Preparation method of surface-enhanced raman spectrum substrate for detecting multi-sulfonamide drug residues
A surface-enhanced Raman spectroscopy substrate was prepared by combining purple phosphorus with gold nanoparticles. By combining it with a one-dimensional convolutional neural network, the problems of purple phosphorus stability and the shortcomings of traditional machine learning were solved, and efficient detection and accurate classification and quantification of polysulfonamide drug residues were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGXI AGRICULTURAL UNIVERSITY
- Filing Date
- 2022-12-23
- Publication Date
- 2026-05-22
AI Technical Summary
In existing technologies, the application of two-dimensional materials such as graphene and black phosphorus in nano-optoelectronics and electronic sensors is limited. Purple phosphorus as a substrate for surface-enhanced Raman spectroscopy has problems with stability and signal distortion. Traditional machine learning algorithms rely on feature selection, which leads to insufficient accuracy in the detection of sulfonamide drug residues.
A surface-enhanced Raman spectroscopy substrate was prepared using a purple phosphorus and gold nanoparticle composite. The detection of polysulfonamide drug residues was performed by combining a one-dimensional convolutional neural network (1-DCNN). The detection conditions were optimized and a deep learning model was established by using SAuNPs/VP seeds and AuNPs/VP composites.
It achieves high sensitivity, strong stability, and high throughput of spectral analysis for the detection of polysulfonamide drug residues, possesses sensitive and reliable identification and quantification capabilities, and the 1-DCNN model achieves 100% classification accuracy and high linear correlation.
Smart Images
Figure CN117169187B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of Raman detection technology, and in particular to a method for preparing a surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues. Background Technology
[0002] Two-dimensional (2D) layered materials, including graphene, black phosphorus (BP), and other graphene-like materials, have attracted increasing attention due to their unique photonic, mechanical, biocompatible, and electronic properties. However, graphene's zero band gap and fingerprint region (600-1800 cm⁻¹) remain significant challenges. -1 The strong background signal of BP limits its practical application in nano-optoelectronics and electronic sensors. Compared to graphene, BP is a p-type semiconductor with a layered structure, exhibiting a larger direct bandgap, and only at 500 cm⁻¹. -1 Several characteristic peaks are shown below. Even so, BP degrades rapidly under light, oxygen, humidity, and temperature conditions, which often hinders its stripping and characterization, thus limiting its practical applications.
[0003] Purple phosphorus, an allotrope of black phosphorus, has attracted worldwide attention since its synthesis in 2019 due to its unique physical properties and potential applications in many fields. In particular, purple phosphorus has demonstrated many excellent properties in the fabrication of surface-enhanced Raman scattering (SERS) substrates, including: (I) each layer possesses only 0.3-0.4 J / m². 2 (I) The binding energy indicates the possibility of exfoliation in experiments. (II) The pyrolysis temperature of purple phosphorus is 52℃ higher than that of black phosphorus, making it the most stable phosphorus isotope. Therefore, it provides a basis for photonics, electronics, and semiconductor applications. (III) The excellent electron transfer rate makes VP an excellent electromagnetic enhancement material, which can greatly increase the response of surface-enhanced Raman spectroscopy, also known as chemical enhancement (CM). (IV) Given the large surface area of two-dimensional materials, well-controlled nanosheets can act as a barrier to prevent direct contact between noble metal nanoparticles and analytes, which can fundamentally prevent the possibility of distortion in surface-enhanced Raman spectroscopy. (V) A layer of purple phosphorene composed of cross-linked sub-nanorods has a Young's modulus of 1512+76 Nm. -1 Its strength is 4.4 times higher than that of graphene and far exceeds that of other reported two-dimensional nanostructures, verifying its extremely high mechanical properties. Therefore, purple phosphorus has the potential to replace other two-dimensional materials in the preparation of surface-enhanced Raman spectroscopy substrates.
