A microarray chip-based sers serum analysis method

Abstract

The application relates to a microarray chip-based SERS serum analysis method, belonging to the cross field of medicine and artificial intelligence, and a technical scheme is as follows: by using the prepared microarray chip composed of AuNOs arrays, high-throughput spectrum acquisition can be realized, and the signal intensity of extremely weak biological components can be significantly amplified; SERS is combined with the microarray chip to overcome the problems of changes in spectrum peak position and intensity caused by the exposed environment in serum SERS detection; a spectrum recognition analysis model is proposed, which can be used for recognizing serum spectra at different stages, capturing and distinguishing several key spectral characteristics at different stages, effectively improving the accuracy, sensitivity and specificity of recognizing similar SERS spectra, so that the liver injury caused by cisplatin (CILI) can be accurately recognized and analyzed at the molecular level.

Description

Technical Field

[0001] This invention relates to a method at the intersection of medicine and artificial intelligence, and more particularly to a microarray chip and a method for serum analysis using the same. Background Technology

[0002] Drug-related problems (DRPs) are events involving drug treatment that actually or potentially interfere with intended health outcomes at any point in time when the drug is used. Cisplatin, an effective chemotherapy drug that inhibits tumor cell proliferation and triggers cell death by forming adducts with nuclear and mitochondrial DNA, has been widely used in cancer treatment. The liver, as the body's most vital metabolic and detoxification organ, is more susceptible to drug damage than other organs, making cisplatin-induced liver injury (CILI) one of the most serious complications.

[0003] Recent studies have shown that major byproducts of drug metabolism in the liver, reactive oxygen species (ROS) and reactive nitrogen species (RNS), can lead to cell death or other acute liver failures. Therefore, drug metabolites can reflect the diagnosis of CILI to some extent. Furthermore, the evaluation of enzyme activities such as aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), alanine aminotransferase (ALT), and alkaline phosphatase (ALP) has been used in the diagnosis of CILI. Nevertheless, the application of drug metabolites or certain enzymes alone cannot serve as a specific indicator for predicting complex CILI processes.

[0004] Serum, as a bodily fluid, is the most commonly used clinical specimen, containing various biomolecules such as proteins, carbohydrates, lipids, enzymes, electrolytes, and nucleic acids. Its content and properties change with biological metabolism or the occurrence and development of disease. Unfortunately, due to the complex pathophysiological mechanisms of CILI, traditional techniques struggle to achieve accurate early diagnosis in serum. Furthermore, serum samples are inherently complex and easily affected by other factors in the exposed environment. Therefore, there is an urgent need to develop an innovative, simple, accurate, and reliable method to identify changes in serum composition for early diagnosis of CILI. Summary of the Invention

[0005] To address some or all of the aforementioned problems in the existing technology, the technical objective of this invention is to propose a microarray chip and a method for SERS serum analysis using it. The microarray chip, composed of an AuNOs array, enables high-throughput spectral acquisition and significantly amplifies the extremely weak signal intensity of biological components. The combination of SERS and the microarray chip overcomes the problem of variations in spectral peak position and intensity caused by environmental exposure in serum SERS detection. A spectral recognition and analysis model is proposed, capable of separating serum spectra at different stages and capturing several key spectral features distinguishing different stages, effectively improving the accuracy, sensitivity, and specificity in identifying similar SERS spectra, thereby ultimately enabling accurate identification of CILI at the molecular level. The technical solution to achieve the above technical objectives is as follows.

[0006] In a first aspect, the present invention proposes a microarray chip for surface-enhanced Raman scattering (SERS), which is obtained by adhering a polydimethylsiloxane (PDMS) layer to a plasma-treated glass slide, and then embedding a gold nano-octahedral (AuNOs) array into the glass slide by laser etching.

[0007] In the above technical solution, one method for obtaining a gold nano-octahedral (AuNOs) array includes the following steps:

[0008] The prepared AuNOs solution was added to n-hexane to form an organic / water interface;

[0009] Ethanol was injected into the solution as an inducer to induce AuNOs precipitation and form a dense and ordered monolayer gold nanoparticle film.

[0010] The prepared hydrophilic glass is inserted obliquely into the bottom of the formed AuNOs membrane, and then slowly lifted out to retrieve the AuNOs monolayer membrane floating at the liquid-liquid interface. After the n-hexane has completely evaporated, the AuNOs array is obtained.

