Rapid identification method for cow milk adulteration based on particle size fingerprint spectrum

Abstract

The invention provides a rapid identification method for cow milk adulteration based on a particle size fingerprint spectrum, which comprises the following steps: taking a sample to be detected, carrying out centrifugal treatment, and taking supernate after centrifugation; carrying out particle size distribution parameter characterization on the obtained supernate; according to the obtained particle size distribution parameter characterization data, establishing a random forest model taking a particle size fingerprint as a discrimination basis, and taking the random forest model as a particle size distribution parameter model; comparing the particle size fingerprint spectrum of the sample to be detected with the particle size fingerprint spectrum of cow milk by utilizing the constructed particle size distribution parameter model to obtain a comparison result; and judging whether the to-be-detected sample is adulterated milk or not according to a comparison result. According to the method disclosed by the invention, by detecting endogenous natural nanoparticles in cow milk and combining machine learning and a random forest algorithm, a random forest model taking a particle size fingerprint spectrum as a judgment basis is constructed, and through database modeling analysis, cow milk adulteration can be rapidly identified on the premise that adulterants do not need to be known.

Description

Technical Field

[0001] The present invention belongs to the field of food detection, and in particular relates to a method for detecting cow's milk, and more particularly to a method for rapidly identifying cow's milk adulteration based on a particle size fingerprint. Background Art

[0002] Milk testing in my country primarily encompasses three categories: nutrient content testing, typically using dairy analyzers and the Kjeldahl method; antibiotic and microbial residue testing, typically using methods such as high-performance liquid chromatography (HPLC), liquid chromatography-mass spectrometry (LC-MS / MS), and ELISA; and adulteration and illegal additive testing, primarily relying on chemical analysis, chromatography, and spectroscopy to comprehensively ensure milk quality and safety. Existing testing methods primarily rely on single-molecule detection, and these molecular indicators often represent molecules already present in milk. Existing technologies are unable to determine whether milk has been artificially adulterated to manipulate these indicators. Furthermore, existing technologies employ localized single-point sampling for milk adulteration analysis, requiring different detection methods for different adulterants. This makes detection relatively difficult, complex, and time-consuming. Chinese patent CN102253037A discloses a rapid milk adulteration test strip packaged with four test strips for starch, urea, alkali, and salt. This strip can simultaneously detect all four adulterants, but the accuracy of this method is low and the detection limit is high.

[0003] Therefore, there is an urgent need for a specific detection method that can directly determine adulteration of target samples, achieving efficient detection, significant results, and cost-effectiveness. Summary of the Invention

[0004] To address the challenges of existing milk analysis techniques, which include relatively difficult detection techniques, cumbersome processes, and lengthy processing times, the present invention provides a rapid method for identifying milk adulteration based on particle size fingerprinting. This method detects endogenous natural nanoparticles in milk and combines machine learning with a random forest algorithm, using database modeling and analysis. This method allows for rapid identification of milk adulteration without requiring knowledge of the adulterant.

[0005] The technical solution adopted by the present invention is: a rapid identification method for milk adulteration based on particle size fingerprint, comprising the following steps: S1. Centrifuge the test sample and standard milk separately, and take the supernatant after centrifugation; S2. Characterizing the particle size distribution parameters of the obtained supernatant to obtain the particle size distribution parameter characterization data of the test sample and the particle size distribution parameter characterization data of the standard milk; S3. Based on the obtained particle size distribution parameter characterization data of the test sample and the standard milk, a random forest model based on the particle size fingerprint was established as a particle size distribution parameter model; S4. Using the constructed particle size distribution parameter model, compare the particle size fingerprint of the test sample with the particle size fingerprint of standard milk to obtain a comparison result; and determine whether the test sample is adulterated milk based on the comparison result.