[0004] Nevertheless, the electron transfer provided by two-dimensional materials can only reach up to 10. 3 The electromagnetic enhancement (EM) is limited by the chemical composition of the analyte and the interaction between the analyte and the substrate. Noble metal nanomaterials, as major contributors to Raman response, have attracted considerable attention, with their electromagnetic enhancement (EM) reaching 10 times the standard value.6 Among them, gold nanoparticles, with their extremely strong stability and enhancing properties, have become an ideal choice for nanomaterials. However, metal nanostructures are easily oxidized, making it difficult to maintain long-term stability, and the interaction between metals and molecules can cause signal distortion and molecular deformation, all of which greatly limit their applications. As a popular trend, combining flexible substrates with nanomaterials as ideal support structures has proven to be an attractive option. This not only prevents the aggregation of noble metal nanoparticles but also improves the response of surface-enhanced Raman spectroscopy. Furthermore, when the surface-enhanced Raman spectroscopy substrate achieves a combination mechanism of CM and EM, a synergistic effect is generated, resulting in an even higher surface-enhanced Raman spectroscopy enhancement effect.
[0005] Raman spectroscopy typically involves over 1000 Raman bands, containing a wealth of sample information. In previous work, selecting the strongest peaks in the surface-enhanced Raman (SMR) spectra of a substance for analysis has become a fixed approach for trace detection using SMR, and it has been widely used to detect residues of certain sulfonamides. While this approach can achieve low detection limits, it also results in significant information loss. Furthermore, substances of the same type often share similar vibrational spectra, making it impossible to quantify or classify spectral differences invisible to the naked eye. Recently, the development of machine learning techniques has provided a new strategy to overcome this problem. By combining traditional SMR detection with machine learning (ML)-driven spectral analysis, SMR has become a new intelligent technique with automation, predictability, robustness, and higher accuracy. Algorithms such as Support Vector Machines (SVM), Random Forests (RF), and Artificial Neural Networks (ANN) have successfully achieved spectral classification and quantification of unknown substances. However, traditional machine learning algorithms rely on the selection of effective features. Deep learning, a branch of machine learning, leverages the power of big data by extracting features from the machine itself, demonstrating outstanding performance in nonlinear tasks that are difficult to analyze mathematically. One-dimensional convolutional neural networks (1-D CNNs) are among the most popular deep learning algorithms. They combine preprocessing, feature extraction, and classification / regression into a single architecture and are trained end-to-end in a hierarchical manner without manual tuning. Their modular structure simplifies model development and achieves higher accuracy than traditional methods in classification tasks. In recent years, numerous works have achieved breakthroughs with the help of 1-D CNNs, such as basal body identification and cancer diagnosis, as well as the determination of the concentration of unknown additives. Summary of the Invention
[0006] To address the shortcomings of existing technologies, the present invention aims to provide a method for preparing a surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues. The substrate of the present invention has the advantages of simple preparation, high detection sensitivity, strong preservation stability, and high spectral analysis throughput, and can become a sensitive, reliable, and effective means for the identification and quantification of various sulfonamide residues.
[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:
[0008] On the one hand, the present invention provides a method for preparing a surface-enhanced Raman spectroscopy (SERS) substrate for detecting polysulfonamide drug residues, comprising two steps: the synthesis of SAuNPs / VP seeds and the synthesis of AuNPs / VP complex substrates.
[0009] 1. Preparation of SAuNPs seeds by in-situ deposition method, as detailed below:
[0010] (1) Add 1-10 mg of VP crystals and 4.45 mL of deionized water to a 50 mL double-necked flask and sonicate under nitrogen to remove residual oxygen in the solution and the double-necked flask.
[0011] (2) Add 10-100 μL of HAuCl4·3H2O solution (5 mM) to provide gold ions, and sonicate under nitrogen purging to completely remove residual oxygen in the HAuCl4·3H2O solution and the double-necked flask.