[0011] In the above technical solution, one method for preparing AuNO3 solution includes the following steps:

[0012] Under stirring, the newly prepared NaBH4 with a temperature range of 0-4℃ was added to a mixed solution containing HAuCl4 and CTAC. After stirring for 2 minutes and aging at a set temperature for 1 hour, a seed solution was obtained. The set temperature range was 25-30℃.

[0013] The growth solution was prepared in parallel in two beakers;

[0014] HAuCl4, KI and AA were added to the two beakers one after another;

[0015] While stirring, the synthesized seed solution was added to the mixed solution in the beaker until the solution turned light pink. Then the mixed solution in the beaker was transferred to another beaker, stirred thoroughly for about 8-12 seconds, and allowed to stand to ensure that the reaction was complete, thus obtaining AuNOs solution.

[0016] In one embodiment of the above technical solution for preparing the seed solution, the concentration of NaBH4 is 10 mM, the concentration of HAuCl4 is 0.25 mM, and the concentration of CTAC is 0.10 M.

[0017] In one embodiment of the above technical solution for preparing the growth solution, the concentration of HAuCl4 is 0.01M, the concentration of KI is 1mM, and the concentration of AA is 40mM.

[0018] In the above technical solution, one method for preparing the PDMS layer includes:

[0019] A two-dimensional drawing template is made, negative photoresist is spin-coated onto a silicon wafer, and then soft-baked at 60-70℃ and 90-95℃ for 2-5 minutes and 8-10 minutes respectively. The photoresist film is then photolithographically etched through a photolithography machine for 8 seconds, and baked at 60-70℃ and 90-95℃ for 2-5 minutes and 6-8 minutes respectively to obtain the prepared mold.

[0020] Mix the polydimethylsiloxane (PDMS) prepolymer and curing agent in a 10:1 ratio, pour the mixture onto the prepared mold, then apply vacuum and cure at 80-90℃ for 15-20 minutes.

[0021] After the PDMS layer is cooled, it is peeled off from the silicon wafer and then subjected to plasma treatment to obtain the prepared PDMS layer.

[0022] Secondly, the present invention proposes a method for SERS serum analysis using a microarray chip. The method uses any of the microarray chips described above to acquire surface-enhanced Raman scattering (SERS) spectra of the sample serum, utilizes the changes in the composition of the sample serum that lead to corresponding changes in the intensity of the SERS spectral signal, and then judges the changes in serum composition by detecting changes in the characteristic peaks of the SERS spectral signal.

[0023] In the above technical solution, one method for detecting SERS spectral signals includes the following steps:

[0024] PCA was used to project the SERS spectra of the samples onto the principal components (PCs) for data analysis. The top N PCs that accounted for more than 95% of the total variance were selected as features of the spectral identification model.

[0025] Wherein: the spectral features are linear combinations of the K nearest centroid neighbors found by the nearest centroid neighborhood (NCN) criterion.

[0026] In the above technical solution, the representation coefficients in the linear combination have the following relationship with the classification contribution:

[0027] By comparing the sum of the representation coefficients of each class in the centroid neighbors, the larger the sum of the representation coefficients, the greater the classification contribution, and the closer the test sample is to this class.

[0028] In the above technical solution, the characteristic peak analysis involves analyzing the characteristic peaks of a representative spectrum, and the steps for obtaining the representative spectrum include:

[0029] SERS spectral data were smoothed using Savitzky-Golay, baseline corrected using airPLS, and vector normalized.

[0030] After normalization, the average of the spectra is taken as the representative spectrum.

[0031] In the above technical solution, PCA is used to perform data analysis on the SERS spectrum of the sample for the first and second components, and the positions of seven different key characteristic peaks under the first and second component loading are used as key identification factors for judging cisplatin-induced liver injury.

[0032] The positions of the key characteristic peaks are: 637 cm⁻¹ -1 760cm -1 854cm -1 936cm -1 1004cm -1 1087cm -1 1120cm -1 1210cm -1 1252cm -1 1320cm -1 1357cm -1 1434cm -1 1534cm -1 and 1655cm -1 .

[0033] In the above technical solution, the steps for acquiring surface-enhanced Raman scattering (SERS) spectra of serum samples include:

[0034] SERS measurements were performed using a 785nm laser excitation method and a Raman microscope. After exposing the SERS spectrum for a certain time, the spot size and laser power were set, and measurements were taken at five different points.

