride (AlCl 3 ), and silver nitrate (AgNO 3 ), and mix them evenly with the 5 mg / mL AuNCs element in a ratio of 1:1, and incubate for 5 min to obtain four kinds of sensing elements with different peroxidase activities. The 5 mg / mL AuNCs solution is directly used as the 5th sensing element.

[0048] S2. Establish a protein powder detection training set;

[0049] Dissolve various commercially available common protein powders in hot water at 50 °C respectively to prepare solutions with a final concentration of 0.05 mg / mL as the analytes for the discrimination experiment. Add the sensing element (10 μL), different protein powder solutions (10 μL), acetic acid-sodium acetate buffer (HAC-NaAC, 140 μL, 0.2 M, pH 4.0), TMB solution (20 μL, 5 mM), and H 2 O 2Solution (20 μL, 200 mM), mix well thoroughly at room temperature. Immediately after mixing, use a microplate reader to record the absorbance (A) of oxidized TMB (ox TMB) at 652 nm. Repeat each experiment 6 times. According to the formula ΔA / A 0 =(A - A 0 ) / A 0 calculate the change rate of absorbance, where A 0 is the absorbance value of the blank control. Use the data of different protein powder types and the corresponding measured change rates of absorbance as the protein powder detection training set.

[0050] S3. Conduct model training;

[0051] Select a machine learning model and train it using the protein powder detection training set established in S2 to obtain a trained machine learning model.

[0052] The machine learning models used are models among k-nearest neighbor, random forest, decision tree, support vector machine, multi-layer perceptron, and linear discriminant analysis algorithms.

[0053] For example, use classical linear discriminant analysis (LDA) in SYSTAT (version 13.0) for linear discriminant analysis. In LDA, all variables are used in the model (complete model), and the tolerance is set to 0.001. The colorimetric response pattern is converted to a canonical discriminant pattern. The Mahalanobis distance from each individual pattern in the multi-dimensional space to the centroid of each group is calculated, and samples are assigned based on the shortest Mahalanobis distance. After machine learning analysis and processing, different types of protein powders are distinguished.

[0054] S4. Conduct prediction of protein powder types;

[0055] Prepare the protein powder to be predicted into a detection solution, and mix the detection solution with the gold nanocluster sensing solution, buffer solution, TMB solution, and H 2 O 2 solution well, and use a microplate reader to measure the change rate of absorbance; input the measured change rate value of absorbance into the machine learning model trained in S3 to obtain the prediction result of the protein powder type.

[0056] Specifically, dissolve the protein powder to be predicted in hot water at 50 °C to prepare a protein powder solution with a concentration of 0.05 mg / mL; sequentially add 10 μL of the sensing solution, 10 μL of the protein powder solution, 140 μL of an acetic acid-sodium acetate buffer solution with a concentration of 0.2 M, 20 μL of a TMB solution with a concentration of 5 mM, and 20 μL of an H 2 O 2 solution into a 96-well plate, and mix well thoroughly at room temperature; immediately after mixing, use a microplate reader to record the absorbance of oxidized TMB at 652 nm; according to the formula ΔA / A 0 =(A - A0 ) / A 0 Calculate the absorbance change rate, where A 0 is the absorbance value of the blank control.

[0057] On the basis of the above method, the present invention also conducted an enzymatic reaction kinetics experiment.

[0058] Using TMB as the substrate: sequentially add the sensing element (50 μL), ultrapure water (50 μL), acetic acid-sodium acetate buffer (HAC-NaAC, 700 μL, 0.2 M, pH 4.0), TMB solutions with different concentrations (0 - 10 mM, 100 μL), and H 2 O 2 solution (200 mM, 100 μL) into a centrifuge tube, and mix well at room temperature. After incubating for 15 minutes, transfer 200 μL of the mixture to a 96-well plate, and use a microplate reader (BioTek Cytation 5) to detect the absorbance of oxidized TMB (oxTMB) at 652 nm.

[0059] Using H 2 O 2 as the substrate: The concentration gradient of H 2 O 2 is 0 - 500 mM, the concentration of TMB is fixed at 5 mM, and the remaining steps are the same as above. To further study the effect of time on the enzymatic activity of the sensing element, directly mix the sensing element (10 μL), ultrapure water (10 μL), HAC-NaAC buffer (140 μL), TMB solutions with different concentrations (0 - 10 mM, 20 μL), and H 2 O 2 solution (200 mM, 20 μL) in a 96-well plate, and set three replicates for each group of samples. Use a microplate reader to record the absorbance at 625 nm at intervals of 27 seconds to monitor the reaction process in real time. When using H 2 O 2 as the substrate, its concentration gradient is 0 - 500 mM, the concentration of TMB is fixed at 5 mM, and the remaining conditions are the same as the above method.