[0012] (3) After the timing is over, quickly seal the bottle opening with a sealing film to prevent oxygen from entering, and continuously sonicate to provide an environment in which VP can be completely stripped and gold ions can be completely reduced.
[0013] (4) Finally, the VP-small-particle gold nanoparticle composite (SAuNPs / VP) seeds were obtained and stored in a refrigerator at 4°C for later use.
[0014] In the SAuNPs seed preparation process, in step (1), VP crystals are ultrasonically treated for 1-10 minutes under nitrogen purging; in step (2), HAuCl4·3H2O is added and ultrasonic treatment is continued for 10-30 minutes under nitrogen purging; in step (3), the bottle mouth is sealed and ultrasonic treatment is continued for 5-12 hours.
[0015] 2. AuNPs / VP complex substrates were prepared using a seed-mediated method, as detailed below:
[0016] (1) Add 70-140 μL of the above-synthesized seed solution to a 20 mL glass headspace flask, dilute with 705.5 μL of water, add 50-300 μL of HAuCl4·3H2O (1%) solution to provide gold ions, stir at 300-700 rpm for 1-3 min to make it uniform, and obtain the seed growth solution.
[0017] (2) Subsequently, 0.5-3 mL of L-ascorbic acid (10 mM) was added dropwise to the above solution and stirred at 300-700 rpm for 10-40 min to provide an environment for seed growth;
[0018] (3) After the timing is completed, a gold nanoparticle / phosphorus complex (AuNPs / VP) substrate with a larger diameter is obtained and stored in a refrigerator at 4°C for further use.
[0019] On the other hand, the present invention provides a method for detecting polysulfonamide drugs, including the optimization of the method and the establishment of a detection method for polysulfonamide drugs.
[0020] The optimization of the method is as follows:
[0021] (1) Dissolve the analyte SM2 in different solvents to prepare a 5 μg / mL solution, and check the response value of SM2 at the characteristic peak;
[0022] (2) Subsequently, the AuNPs / VP complex substrate was mixed with the above solution at ratios of 1:8, 1:4, 1:2, 1:1, and 2:1 in 500 μL centrifuge tubes, and the response value of SM2 at the characteristic peak was examined.
[0023] In the process of optimizing the sensing method, in step (1), the selected solvents are acetic acid (10mM), acetonitrile, and sodium hydroxide (10mM); the characteristic peak selected in steps (1) and (2) is 591 cm⁻¹. -1 826cm -1 1003cm -1 1125cm -1 1597 cm -1 .
[0024] The establishment of the detection method for the polysulfonamide drugs includes quantitative analysis and qualitative analysis.
[0025] The qualitative analysis of the polysulfonamide drugs is as follows:
[0026] (1) Three sulfonamide drugs, sulfadiazine (SM2), sulfadiazine (SD), and sulfamethoxazole (SMZ), were measured 8 times at gradient concentrations (0.005-10 μg / mg), yielding a total of 192 raw spectra. Savitzky-Golay (SG) spectral preprocessing was used to remove noise from the raw spectra and improve spectral resolution. A one-dimensional convolutional neural network model (1-DCNN) was established as a reference dataset for the qualitative and quantitative analysis of polysulfonamide drugs.
[0027] (2) Select 70% of the spectrum in the reference dataset for the training set and 30% of the spectrum for the test set; construct a multi-channel deep learning model based on a convolutional neural network and train the model using the training set;
[0028] (3) Subsequently, the sulfonamide drugs with the above three gradient concentrations are classified through one input layer, six convolutional layers, three pooling layers, one planarization layer and one fully connected layer.
[0029] In the establishment of the qualitative detection method for polysulfonamide drugs, in step (2), the order of the data was randomly shuffled during the division of the training set and the test set in order to further improve the generalization ability of the qualitative training model; in step (3), two traditional machine learning methods, principal component analysis (PCA) and t-distributed random adjacency embedding (t-SNE), were used to reduce the dimensionality of the data. Compared with the deep learning method 1-DCNN, it was found that 1-DCNN had the best classification effect with a classification accuracy of 100%.