[0035] Compared with existing technical solutions, the present invention has the following technical effects:

[0036] 1) Microarray chips have advantages such as small size, portability, fast response speed and high detection efficiency. Microarray chips have good repeatability and can be fabricated on a large scale. The AuNOs array involved in microarray chips has simple preparation methods, high uniformity and stability, and can be prepared in large batches.

[0037] 2) Applying microarray chips for SERS serum analysis can overcome the problem that serum contamination by the external environment can cause changes in the peak position and intensity of the SERS spectrum, making it undetectable.

[0038] 3) A spectral recognition model is proposed, which can automatically identify and judge the SERS spectral classification. In the application of cisplatin-induced liver injury (CILI) detection, CILI can be accurately identified at the molecular level. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A schematic diagram of an SERS analysis platform in one implementation method; wherein, Figure 1 (a) is a schematic diagram of the microarray chip designed in this embodiment. Figure 1 (b) shows the SERS spectra of CILI mice at different stages in this embodiment. Figure 1 (c) is a training diagram of the PCA-RCKNCN model in this implementation. Figure 1 (d) shows the serum SERS spectrum analyzed by the PCA-TLNN model in this embodiment;

[0041] Figure 2 This is a schematic diagram illustrating information related to the AuNOs prepared in one implementation method; wherein, Figure 2 (a) is a SEM image of AuNOs in this embodiment. Figure 2 (b) is a TEM photograph of AuNOs in this embodiment. Figure 2 (c) is an HRTEM photograph of AuNOs in this embodiment. Figure 2 (d) is the EDS spectrum of AuNOs in this embodiment. Figure 2 Image (e) is the UV-vis spectrum of AuNOs in this embodiment. Figure 2 (f) is the Raman spectrum of AuNOs in this embodiment;

[0042] Figure 3 This is a schematic diagram illustrating information related to the AuNOs substrate in one implementation method; wherein: Figure 3 Image (a) shows SEM images of AuNOs at low and high magnification in this embodiment. Figure 3 Image (b) is a photograph of the AuNOs nanofilm, AuNOs substrate, and microarray chip in this embodiment. Figure 3 (c) is the SERS mapping of the characteristic peak at 1330 cm⁻¹ on the DTNB-labeled AuNOs substrate in this embodiment. Figure 3 Image (d) shows the SERS spectra of 10 randomly selected points on 5 AuNOs substrates prepared in different batches in this embodiment. Figure 3 In this embodiment, (e) is a histogram showing the intensity at 1330 cm⁻¹ for 10 randomly selected points on 5 AuNO₂ substrates prepared in different batches. Figure 3 In this embodiment, (f) represents different concentrations (10) - 8 M, 10 -9 M, 10 -10 M, 10 -11 M, 10 -12 M and 10 -13 SERS spectrum of AuNOs substrate in the presence of DTNB (M). Figure 3 (g) represents different concentrations (10) -8 M, 10 -9 M, 10 -10 M, 10 -11 M, 10 -12 M and 10 -13 The corresponding calibration curve for M) exists under the DTNB condition. Figure 3 In this embodiment, (h) shows the SERS spectra of 0d, 1d, 5d, 10d, and 50d substrates stored at room temperature using a DTNB. Figure 3 In this embodiment, (i) is the SERS spectrum of the AuNOs substrate after treatment with 0.1 M KI;

[0043] Figure 4 One embodiment is a correlation map between a mouse HE-stained liver section and the central vein region; wherein... Figure 4 (a) is an optical image (4×) of an HE-stained liver section from a mouse that did not receive cisplatin treatment in this embodiment. Figure 4 (b) is a magnified image (10×) of the central vein region of mice that did not receive cisplatin treatment in this embodiment. Figure 4 (c) is an optical image (4×) of a liver section stained with hematoxylin and eosin (HE) from a mouse that received cisplatin treatment for 3 days in this embodiment. Figure 4(d) is a magnified image (10×) of the central vein region of mice treated with cisplatin for 3 days in this embodiment. Figure 4 (e) is an optical image (4×) of an HE-stained liver section from a mouse that received cisplatin treatment for 6 days in this embodiment. Figure 4 (f) is a magnified image (10×) of the central vein region of mice treated with cisplatin for 6 days in this embodiment. Figure 4 (g) is a graph showing the change in serum ALT levels in mice after cisplatin treatment in this embodiment, and 4(h) is a graph showing the change in serum AST levels in mice after cisplatin treatment in this embodiment.