[0060] v = v max [S] / K m +[S]

[0061] Calculate the Michaelis constant (K m ) and the maximum reaction rate (v max ) according to the formula.

[0062] The feasibility of the present invention was further verified through the enzymatic reaction kinetics experiment.

[0063] Hydrogen peroxide (H 2 O 2) can generate hydroxyl radicals (*OH), which oxidize colorless TMB to blue ox TMB. However, under natural conditions, the process of generating *OH to oxidize TMB is very slow. AuNCs possess peroxidase-like activity and can catalyze H 2 O 2 to generate *OH, thus accelerating the formation of ox TMB. In addition, metal ions can regulate the peroxidase-like activity of AuNCs. In this invention, four metal ions (Fe 2 O 2 ), Cu 3+ , Al 2+ , and Ag 3+ ) are used to construct sensing elements with five different peroxidase-like activities. Different protein powders will affect the enzyme activities of each sensing element to varying degrees, thereby causing varying degrees of changes in the oxidation degree of the chromogenic substrate TMB. This change will be reflected in the depth of the blue color of the solution and the height of the absorbance. The change in the absorbance of ox TMB at 652 nm in a 96-well plate is measured using a multifunctional microplate reader to obtain a colorimetric response pattern. A machine learning algorithm is used to process the colorimetric response pattern to accurately distinguish protein powders from different sources. + )

[0064] Figure 2 is the characterization diagram of the synthesized gold nanocluster sensing solution. Figure 2 In (A) and (B), it shows that AuNCs are almost monodispersed, and the lattice spacing is 0.0235 nm; (C) indicates that the average particle size of AuNCs is about 2.2 nm; (D) indicates that AuNCs have two excitation wavelengths of 380 nm and 495 nm, and the fluorescence emission wavelength is 648 nm; (E) shows that the CIE chromaticity diagram of AuNCs is (0.57117, 0.32914); (F) shows that under the condition that the pH value is higher than the isoelectric point (PI) of BSA, which is 4.7, AuNCs synthesized using BSA as a template show negative charge, and the Zeta potential is -34.23 mV. As Figure 2 shown in (G), the FT-IR spectrum shows significant absorption peaks in the range of 3200 - 3400 cm -1 , which belong to the stretching vibrations of O-H and N-H bonds, confirming the presence of -NH 2 and -COOH functional groups on the surface of AuNCs. The characteristic peaks of the bending vibration of N-H and the stretching vibration of the amide bond are observed in the range of 1500 - 1700 cm -1 . Further comparing the amide I band (1650 cm -1 ) and amide II band (1540 cm -1), it was found that the peak position changed slightly, indicating that the influence of AuNCs on the secondary structure of BSA after being coated with BSA molecules was negligible. The surface chemical composition of AuNCs and the oxidation state of Au were analyzed by X-ray photoelectron spectroscopy (XPS), as Figure 2 in (H), the results showed that characteristic peaks appeared at binding energies of 531 eV (O 1s), 400 eV (N 1s), 285 eV (C 1s), 163 eV (S2p), and 84 eV (Au 4f) for AuNCs. The Au 4f spectrum in (I) showed a doublet structure: 83.7 eV corresponded to Au 0 , and 84.3 eV corresponded to Au + , indicating that the Au - in the precursor AuCl 3+ was chemically reduced by BSA and transformed into Au 0 and Au + .

[0065] Table 1 shows the enzyme kinetic parameters of five sensing elements.

[0066] Table 1

[0067]

[0068] Figure 3 is for Ag + -AuNCs, Al 3+ -AuNCs, AuNCs, Cu 2+ -AuNCs, Fe 3+ -AuNCs in the TMB-H 2 O 2 system. As Figure 3 shown, the absorbances of the characteristic absorption peaks at 652 nm of ox TMB in the five systems were different, indicating that the peroxidase-like activities of the five sensing elements were different.