[0030] The quantitative analysis of the polysulfonamide drugs is as follows:
[0031] (1) 192 surface-enhanced Raman spectra obtained from gradient concentrations (0.005-10 μg / mL) of sulfadimethylpyrimidine (SM2), sulfadiazine (SD) and sulfamethoxazole (SMZ) were preprocessed using Savitzky-Golay (SG) spectra to remove original spectral noise and improve spectral resolution.
[0032] (2) 70% of the 64 spectra at each gradient concentration of the analyte were used for the training dataset and 30% of the spectra were used for the test dataset.
[0033] (3) Based on Aheto's discussion, determine whether the model is robust. Aheto's discussion states: RPD > 2.0 (Rp2 > 0.75) indicates that the model is robust; RPD = 1.4 - 2.0 (Rp2 = 0.49 - 0.74) indicates that the model is usable; RPD < 1.4 (Rp2 < 0.49) indicates that the model has no predictive ability.
[0034] In the above scheme, within the range of 0.005-10 μg / mL, the classification model for sulfadiazine (SM2), sulfadiazine (SD), and sulfamethoxazole (SMZ) achieved a 100% accuracy rate, which is superior to PCA and t-SNE. In addition, the linear correlation Rp2 of the gradients of the three sulfonamide drugs is greater than 0.9786, and the prediction residual bias RPD is greater than 6.3484, indicating that the 1-DCNN quantitative model of this invention is robust.
[0035] Compared with the prior art, the present invention has the following beneficial effects:
[0036] The AuNPs / VP complex substrate prepared by this invention and the established detection method not only have the characteristics of low sensitivity, strong preservation stability and high spectral analysis throughput, but can also detect a variety of sulfonamide residues. It is a highly sensitive, reliable and effective means of identification and detection. Attached Figure Description
[0037] Figure 1 A schematic diagram of a method for preparing a surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues;
[0038] Figure 2 AB and Figure 2 Morphological characterization and corresponding particle size statistics of the SAuNPs / VP seeds and AuNPs / VP composite substrates prepared in Example 1 of this invention.
[0039] Figure 3 Optimization of the substrate and method in Example 2 of this invention: (A) Monitoring of the substrate assembly process; (B) Optimization of the solvent; (C) Optimization of the amount of VP crystals; (D) Optimization of the volume ratio of substrate to analyte.
[0040] Figure 4 A. SERS spectra of different concentrations of SM2 in the range of 0.005-10 μg / mL (curves b to i) in Example 3 of this invention, and SERS spectra of SM2 with zero addition (curve a). Figure 4 B. The linear relationship between the SERS intensity and the logarithm of the concentration of SM2 in this invention;
[0041] Figure 5 In Example 4 of this invention, the confusion matrix (C) of PCA (A), t-SNE (B), and 1-DCNN results for identifying different concentrations of polysulfonamide drugs of various types and concentrations is presented.
[0042] Figure 6SERS spectra of SM2(A), SD(B), and SMZ(C) at different concentrations (0.005-10 μg / mL) after SG treatment in Example 4 of this invention (a); Relationship between actual and predicted values of SM2(A), SD(B), and SMZ(C) of the 1-DCNN model on the training and test sets (b); Detailed Implementation
[0043] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.
[0044] Unless otherwise specified, all raw materials and reagents used in this invention can be obtained commercially.
[0045] This invention provides a method for preparing a surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues, such as... Figure 1 The schematic diagram illustrates the substrate preparation method and the detection and application method, with specific embodiments as follows.
[0046] Example 1
[0047] Synthesis of AuNPs / VP complex substrate, detailed procedure is described in [link to procedure]. Figure 1 .
