CN117434045A - Method for simultaneously detecting two veterinary drugs based on SERS (surface enhanced Raman Scattering) mark detection and machine learning - Google Patents

Abstract

The invention is applicable to the technical field of Raman spectrum detection, and provides a method for simultaneously detecting two veterinary drugs based on SERS (surface enhanced Raman Scattering) mark detection and machine learning, which comprises the following steps: modifying Raman reporter molecules inside silver-coated gold nanoparticles, modifying drug aptamers on the surfaces, taking Jin Baoci bead nanoflower with modified aptamer complementary chains as a capture probe, constructing a competitive SERS aptamer sensor aiming at different drugs, and uniformly mixing the competitive SERS aptamer sensors in a ratio of 1:1; preparing mixed solution containing two medicines with different concentrations; then uniformly mixing the drug mixed solution with the competitive SERS aptamer sensor of the two drugs, incubating, magnetically separating, washing and re-suspending in PBS buffer solution; dripping the heavy suspension on the surface of a silicon wafer, collecting Raman spectrum, preprocessing the Raman spectrum, and establishing a machine learning model; and finally, testing the contents of the two medicaments in the solution to be tested by using the established prediction model. The detection method improves the efficiency and accuracy of the competitive SERS aptamer sensor for simultaneously detecting two medicaments.

Description

A method for simultaneous detection of two veterinary drugs based on SERS marker detection and machine learning

Technical field

The present invention relates to the technical field of Raman spectroscopy detection, and specifically, to a method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning.

Background technique

Veterinary drugs are often used in the animal husbandry and aquaculture industries to prevent and treat animal diseases or to regulate animal physiological functions in a purposeful manner. In recent years, with the continuous expansion of the scale of breeding and the continuous improvement of the level of intensification, diseases during the breeding process have become increasingly serious. In order to improve economic efficiency, some farmers often excessively use a variety of veterinary drugs during the breeding process, and even abuse banned drugs, resulting in accumulation or incomplete metabolism in the body and the formation of drug residues, causing great harm to consumers' health. Rapid detection is an important way to ensure food safety.

The detection methods of veterinary drugs are mostly based on instrumental detection methods such as high performance liquid chromatography and high performance liquid chromatography-mass spectrometry. However, the instruments used in these detection methods are expensive, the detection process is cumbersome, time-consuming, and requires professional operations, so it is not applicable. For rapid on-site screening of large batches of samples. Colloidal gold immunochromatography test strips are a commonly used rapid detection method for veterinary drug residues in recent years. This method is suitable for on-site rapid screening of large quantities of samples. However, the detection sensitivity is low, the false positive/false negative rate is high, and it is often affected by Due to the influence of solvents and the interference of food matrices, there is an urgent need to establish a high-throughput, highly sensitive, and on-site rapid detection method for veterinary drug residues.

Surface-enhanced Raman spectroscopy (SERS) is a powerful molecular vibration spectroscopy technology that utilizes the greatly enhanced Raman scattering effect of surface nanostructures of precious metals such as gold and silver to detect molecules adsorbed on the surface with extremely high sensitivity. SERS has high sensitivity, fast detection speed, flexible and convenient detection, and can provide information such as molecular "fingerprint" patterns. In recent years, its application in chemical and biological sensing, materials science, medical diagnosis, food safety and other fields has received great attention. SERS detection methods can be divided into non-label detection methods based on target characteristic peaks, and labeled detection methods based on Raman probe molecules. By combining with target-specific recognition elements such as aptamers and antibodies, SERS labeling detection can achieve ultra-sensitive detection of targets. Among them, the competitive SERS aptamer sensor is a typical SERS marker detection method. It has been used to detect various food risk factors such as pesticides and veterinary drugs, hormones, toxins, and microorganisms, and has achieved good results. Although the competitive SERS aptamer sensor has good performance in detecting a drug, its detection efficiency is low and cannot meet the needs of rapid screening of large batches of samples. The rapid detection of two or more drugs at the same time is an urgent need for food safety testing. However, when the competitive SERS aptamer sensor is used to detect two drugs at the same time, there will be a gap between the signal probes of different drugs in the reaction system. They will interfere with each other and affect the Raman "hot spot" distribution of the entire system, making it impossible to accurately and quantitatively detect the contents of the two drugs. The problem of mutual interference between multiple Raman signal probes in the same system reduces the efficiency and accuracy of the competitive SERS aptamer sensor for simultaneous detection of multiple targets.