[0069] Figure 4 and Figure 5 are the experimental characterization diagrams of the enzymatic reaction kinetics of AuNCs for H 2 O 2 and TMB, respectively. Michaelis-Menten kinetics is one of the most well-known enzyme kinetic models. It quantitatively describes the relationship between the substrate concentration ([S]) and the reaction rate (v), which can be expressed by the following equation:

[0070] v = v max [S] / K m +[S] (1)

[0071] v max represents the maximum reaction rate when the nanozyme system is completely saturated with the substrate. The Michaelis constant (Km ) is the substrate concentration when the reaction rate reaches v max at half, reflecting the affinity between the enzyme and the substrate. The smaller the K m value, the higher the substrate affinity, which can be used to screen the optimal substrate for nanozymes. As shown in Table 1, the substrate affinities for H 2 O 2 decrease in the following order: Fe 3+ -AuNCs > AuNCs > Cu 2+ -AuNCs > Al 3+ -AuNCs > Ag + -AuNCs, and the substrate affinities for TMB decrease in the following order: Al 3+ -AuNCs > Ag + -AuNCs > Cu 2+ -AuNCs > Fe 3+ -AuNCs > AuNCs.

[0072] Figure 6 Figure 3 shows the pattern recognition results of the AuNCs-based nanozyme colorimetric array sensor for 11 protein powders. The species sources, countries of origin, and brand information of the 11 protein powders are shown in Table 2.

[0073] Table 2

[0074]

[0075]

[0076] Protein powders are derived from various raw materials and usually contain additives, resulting in only slight differences in their sensory characteristics (such as appearance, texture, odor, and taste). These differences are difficult to accurately distinguish by the human senses, but differentiating protein powders based on their raw material sources is crucial for food safety and quality control. To solve this problem, the invented nanozyme colorimetric array sensor was used to identify 11 common commercially available protein powders. The selected protein powders cover mainstream types such as whey protein, casein, mixed protein, pea protein, and brown rice protein, involving different species, origins, and brands (as shown in Table 2). The final concentration of the protein powder solution was 0.05 mg / mL. The colorimetric responses of the array sensor to different protein powders are as Figure 6 (A) shown in the figure. Each response curve corresponds to a single sensing element, and its size and shape reflect the response intensity and pattern. The radar chart shows the specific response profiles of the sensing units, and the differences between samples are significant: animal proteins (such as whey and casein) trigger stronger responses in most sensing elements, while plant proteins (such as soy and pea) generally have weaker responses; mixed proteins (such as whey + soy) show medium response intensities due to their compositional complexity.

[0077] Canonical scores were calculated by linear discriminant analysis (LDA) (training matrix: 5 sensing elements × 11 protein powders × 6 replicates), and the LDA score plot based on the first two principal components of the response pattern ( Figure 6 in (B)) showed that the protein powders were clearly clustered by source: animal protein powders were distributed in the upper region of the graph, plant protein powders were concentrated in the lower region, and mixed protein powders (B1, B2) were located in the boundary region between animal and plant protein powders, reflecting their mixed characteristics. The accuracy of the jackknife classification matrix by cross-validation reached 98.48%, and the recognition accuracy for unknown samples was 97.73%, confirming that the array sensor could accurately distinguish the sources of protein powders.

Claims

1. A method for identifying different types of protein powder, characterized in that The following steps are involved: S1, synthesis of gold nanocluster sensing solution; S2, establish a protein powder detection training set; Different types of protein powder were prepared into analyte solutions, and the analyte solutions were mixed with the gold nanocluster sensing solution synthesized by S1, buffer solution, TMB solution and H2O2 solution, and the absorbance change rate was measured using an enzyme marker; different types of protein powder and the corresponding measured absorbance change rate values ​​were used as protein powder detection training sets; S3, perform model training; Select a machine learning model and use the protein powder detection training set established by S2 for training to obtain a trained machine learning model; S4, predicting the types of protein powder; The protein powder to be predicted is prepared into a detection solution, and the detection solution is mixed with the gold nanocluster sensor solution, buffer solution, TMB solution and H2O2 solution, and the absorbance change rate is measured using an enzyme marker. The measured absorbance change rate value is input into the machine learning model trained by S3 to obtain the protein powder type prediction result.