[0048] First, SAuNPs / VP seeds were prepared: 3 mg of VP crystals were weighed and added to a 50 mL double-necked flask, followed by 4.45 mL of deionized water. A stable nitrogen flow rate and strong ultrasonic exfoliation were maintained for 2 minutes. 50 μL of HAuCl4·3H2O solution (5 mM) was added, and exfoliation was carried out for another 20 minutes. The flask was quickly sealed with a sealing film to isolate oxygen, and continuous ultrasonic treatment was performed for 7 hours to obtain small-sized gold nanoparticle-modified phosphane complex (SAuNPs / VP) seeds.
[0049] Next, the synthesis of the AuNPs / VP complex substrate was carried out: 705.5 μL of deionized water and 125 μL of the above seed solution were placed in a 20 mL glass headspace flask, and 144.5 μL of HAuCl4·3H2O (1%) aqueous solution was added and mixed (500 rpm, 1 min). Then, 1.025 mL of L-ascorbic acid (10 mM) was added dropwise to the above solution and the mixture was stirred for 15 min to obtain a nanocomposite (AuNPs / VP) with a larger diameter of gold nanoparticles.
[0050] Comparison of SAuNPs / VP and AuNPs / VP scanning electron microscope images ( Figure 2 A and Figure 2 C), by Figure 2 B and Figure 2 D, Gold nanoparticles on the composite material grow from 10 nm to 27 nm.
[0051] Example 2
[0052] Optimization of AuNPs / VP complex substrate and detection method:
[0053] (1) Dissolve the analyte SM2 in different solvents to prepare a 5 μg / mL solution, and check the response value of SM2 at the characteristic peak;
[0054] (2) Subsequently, the AuNPs / VP complex substrate was mixed with the above solution at ratios of 1:8, 1:4, 1:2, 1:1, and 2:1 in 500 μL centrifuge tubes, and the response value of SM2 at the characteristic peak was checked.
[0055] In the process of optimizing the sensing method, in step (1), the selected solvents are acetic acid (10mM), acetonitrile, and sodium hydroxide (10mM); the characteristic peak selected in steps (1) and (2) is 591 cm⁻¹. -1 826cm -1 1003cm -1 1125cm -1 1597 cm -1 .
[0056] Test results are shown Figure 3 Specifically: (1) Optimization of the AuNPs / VP complex substrate: such as Figure 3 As shown in Aa, the tin foil did not produce a Raman peak, therefore it did not interfere with the detection of the analyte. However, low concentrations of SM2 (5 μg / mL) also showed no peak. Figure 3 This necessitates the use of high-performance substrates. Monitoring the substrate assembly process revealed that neither SAuNPs / VP seeds nor AuNPs / VP composite substrates produced interfering peaks. Figure 3 Ab-Ac). Simultaneously, by comparing the SERS response of SAuNPs / VP and AuNPs / VP to SM2 (5 μg / mL) (... Figure 3 Ae-Af) revealed that only AuNPs / VPs responded to SM2, especially at 591 cm⁻¹. -1 826cm -1 1003cm -1 1125cm -1 and 1597cm -1 The response was very strong, so these five Raman bands were selected as characteristic peaks of SM2 for further analysis.
[0057] (2) Solvent optimization. Three classic solvents (acid, base, and organic solvent) were selected to explore their effects on SERS spectra. From... Figure 3B observed that, compared to acetic acid and acetonitrile, the SERS spectrum of SM2 using NaOH (10 mM) as solvent showed the highest SERS intensity at the selected characteristic peaks.
[0058] (3) Optimization of VP crystal quantity. From... Figure 3 As can be seen from C, the SERS intensity at the selected characteristic peaks initially increases with increasing VP crystal content, reaching a maximum at 3 mg of VP, and then decreases with further increases in VP crystal content. This may be because too little VP cannot facilitate electron transfer, while too much VP weakens LSPR, and 3 mg of VP produces the optimal effect of both enhancement mechanisms. Therefore, 3 mg of VP crystals was selected for subsequent analysis.