Contents of the invention

In order to make up for the shortcomings of the existing technology, the present invention provides a method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning.

The present invention is achieved through the following technical solution: a method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning, which specifically includes the following steps:

S1: Silver-coated gold nanoparticles (Au@AgNPs) modified with drug aptamers are used as signal probes, and gold-coated magnetic bead nanoflowers (Fe ₃ O ₄ @AuNFs) modified with complementary chains of aptamers are used as capture probes. Mix the signal probe and the capture probe evenly and incubate at 37°C to obtain a competitive SERS aptamer sensor for a single drug; in step S1, for different drugs, the surface of the gold core is modified with different Raman reporter molecules, and different The characteristic peaks of Raman reporter molecules do not overlap with each other;

S2: Mix the competitive SERS aptamer sensors for different drugs in S1 evenly and incubate them at 37°C for 8 hours to obtain a competitive SERS aptamer sensor that can detect two drugs at the same time;

S3: Mix the two drug solutions to be tested in different proportions to obtain drug mixtures containing different concentrations;

S4: Mix the drug mixture in S3 and the competitive SERS aptamer sensor in S2 evenly, incubate at 37°C for 1 hour, then separate by magnet, wash 3 times with PBS buffer, and resuspend in 15 μL in PBS buffer;

S5: Add the resuspension solution in S4 dropwise on the surface of the silicon wafer, aim the Raman probe at the surface of the solution, collect the Raman spectrum of the solution, preprocess the Raman spectrum, and compare the collected N drug mixtures with different concentrations. Raman spectrum is used as a database to build a machine learning model;

S6: Use the machine learning model established in S5 to predict the contents of the two drugs in the solution to be tested.

As a preferred solution, the gold core of the silver-coated gold nanoparticles in step S1 can be spherical, rod-shaped, regular polyhedron, star, etc., the particle size of Au@AgNPs is 30-60 nm, and the particle size of Fe ₃ O ₄ @AuNFs is 200-500 nm. , and the drug aptamer and its complementary chain modify the thiol group at the 5' end, and connect it to the surface of the nanomaterial through the thiol group. The incubation time of the signal probe and capture probe is 1 to 2 hours;

As a preferred solution, the mixing ratio of the two drugs' competitive SERS aptamer sensors in step S2 is 1:1.

As a preferred solution, the concentration range of the two drugs in step S3 is 0.001~1000 μg/L, and the two drug mixtures used for modeling contain different concentration combinations as much as possible.

As a preferred solution, in step S4, the volume ratio of the drug mixture to the competitive SERS aptamer sensor is 1:1, the pH of the PBS buffer is 7.4, and the concentration is 10 mM.

As a preferred solution, the amount of solution dropped on the surface of the silicon wafer in step S5 is set to 5~10 μL, and a laser with a wavelength of 785 nm is used to obtain a Raman spectrum. The detection range is 400 - 3200 cm ^-1 and the number of spectra is N> 100.

As a preferred solution, the preprocessing of the Raman spectrum in step S5 includes baseline correction and smoothing filtering, and the spectral interval containing the characteristic peaks of the signal molecules is subjected to maximum-minimum normalization processing, and then the PyTorch framework is used to establish a machine learning model .

As a preferred solution, after the model prediction is completed in step S5, denormalization processing needs to be performed to obtain the actual concentration value, and the test data set is used to further evaluate its performance.