[0006] Previous studies have found that there are a large number of natural micro-nano colloidal particles in cow's milk, which are mainly composed of proteins. These proteins will form casein micelles through hydrophobic interactions. Therefore, each type of cow's milk has its own specific particle size information. Once the milk is adulterated, the properties of the nano-colloidal particles will inevitably change. Among them, the most significant change is the change in particle size distribution. Therefore, the present invention detects the endogenous natural nanoparticles in cow's milk and combines machine learning with the random forest algorithm to construct a random forest model based on particle size fingerprints, i.e., a particle size distribution parameter model. Through database modeling and analysis, it can quickly identify cow's milk adulteration without knowing the adulterant.

[0007] Preferably, the milk adulteration comprises adding adulterants to the milk; the adulterants comprise one or more of isolated whey protein, starch, urea, formaldehyde, and water, preferably one or more of isolated whey protein, starch, urea, and formaldehyde.

[0008] Preferably, in step S1, the sample is selected from liquid milk or milk powder.

[0009] Preferably, the milk adulteration comprises adding an adulterant, whey protein isolate, to the milk; and the detection limit of the adulterant, whey protein isolate, by the method is 0.5 mL / L, more preferably 5 mL / L.

[0010] Preferably, the milk adulteration comprises adding adulterated starch to the milk; the detection limit of the adulterated starch by the method is 0.1 g / L, more preferably 1 g / L.

[0011] Preferably, the milk adulteration comprises adding urea as an adulterant to the milk; wherein the lower limit of detection of the adulterant urea by the method is 0.1 g / L, more preferably 0.75 g / L.

[0012] Preferably, the milk adulteration comprises adding the adulterant formaldehyde to the milk; wherein the detection limit of the adulterant formaldehyde by the method is 0.5 mL / L, more preferably 5 mL / L.

[0013] Preferably, the milk adulteration comprises adding adulterated water to the milk; wherein the detection limit of the adulterated water by the method is 10 mL / L, more preferably 300 mL / L.

[0014] Preferably, in step S1, the rotation speed of the centrifugal treatment is 3000-5000 g, and the time of the centrifugal treatment is 10-20 min.

[0015] Preferably, in step S2, the characterization method of the particle size distribution parameters includes one or more of light scattering measurement, electron microscopy measurement, and X-ray measurement, more preferably light scattering measurement.

[0016] Preferably, in step S2, the characterization condition of the particle size distribution parameter is 20-30°C, more preferably 25°C.

[0017] Preferably, in step S3, the particle size distribution parameter characterization data includes one or more of particle size-volume distribution, particle size-number distribution, and particle size-light scattering intensity distribution, more preferably particle size-volume distribution.

[0018] Preferably, in step S3, the number of decision trees of the random forest model is 300-800, more preferably 600-800.

[0019] Preferably, in step S3, the number of nodes of the random forest model is 10-20, more preferably 12-17.

[0020] Preferably, in step S3, the AUC of the random forest model is not less than 0.95. AUC, defined as the area under the ROC curve, is an evaluation metric for measuring the quality of a binary classification model, indicating the probability that a predicted positive example will be ranked before a negative example. The value range is between 0.5 and 1. The closer the AUC is to 1.0, the better the model performance; when it is equal to 0.5, the model is equivalent to random guessing and has no application value. In a preferred embodiment of the present invention, the AUC value of the constructed random forest model for particle size distribution parameters reached 0.974, indicating that the model has excellent predictive effect.

[0021] Preferably, in step S4, if the comparison result is less than a preset similarity threshold, the sample to be tested is adulterated milk. The similarity threshold is preferably 90% to 95%. The similarity threshold is adjusted based on the number of nodes and decision trees set in the random forest model.

[0022] The present invention provides a rapid, efficient, and simple method for identifying milk adulteration based on particle size fingerprints. It provides accurate results, good reproducibility, and high sensitivity, enabling identification of milk adulteration without requiring knowledge of the adulterant. Experiments have demonstrated that the method has a high discrimination rate for adulterants including whey protein isolate, starch, urea, formaldehyde, and water, requires few parameters for optimization, exhibits excellent anti-fitting properties, and delivers excellent prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a diagram showing particle size distribution parameters of adulterated cow's milk mixed with whey protein isolate according to Example 1 of the present invention.