[0059] (4) Optimization of the volume ratio of substrate to analyte. From Figure 3 D observed that the SERS intensity at the selected characteristic peaks was proportional to the substrate-to-analyte volume ratio in the range of 1:8 to 1:2. The SERS response was strongest at a substrate-to-analyte volume ratio of 1:2, and then weakened as the volume ratio increased. Therefore, a substrate-to-analyte volume ratio of 1:2 was selected for subsequent analysis.
[0060] Example 3
[0061] Substrate performance analysis
[0062] (1) Preparation of standard solutions: First, dissolve 50 μg / mL stock solutions of three sulfonamide drugs (SM2, SD, and SMZ) in 10 mM NaOH solution, and then dilute with the same solution to a series of concentrations (10, 5, 1, 0.5, 0.1, 0.05, 0.01, 0.005 μg / mL). Then, introduce the standard analytes and substrate at different concentrations into a 500 μL centrifuge tube at a volume ratio of 2:1, and vortex continuously until homogeneous. Finally, transfer 6 μL of the above compound to a glass slide wrapped with aluminum foil and dry it in an oven at 60 °C for subsequent use.
[0063] (2) Raman detection: Place the glass slide wrapped in the dried tin foil onto the stage of the Raman spectrometer for Raman detection. The excitation wavelength of the Raman spectrometer is 785 nm, the integration time is 2 s, the laser power is 5 mW, and the spectral range is 2000-500 cm⁻¹. -1 .
[0064] (3) Construction of the SM2 standard curve: From Figure 4From A, the intensity of this region decreases with decreasing SM2 concentration; however, most characteristic peaks remain visible at molecular concentrations below 0.005 μg / mL, which is far below the EU's Maximum Residue Limit (MRL) of 100 μg / kg. Based on the characteristic peak (591 cm⁻¹) most likely to yield the best linear relationship in Raman spectra for different SM2 molecules... -1 A linear model was constructed to establish the relationship between intensity and concentration (LOGC). Figure 4 B), the fitted regression equation is Y = 376.46X + 768.28, R0 2 The value is 0.9429. It is worth noting that the LOD was calculated to be 0.0047 μg / mL.
[0065] (4) Calculation of substrate enhancement factor (EF): Based on the 0.5 μg / mL SM2 standard solution and solid SM2 at 591 cm⁻¹, respectively... -1 By comparing the surface-enhanced Raman spectrum and the ordinary Raman response at the substrate, the EF of the substrate was calculated to be 1.66 × 10⁻⁶. 6 The surface-enhanced Raman spectroscopy substrate of this invention achieves a combination mechanism of CM and EM, producing a synergistic effect and resulting in a higher surface-enhanced Raman spectroscopy enhancement effect.
[0066] Example 4
[0067] Performance evaluation of 1-DCNN models
[0068] (1) Qualitative Identification: First, three sulfonamide drugs at gradient concentrations (0.005-10 μg / mg): sulfadiazine (SM2), sulfadiazine (SD), and sulfamethoxazole (SMZ) were measured eight times at each concentration (10, 5, 1, 0.5, 0.1, 0.05, 0.01, 0.005 μg / mL), resulting in a total of 192 raw spectra. Second, Savitzky-Golay (SG) spectral preprocessing was used to remove noise from the raw spectra and improve spectral resolution, establishing a one-dimensional convolutional neural network model (1-DCNN) as a reference dataset for the qualitative identification of polysulfonamide drugs. Subsequently, 70% of the spectra in the reference dataset were selected for the training set, and 30% were used for the test set. Finally, the three gradient concentration sulfonamide drugs were classified using one input layer, six convolutional layers, three pooling layers, one planarization layer, and one fully connected layer. By examining the detailed classification accuracy of each sulfonamide drug shown in the confusion matrix, such as... Figure 5 As shown in C, the recognition accuracy is 100%.