Since the present invention adopts the above technical solution, it has the following beneficial effects compared with the existing technology: The present invention proposes a new, fast, simple, sensitive and accurate method for detecting two drug residues at the same time, which is based on silver. Gold-coated nanoparticles and gold-coated magnetic bead nanoflower materials are used to construct a competitive SERS aptasensor, which uses magnets to quickly magnetically separate the SERS substrate, simplifying the experimental steps; the constructed competitive SERS aptasensor can achieve For simultaneous rapid detection of two targets, the accuracy of simultaneous detection of two targets is further improved by combining machine learning methods with competitive aptamer detection; by changing aptamers and Raman signal molecules, This detection mode can be expanded to the simultaneous high-sensitivity detection of three or even multiple targets, which has important development significance for improving the detection efficiency of multiple drug residues in food and the environment.

Additional aspects and advantages of the invention will be apparent from the description which follows, or may be learned by practice of the invention.

Description of the drawings

The above and/or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

Figure 1: (a) TEM image of Fe ₃ O ₄ @AuNFs, (b) TEM image of Au@AgNP

Figure 2: Artificial neural network model performance evaluation for chloramphenicol and estradiol: (ad) ^R values for training, validation, test and overall data sets respectively, (e) actual versus predicted values for chloramphenicol Histogram of, (f) Histogram of actual and predicted values of estradiol, (g) model training loss and validation loss

Figure 3: Random forest model performance evaluation of chloramphenicol and estradiol: (ac) ^R2 values for training, test, and validation sets respectively (d), histogram of actual and predicted values of chloramphenicol ( e), histogram of actual and predicted values of estradiol

Figure 4: Quantitative model of chloramphenicol and estradiol: (a) Raman spectra of different chloramphenicol concentrations at an estradiol concentration of 0.01 μg/L, (b) chloramphenicol at a displacement of 1075 cm ^-1 Quantitative regression curve, (c) Raman spectrum of different estradiol concentrations when the chloramphenicol concentration is 0.01 μg/L, (d) Regression curve for estradiol quantification at a displacement of 1330 cm ^-1

Figure 5: Theoretical simulation of competitive SERS aptamer sensor Raman hotspot changes by finite difference method: (a) xy view of the model without adding drugs, (b) xy view of the model after adding drugs, (c) without Electric field simulation results of the model when adding drugs, (d) Electric field simulation results of the model after adding drugs

Detailed ways

In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, as long as there is no conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

Many specific details are set forth in the following description to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described here. Therefore, the protection scope of the present invention is not limited to the specific implementation disclosed below. Example limitations.

The method of simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to the embodiment of the present invention will be described in detail below with reference to FIGS. 1 to 5 .

Example 1:

This embodiment provides a method for simultaneously detecting chloramphenicol and estradiol based on a competitive SERS aptamer sensor and an artificial neural network, including the following steps:

S1: Using spherical Au@AgNPs with a particle size of about 40 nm modified with chloramphenicol and estradiol aptamers (as shown in Figure 1a) as a signal probe, modifying the complementary chains of chloramphenicol and estradiol aptamers. Fe ₃ O ₄ @AuNFs with a particle size of about 300 nm (shown in Figure 1b) is used as a capture probe, and signal probes for chloramphenicol and estradiol are modified with 4-mercaptobenzoic acid and 5,5'-disulfide respectively. Bis(2-nitrobenzoic acid) is used as a Raman reporter molecule; mix the signal probe and the capture probe evenly, and incubate them at 37°C for 1 hour to obtain a competitive SERS aptamer sensor for a specific drug;

S2: Mix the competitive SERS aptamer sensors for chloramphenicol and estradiol in S1 at a ratio of 1:1 and incubate at 37°C for 8 hours to obtain a sensor that can simultaneously detect chloramphenicol and estradiol. Competitive SERS aptamer sensor;