[0024] Figure 2 This is a diagram showing particle size distribution parameters of adulterated milk mixed with starch according to Example 2 of the present invention.

[0025] Figure 3 This is a diagram showing particle size distribution parameters of adulterated milk mixed with urea according to Example 3 of the present invention.

[0026] Figure 4 This is a diagram showing particle size distribution parameters of adulterated milk mixed with formaldehyde according to Example 4 of the present invention.

[0027] Figure 5 This is a diagram showing particle size distribution parameters of adulterated milk mixed with water according to Example 5 of the present invention.

[0028] Figure 6 This is a graph showing the relationship between the number of nodes and the error in the particle size distribution parameter model constructed in an embodiment of the present invention.

[0029] Figure 7 A graph showing the relationship between the number of decision trees and the error in the particle size distribution parameter model constructed in an embodiment of the present invention.

[0030] Figure 8 This is a ROC (Receiver Operating Characteristic) curve diagram of the particle size distribution parameter model constructed in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The following describes the embodiments of the present invention by specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the case of no conflict, the features in the following examples and embodiments can be combined with each other. In the embodiments of the present invention, unless otherwise specified, the methods used are all conventional methods, and the reagents used can be obtained from commercial sources.

[0032] Example 1: This embodiment provides a method for rapidly identifying milk adulteration by whey protein adulteration, comprising the following steps: S1. The test samples were milk spiked with 0 (real milk), 5, 10, 20, 30, and 40 mL / L of whey protein isolate; the test samples were centrifuged at 4000 g for 15 min, and the supernatant was collected after centrifugation. S2. The sample filtrate was analyzed using a dynamic light scattering (DLS) instrument (Zetasizer Nano-ZS, Malven Instruments Ltd., UK) to obtain raw particle size distribution data, including size-volume distribution, size-number distribution, and size-light scattering intensity distribution. The measurement temperature was 25°C. S3. Based on the particle size distribution parameter characterization data of the sample obtained in step 2, Python modeling was performed. The number of nodes was set to 12, the number of decision trees was set to 600, and the threshold for stopping the splitting of the decision tree nodes was set to 1. A random forest model based on the particle size fingerprint was established as the particle size distribution parameter model. The results are shown in Figure 2. Figure 1 As shown; S4. Using the constructed particle size distribution parameter model, the particle size fingerprint of the test sample was compared with the particle size fingerprint of milk to obtain a comparison result. If the comparison result was less than a preset similarity threshold of 90%, the test sample was adulterated with milk. The results are shown in Table 1.

[0033] Example 2: This example provides a method for rapidly identifying adulterated milk from starch-adulterated milk. Compared to Example 1, this example differs in that the test samples in step S1 are milk adulterated with 0 (genuine milk), 1, 2, 3, 4, and 5 g / L of starch, respectively. The particle size distribution parameter model established in this example shows Figure 2 The discrimination results are shown in Table 1.

[0034] Example 3: This example provides a rapid method for identifying milk adulteration caused by urea. Compared with Example 1, the difference in this example is that the test samples in step S1 are milk adulterated with 0 (genuine milk), 0.75, 1.5, 3, 7.5, and 15 g / L of urea, respectively. The particle size distribution parameter model established in this example is shown in Figure 1. Figure 3 The discrimination results are shown in Table 1.

[0035] Example 4: This example provides a rapid method for identifying milk adulteration caused by formaldehyde-adulterated milk. Compared to Example 1, this example differs in that the test samples in step S1 are milk adulterated with 0 (genuine milk), 5, 10, 20, 30, and 40 mL / L of formaldehyde, respectively. The particle size distribution parameter model established in this example is shown in Figure 1. Figure 4 The discrimination results are shown in Table 1.