[0069] Comparing two traditional machine learning methods: one is Principal Component Analysis (PCA), where the two-dimensional scatter distribution of the first two principal components (PCs) is as follows... Figure 5As shown in Figure A, the contribution rates of PC1 and PC2 were 52.5% and 25.0%, respectively, with a total contribution rate of 77.5%, indicating that both PC1 and PC2 reflected most of the information in the original data. However, at 0.005 μg / mL, the classification results of the PCA models for the three analytes overlapped. Another method is t-distributed random neighbor embedding (t-SNE), which, compared to PCA, showed a reduced degree of data clustering, but still revealed clustering of certain substances at certain concentrations. Figure 5 B).
[0070] The above results verify that the 1-DCNN model can achieve relatively accurate differentiation of multiple analytes and multiple concentrations, that is, it achieves qualitative differentiation of polysulfonamide drugs.
[0071] (2) Quantitative Analysis: Surface-enhanced Raman spectroscopy is a detection technique capable of semi-quantitative analysis. If 1-DCNN is introduced into the implementation process of surface-enhanced Raman spectroscopy spectral identification, a qualitative analysis model alone is insufficient. Under the premise of achieving 100% accuracy in distinguishing three sulfonamide drugs, a 1-DCNN quantitative model is established for the quantitative prediction of analyte concentration.
[0072] 70% of the spectra from 64 spectra at each sample's concentration gradient were selected for the training set, and 30% were used for the test set. The relationship between actual and predicted values is as follows: Figure 6 As shown in AC, the linearity between the true and predicted values of SM2 is very high, with a small deviation: Rp2 = 0.9983 and RPD = 7.8441. Figure 6 A). Compared to the 1-DCNN model, Figure 4 In section B, a conventional univariate linear relationship is established by selecting the Raman band most likely to yield the best linear relationship. Figure 4 B shows that it is still not as linear as the 1-DCNN model, therefore the 1-DCNN model has an absolute advantage in the quantitative processing of analytes.
[0073] Figure 6 B and Figure 6 C shows the training and prediction results of our 1-DCNN model based on SD and SMZ preprocessed data, respectively. We can see that SD has an Rp² of 0.9878 and an RPD of 6.9872. The quantitative performance of SMZ is similar to that of SM2 and SD (Rp² = 0.9786, RPD = 6.3484). In summary, all the above outputs demonstrate Rp² > 0.9786 and RPD > 6.3484, indicating that the model is robust, according to Aheto's discussion. This suggests that the model can predict sulfonamide drug concentrations well using surface-enhanced Raman spectroscopy sensors.
[0074] In summary, this invention provides a method for preparing a surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues. This method not only features low sensitivity, strong preservation stability, and high spectral analysis throughput, but also offers better linear fitting and the ability to classify various types and concentrations of sulfonamide drugs.
[0075] The above description is a preferred embodiment of the present invention. For those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.
Claims
1. A method for preparing a surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues, characterized in that, Includes the following steps: (11) Preparation of SAuNPs seeds by in-situ deposition method, including: (a1) Add 1-10 mg of VP crystals and 4.45 mL of deionized water to a 50 mL double-necked flask and sonicate under nitrogen purging to remove residual oxygen in the solution and the double-necked flask. (a2) Add 10-100 μL of 5 mM HAuCl4⋅3H2O solution to provide gold ions, and sonicate under nitrogen purging to completely remove residual oxygen in the HAuCl4⋅3H2O solution and the double-necked flask. (a3) After the timing is over, quickly seal the bottle opening with a sealing film to prevent oxygen from entering, and continuously sonicate to provide an environment in which VP can be completely stripped and gold ions can be completely reduced. (a4) Finally, the VP-small-particle gold nanoparticle composite (SAuNPs / VP) seeds were obtained and stored in a refrigerator at 4°C for later use; (12) Preparation of AuNPs / VP complex substrates by seed-mediated method, including: (b1) Add 70-140 μL of the above-synthesized seed solution to a 20 mL glass headspace flask, dilute with 705.5 μL of water, add 1 wt% 50-300 μL of HAuCl4⋅3H2O solution to provide gold ions, stir at 300-700 rpm for 1-3 min to make it uniform, and obtain the seed growth solution; (b2) Add 10 mM 0.5-3 mL L-ascorbic acid dropwise to the above solution and continue stirring at 300-700 rpm for 10-40 min to provide an environment for seed growth; (b3) After the timing is completed, the AuNPs / VP complex substrate is obtained and stored in a refrigerator at 4°C.