S3: Mix two drug solutions of chloramphenicol and estradiol with a concentration range of 0.001~1000 μg/L according to different proportions. The ratio of chloramphenicol and estradiol in the mixed solution is chloramphenicol (μg/L) :Estradiol (μg/L)=100:500, 80:6200, 100:2000, 0.1:700, 50:50, 50:400, 300:700, 20:600, 0.01:0.01, 10:800, 5:200, 10:100, 100:300, 10:10, 1:500, 5:100, 1:1200, 0.1:400, 1200:10, 10:10, 1:500, 60:130, 50: 300, 5:50, 1:1, 20:200, 60:130, 1200:20, 5:5, 50:0.05, 70:80, 100:700, 10:400, 20:1000, 4300:1500, 1220:1500, 2000:3200, 2000:10000, 1500:3200, 3000:1800, 30:200, 130:1500, 120:80, 500:3300, 300:100, 1:100, 10:60, 1000: 100, 1000:1000 to obtain a drug mixture containing different concentrations of chloramphenicol and estradiol;

S4: Mix the drug mixture in S3 and the competitive SERS aptamer sensor of the two drugs in S2 at a ratio of 1:1, incubate at 37°C for 1 hour, then separate with a magnet, and use PBS buffer Wash three times and then resuspend in 15 μL of 10mM PBS buffer (pH=7.4);

S5: Take 5 μL of the resuspension in S4 and drop it on the surface of the silicon wafer. Aim the Raman probe at the surface of the solution, collect the Raman spectrum of the solution, preprocess the Raman spectrum, and mix the collected 279 drugs of different concentrations. The Raman spectrum of the liquid is used as a database to build a machine learning model. The machine learning model used in this embodiment is an artificial neural network model. The specific modeling process is as follows:

(1) Data preprocessing: First perform baseline correction and smoothing filter preprocessing on the Raman spectrum, and then perform maximum-minimum normalization processing on the peaks in the spectral interval 1000-1400 cm ^-1 to ensure that all feature values are within [ 0, 1], the normalization formula is as follows:

(1)

Among them, the minimum value and maximum value in formula (1) are the minimum and maximum values of the feature respectively.

(2) Model construction and configuration: Use the PyTorch framework to build an artificial neural network model. The architecture of the model includes several fully connected layers, and activation functions are inserted between these layers to increase the nonlinear expression ability of the network. Specifically, the input layer contains 400 nodes, corresponding to the number of features of the data. The model contains four hidden layers with 312, 144, 72 and 36 nodes respectively. Each layer is followed by a LeakyReLU activation function. The final layer has two nodes that predict the concentrations of two chemicals.

(3) Training parameters and process: Select the mean square error (MSE) as the loss function of the error, and perform error backpropagation and weight adjustment through the stochastic gradient descent (SGD) algorithm (momentum is 0.9, learning rate is 0.001). The data set is divided into 70% training data, 15% validation data, and 15% testing data. The model was trained for 500 iterations. In order to prevent overfitting, an early stopping patience value of 20 iterations was set by monitoring the verification loss.

(4) Model prediction ability evaluation: After training is completed, use the R² value to evaluate the model prediction ability on the test set. As shown in Figure 2, the R² of the training, validation, testing and overall data sets are 0.959, 0.991, 0.976 and 0.970 respectively. The loss trajectory during the training process shows effective learning and no obvious overfitting, indicating that the method It has excellent performance for predicting the content of two drugs at the same time. All calculations were performed in a Python 3.8 environment, using the TensorFlow 2.x framework.

S6: Use the machine learning model established in S5 to predict the content of the two drugs in the sample to be tested, perform denormalization processing to obtain the actual concentration value, and qualitatively evaluate the model by comparing the predicted concentration value and the actual value of the test set. Accuracy, as shown in Figure 2, the predicted value and the actual value are in good agreement, with an average absolute error of 244 μg/L.