2. A method for identifying different types of protein powder according to claim 1, characterized in that, In step S1: the bovine serum albumin aqueous solution and the chloroauric acid solution are mixed and stirred to obtain a mixed solution; then the pH value of the mixed solution is adjusted with an alkaline solution, and the reaction is continuously stirred at 70°C-80°C. After the reaction is completed, the obtained AuNCs solution is purified by dialysis, and then centrifuged to remove unreacted products, and finally a gold nanocluster sensing solution is obtained.

3. A method for identifying different types of protein powder according to claim 2, characterized in that: Add Fe to the gold nanocluster sensing solution 3+ , Cu 2+ 、Al 3+ 、Ag + The four metal ions, plus the gold nanocluster sensing solution without added metal ions, constructed a total of five sensing solutions with different degrees of nanozyme enzymatic activity.

4. A method for identifying different types of protein powder according to claim 3, characterized in that: The concentration of the bovine serum albumin aqueous solution is 40-60 mg / mL, the concentration of the chloroauric acid solution is 8-15 mM, the volume ratio of the bovine serum albumin aqueous solution to the chloroauric acid solution is controlled to be 1:1-3, and the stirring time is 5-10 minutes; then, the pH value of the mixed solution is adjusted to 12 with NaOH, and the mixture is continuously stirred and incubated at 60-70° C. for 15-20 minutes to promote the formation of AuNCs; after the reaction is completed, the obtained AuNCs solution is purified by dialysis for 20-24 hours, and then centrifuged at 8000-10,000 rpm for 10-15 minutes to remove unreacted products; the finally purified AuNCs solution is stored at 4° C. for standby use, which is the gold nanocluster sensing solution; 0.1 mM ferric chloride, copper chloride, aluminum chloride and silver nitrate solutions were prepared respectively, mixed evenly with 5 mg / mL AuNCs solution in a volume ratio of 1:1, and incubated for 5-10 min to obtain four sensing solutions with different peroxidase activities; 5 mg / mL AuNCs solution was directly used as the fifth sensing solution.

5. A method for identifying different types of protein powder according to claim 1, characterized in that, In step S2: a plurality of commercially available common protein powder standard samples are dissolved in 50-55° C. hot water to prepare a protein powder solution with a concentration of 0.05-0.07 mg / mL as the analyte solution for the identification experiment; a sensing solution, different types of protein powder solutions, acetic acid-sodium acetate buffer, TMB solution and H2O2 solution are sequentially added to a 96-well plate, mixed thoroughly at room temperature, and then allowed to stand for 10-15 minutes, and the absorbance of oxidized TMB at 652 nm is recorded using an enzyme reader, and each group of experiments is repeated multiple times; The absorbance change rate was calculated according to the formula ΔA / A0=(A-A0) / A0, where A0 is the absorbance value of the blank control.

6. A method for identifying different types of protein powder according to claim 5, characterized in that: The protein powder was dissolved in 50°C hot water to prepare a protein powder solution with a concentration of 0.05 mg / mL; 10 μL of sensing solution, 10 μL of different types of protein powder solutions, 140 μL of acetic acid-sodium acetate buffer, 20 μL of TMB solution and 20 μL of H2O2 solution were added to a 96-well plate in sequence and mixed thoroughly at room temperature; after mixing, the absorbance of oxidized TMB at 652 nm was immediately recorded using an ELISA reader, and each group of experiments was repeated 6 times; The concentration of the acetic acid-sodium acetate buffer solution is 0.2 M, and the pH is 4.0; the concentration of the TMB solution is 5 mM, and the concentration of the H2O2 solution is 200 mM.

7. A method for identifying different types of protein powder according to claim 1, characterized in that, In step S3: the machine learning models used are models in k-nearest neighbor, random forest, decision tree, support vector machine, multilayer perceptron and linear discriminant analysis algorithm.

8. A method for identifying different types of protein powder according to claim 1, characterized in that: In step S4: the protein powder to be predicted is dissolved in 50°C hot water to prepare a protein powder solution with a concentration of 0.05 mg / mL; 10 μL of sensing solution, 10 μL of protein powder solution, 140 μL of 0.2M acetic acid-sodium acetate buffer, 20 μL of 5mM TMB solution and 20 μL of 200mM H2O2 solution are sequentially added to a 96-well plate, and mixed thoroughly at room temperature; immediately after mixing, the absorbance of oxidized TMB at 652nm is recorded using an ELISA instrument; the absorbance change rate is calculated according to the formula ΔA / A0=(A-A0) / A0, where A0 is the absorbance value of the blank control.

Images