[0036] Example 5: This example provides a rapid method for identifying milk adulteration by water-adulterated milk. Compared to Example 1, this example differs in that the test samples in step S1 are milk adulterated with 0 (genuine milk), 10, 50, 100, 150, and 300 mL / L of water, respectively. The particle size distribution parameter model established in this example is shown in Figure 1. Figure 5 The discrimination results are shown in Table 1.

[0037] Table 1. Particle size distribution parameter model identification results from Figure 6 It can be seen that the out-of-bag data error OOB is used to judge the generalization ability of the model. For the particle size fingerprint model, when the number of nodes is 12, the error is the smallest, which is approximately equal to 0.08. Figure 7 It can be seen that the out-of-bag data error OOB is used to judge the quality. For the particle size fingerprint model, the number of nodes is 12. When the number of decision trees is greater than or equal to 600, the error rates of OOB, T (genuine milk), and NT (adulterated milk) tend to be stable, indicating that this model has good anti-fitting ability. Figure 8 As can be seen from the figure, based on the receiver operating characteristic (ROC) curve, the AUC value (area under the ROC curve) of the particle size distribution parameter model is 0.974, indicating that the model has a good predictive effect.

[0038] Combine Figures 6 to 8 As shown in Table 1, the method provided by the present invention for rapid identification of milk adulteration based on particle size fingerprints is fast, efficient, simple, and provides accurate results, good reproducibility, and high sensitivity. It can identify milk adulteration without requiring knowledge of the adulterant. Experiments have demonstrated that the method provided by the present invention has a high discrimination rate for adulterants including whey protein isolate, starch, urea, formaldehyde, and water. Furthermore, it requires few parameters to be optimized, exhibits good anti-fitting properties, and exhibits excellent prediction results. It is widely applicable to adulteration detection in milk and even the dairy product industry.

[0039] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection of the present invention.

Claims

1. A rapid identification method for milk adulteration based on particle size fingerprint, characterized in that: The following steps are involved: S1. Centrifuge the test sample and standard milk separately, and take the supernatant after centrifugation; S2. Characterizing the particle size distribution parameters of the obtained supernatant to obtain the particle size distribution parameter characterization data of the test sample and the particle size distribution parameter characterization data of the standard milk; S3. Based on the obtained particle size distribution parameter characterization data of the test sample and the standard milk, a random forest model based on the particle size fingerprint was established as a particle size distribution parameter model; S4. Using the constructed particle size distribution parameter model, compare the particle size fingerprint of the test sample with the particle size fingerprint of standard milk to obtain a comparison result; and determine whether the test sample is adulterated milk based on the comparison result.

2. The method according to claim 1, wherein The milk adulteration includes adding adulterants into the milk; the adulterants include one or more of isolated whey protein, starch, urea, formaldehyde, and water.

3. The method according to claim 1, wherein In step S1, the sample is selected from liquid milk or milk powder.

4. The method according to claim 1, wherein In step S1, the rotation speed of the centrifugal treatment is 3000-5000 g, and the time of the centrifugal treatment is 10-20 min.

5. The method according to claim 1, wherein In step S2, the characterization method of the particle size distribution parameters includes one or more of light scattering measurement, electron microscopy measurement, and X-ray measurement.

6. The method according to claim 1, wherein In step S3, the particle size distribution parameter characterization data includes one or more of particle size-volume distribution, particle size-number distribution, and particle size-light scattering intensity distribution.

7. The method according to claim 1, wherein In step S3, the number of decision trees of the random forest model is 300-800.

8. The method according to claim 1, wherein In step S3, the number of nodes of the random forest model is 10-20.

9. The method according to claim 1, wherein In step S3, the AUC of the random forest model is not less than 0.

95.

10. The method according to claim 1, wherein In step S4, if the comparison result is less than a preset similarity threshold, the sample to be tested is adulterated milk.

Images