Example 2

This embodiment provides a method for simultaneously detecting chloramphenicol and estradiol based on a competitive SERS aptamer sensor and a random forest model. The detection steps in this embodiment are the same as those in Embodiment 1. The difference is that the machine learning model used is Random forest model, the specific modeling process is as follows:

(1) Data preprocessing: First, the Raman spectrum baseline correction and smoothing filtering are preprocessed, and then the peaks in the spectral interval 1000-1400 cm ^-1 are subjected to maximum-minimum normalization processing to ensure that all eigenvalues are within [0 , 1] within the range.

(2) Model construction: Choose random forest, which is an ensemble learning method that generates the final prediction result by building multiple decision trees and integrating their outputs.

(3) Multi-output regression processing: In order to be able to handle multi-output regression problems, the random forest model is encapsulated using MultiOutputRegressor to ensure that the model can provide predictions for multiple target variables.

(3) Model parameters and training: The random forest model is built based on 100 decision trees. These trees are trained on different data subsets. This strategy of randomly selecting data increases the diversity of the model, thereby improving the performance of the model. Generalization performance.

(4) Performance evaluation: After completing training, use R² to evaluate the model performance on the training, validation and test data sets. As shown in Figure 3, the R² on the training, validation, and test data sets are 0.959, 0.580, and 0.796 respectively. It can be seen that the model shows relatively robust performance. In order to further verify the accuracy of the model, 10 specific samples were selected for concentration prediction, and the agreement between the predicted values and the actual values was good. This example shows that random forest, as an integrated learning method, can be combined with a competitive SERS aptamer sensor to achieve simultaneous detection of two drugs, but the detection performance is slightly worse than that of artificial neural networks.

Example 3

This embodiment provides a method for detecting chloramphenicol and estradiol based on a competitive SERS aptamer sensor. The detection steps of this embodiment are the same as those of Embodiment 1. The difference is that the model used is established through the intensity of Raman characteristic peaks. Linear model, the specific modeling process is as follows:

(1) Spectral data collection: The material preparation process is the same as steps S1 and S2 in Example 1. In the detection system, the added amount of estradiol is fixed at 0.01 μg/L, and the added amounts of chloramphenicol are 0.001 μg/L, 0.01 μg/L, 0.1 μg/L, 1 μg/L, 10 μg/L, and 100μg/L and 1000μg/L. Similarly, the addition amount of chloramphenicol is fixed at 0.01 μg/L, and the addition amount of estradiol is 0.001 μg/L, 0.01 μg/L, 0.1 μg/L, 1 μg/L, 10 μg/L, 100 μg/L and 1000 μg/L. SERS detection conditions are the same as in Example 1.

(2) Model establishment: For the detection of chloramphenicol, the Raman characteristic peak intensity of 4-mercaptobenzoic acid at 1075 cm ^-1 was selected for linear regression detection of chloramphenicol. For the detection of estradiol, the Raman characteristic peak intensity of 5,5'-dithiobis(2-nitrobenzoic acid) at 1330 cm ^-1 was selected for linear regression detection of estradiol.

As shown in Figure 4, when the estradiol content is fixed and the chloramphenicol content is different, the Raman reporter molecule 5,5'-disulfobis(2-nitrobenzoic acid) used to detect estradiol is The Raman signal intensity has a large difference. For quantitative analysis of chloramphenicol, the correlation coefficient of linear regression is only 0.6763. Similarly, when the chloramphenicol content is fixed and the estradiol content is different, the Raman signal intensity of the Raman reporter molecule 4-mercaptobenzoic acid used to detect chloramphenicol has a large difference, and estradiol is quantified. Analysis, the correlation coefficient of its linear regression is only 0.8146. The results show that the detection accuracy of using the traditional linear regression model to detect two compounds simultaneously is poor.