2. The method for preparing the surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues according to claim 1, characterized in that, In step (a1), the VP crystals are ultrasonically treated under nitrogen purging for 1-10 minutes; in step (a2), after adding HAuCl4⋅3H2O, the ultrasonic treatment is continued under nitrogen purging for 10-30 minutes; in step (a3), after sealing the bottle mouth, the ultrasonic treatment is continued for 5-12 hours.
3. A method for rapid detection of polysulfonamide drugs, characterized in that, The surface-enhanced Raman spectroscopy substrate for detecting polysulfonamide drug residues prepared using the method described in claim 1 or 2 specifically includes the following steps: Step 1: Prepare a polysulfonamide drug solution by mixing the AuNPs / VP complex substrate described in claim 1 or 2 with the polysulfonamide drug solution at a volume ratio of 1:2 and detecting its Raman spectrum. Step 2: Use Savitzky-Golay (SG) spectral preprocessing to remove noise from the original spectrum, improve spectral resolution, and build a one-dimensional convolutional neural network model (1-D CNN) as a reference dataset for qualitative and quantitative analysis of polysulfonamide drugs. Step 3: Select 70% of the spectrum in the reference dataset for the training set and 30% for the test set; construct a multi-channel deep learning model based on a convolutional neural network and train the model using the training set; Step 4: Qualitative and quantitative analysis of polysulfonamide drugs.
4. The method for rapid detection of polysulfonamide drugs according to claim 3, characterized in that, In step 1, the polysulfonamide drugs are sulfadiazine (SM2), sulfadiazine (SD), and sulfamethoxazole (SMZ); the solvent of the polysulfonamide drug solution is sodium hydroxide, and the concentration of the polysulfonamide drug is 0.005-10 μg / mL.
5. The method for rapid detection of polysulfonamide drugs according to claim 4, characterized in that, In step 3, the order of the data is randomly shuffled during the process of dividing the training set and the test set in order to further improve the generalization ability of the qualitative training model.
6. The method for rapid detection of polysulfonamide drugs according to claim 5, characterized in that, In step 4, the qualitative analysis of polysulfonamide drugs specifically involves: analyzing the three sulfonamide drugs at the above concentrations using one input layer, six convolutional layers, three pooling layers, one planarization layer, and one fully connected layer to establish a mixture matrix model of polysulfonamide drugs. The qualitative analysis of polysulfonamide drugs is achieved by checking the detailed classification accuracy of each sulfonamide drug displayed in the confusion matrix.
7. The method for rapid detection of polysulfonamide drugs according to claim 5, characterized in that, In step 4, the quantitative analysis of polysulfonamide drugs: After the qualitative analysis of polysulfonamide drugs is achieved, 70% of the spectra of 64 spectra at each sample gradient concentration are selected for the training set and 30% of the spectra are selected for the test set. The actual and predicted values of polysulfonamide drugs are linearly fitted to obtain the linear relationship between the actual and predicted values of polysulfonamide drug concentration. According to Aheto's discussion, the model is judged to be robust. The predicted values were obtained as concentrations of polysulfonamide drugs, enabling quantitative analysis of polysulfonamide drugs.