Example 4

This embodiment uses the finite difference time domain method to theoretically simulate the changing rules of the Raman "hot spots" of the competitive SERS aptasensor after adding drug molecules, as shown in Figure 5. (a) and (b) are before and after the drug is added. XY cross-section of the constructed model. (c) and (d) are the electric field simulation results of the model before and after adding the drug, in which the color chart scale from dark to light represents the number of Raman "hot spots". It can be seen from the simulation results that when no drug is added, the signal probe and the capture probe are combined, and the Raman "hot spot" is the highest at this time. However, when one of the drugs is added, due to the specific recognition of the aptamer, the signal probe and capture probe originally bound together dissociate, resulting in a reduction in the overall Raman "hot spots" in the system. Therefore, the Raman signals of the two signal probes in the detection system will affect each other. In the actual testing environment, since the contents of the two drugs are unknown, it is impossible to use the traditional linear regression model to perform accurate quantitative analysis of the two drugs at the same time.

In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiments," etc., mean that a particular feature, structure, material or characteristic described in connection with the embodiment or example is included in the invention. in at least one embodiment or example. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (8)

1. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning, which is characterized by including the following steps: S1: Silver-coated gold nanoparticles (Au@AgNPs) modified with drug aptamers are used as signal probes, and gold-coated magnetic bead nanoflowers (Fe ₃ O ₄ @AuNFs) modified with complementary chains of aptamers are used as capture probes. Mix the signal probe and the capture probe evenly and incubate them at 37°C to obtain a competitive SERS aptamer sensor for a single drug; S2: Mix the competitive SERS aptamer sensors for different drugs in S1 evenly and incubate them at 37°C for 8 hours to obtain a competitive SERS aptamer sensor that can detect two drugs at the same time; S3: Mix the two drug solutions to be tested in different proportions to obtain drug mixtures containing different concentrations; S4: Mix the drug mixture in S3 and the competitive SERS aptamer sensor in S2 evenly, incubate at 37°C for 1 hour, then separate by magnet, wash 3 times with PBS buffer, and resuspend in 15 μL in PBS buffer; S5: Add the resuspension solution in S4 dropwise on the surface of the silicon wafer, aim the Raman probe at the surface of the solution, collect the Raman spectrum of the solution, preprocess the Raman spectrum, and compare the collected N drug mixtures with different concentrations. Raman spectrum is used as a database to build a machine learning model; S6: Use the machine learning model established in S5 to predict the contents of the two drugs in the solution to be tested. 2. A method for simultaneously detecting two veterinary drugs based on SERS labeling detection and machine learning according to claim 1, characterized in that the gold core of the silver-coated gold nanoparticles in step S1 can be spherical, rod-shaped, or normal. Polyhedral, star and other shapes, Au@AgNPs particle size 30-60 nm, Fe ₃ O ₄ @AuNFs particle size 200-500 nm, and the drug's aptamer and its complementary chain modify the thiol group at the 5' end, through the thiol group It is connected to the surface of the nanomaterial, and the incubation time of the signal probe and capture probe is 1 to 2 hours. 3. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to claim 1, characterized in that the competitive SERS aptamer sensor mixing ratio of the two drugs is 1:1. . 4. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to claim 1, characterized in that the concentration range of the two drugs in step S3 is 0.001~1000 μg/L. . 5. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to claim 1, characterized in that in step S4, the volume ratio of the drug mixture to the competitive SERS aptamer sensor is 1:1, PBS buffer pH 7.4, concentration 10 mM. 6. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to claim 1, characterized in that the amount of solution dropped on the surface of the silicon wafer in step S5 is defined as 5~ 10 μL, and use a laser with a wavelength of 785 nm to obtain a Raman spectrum. Its detection range is 400 - 3200 cm ^-1 and the number of spectra is N>100. 7. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to claim 1, characterized in that the preprocessing of the Raman spectrum in step S5 includes baseline correction and smoothing. Filter and perform maximum-minimum normalization on the spectral interval containing the characteristic peaks of signal molecules, and then use the PyTorch framework to build a machine learning model. 8. A method for simultaneously detecting two veterinary drugs based on SERS marker detection and machine learning according to claim 1, characterized in that after the model prediction in step S5 is completed, denormalization processing is performed to obtain Actual concentration value.