[0044] Figure 5 This is a schematic diagram of the near-field electric intensity distribution in one implementation method using FDTD simulation;

[0045] Figure 6 This is an example of a normalized SERS spectrum of mouse serum at different stages on a microarray chip in one implementation method.

[0046] Figure 7 One implementation method uses PCA-related information graphs of 1-day and 3-day mouse serum SERS data; wherein, Figure 7 (a) is a scree plot showing the change of PCA component count as a function of eigenvalues ​​in the serum SERS data of mice at 1 day and 3 days in this embodiment. Figure 7 (b) is a PCA score graph of mouse serum SERS data at 1 day and 3 days in this embodiment. Figure 7 (c) is the PC1 loading plot of mouse serum SERS data at 1 day and 3 days in this embodiment. Figure 7 (d) is the PC2 loading plot of mouse serum SERS data at 1 day and 3 days in this embodiment. Figure 7 (eh) is a scatter plot describing the SERS spectral intensities at different stages in this embodiment. Through t-test (*p < 0.05, **p < 0.01, ***p < 0.001), the obvious characteristic peak (854 cm⁻¹) of the PC1 and PC2 loading plots is shown. -1 1543cm -1 637cm -1 1210cm -1 The SERS spectra at different stages (1d and 3d) were quantitatively distinguished.

[0047] Figure 8 This is a PCA-related information graph of serum SERS data from 3-day and 6-day mice in one implementation method; wherein, Figure 8 (a) is a scree plot showing the change of PCA component count as a function of eigenvalues ​​in the serum SERS data of mice at 3 days and 6 days in this embodiment. Figure 8(b) is a PCA score graph of mouse serum SERS data at 3 days and 6 days in this embodiment. Figure 8 (c) is a PC1 loading plot of serum SERS data from mice at 3 days and 6 days in this embodiment. Figure 8 (d) is the PC2 loading plot of serum SERS data from mice at 3 days and 6 days in this embodiment. Figure 8 (eh) is a scatter plot describing the SERS spectral intensities at different stages in this embodiment. Through t-test (*p < 0.05, **p < 0.01, ***p < 0.001), the obvious characteristic peak (854 cm⁻¹) of the PC1 and PC2 loading plots is shown. -1 1543cm -1 637cm -1 1210cm -1 The SERS spectra at different stages (3d and 6d) were quantitatively distinguished.

[0048] Figure 9 This is a PCA-related information graph of serum SERS data from mice at 1 day and 6 days in one implementation method; wherein, Figure 9 (a) is a scree plot showing the change of PCA component count as a function of eigenvalues ​​in the serum SERS data of mice at 1 day and 6 days in this embodiment. Figure 9 (b) is a PCA score graph of mouse serum SERS data at 1 day and 6 days in this embodiment. Figure 9 (c) is a PC1 loading plot of serum SERS data from mice at 1 day and 6 days in this embodiment. Figure 9 (d) is the PC2 loading plot of serum SERS data from mice at 1 day and 6 days in this embodiment. Figure 9 (eh) is a scatter plot describing the SERS spectral intensities at different stages in this embodiment. Through t-test (*p < 0.05, **p < 0.01, ***p < 0.001), the obvious characteristic peak (936 cm⁻¹) of the PC1 and PC2 loading plots is shown. -1 1320cm -1 1210cm -1 1534cm -1 The SERS spectra at different stages (1d and 6d) were quantitatively distinguished.

[0049] Figure 10 This is a schematic diagram illustrating the RCKNCN classification principle in a two-dimensional space in one implementation method; wherein... Figure 10 (a) is a schematic diagram of the test sample selecting K=6 nearest centroid neighbors according to the NCN criterion in this embodiment. Figure 10 (b) is a schematic diagram in this embodiment where the test sample is represented as a linear combination of the selected centroid neighbors;

[0050] Figure 11 (ad) is a schematic diagram comparing the accuracy, AUC, sensitivity, and specificity of the spectral identification analysis model for serum SERS data of CILI mice at 1d & 3d, 3d & 6d, and 1d & 6d with the traditional PCA-KNN, PCA-SVM, and PCA-LDA models in one implementation method.

[0051] Figure 12 (ac) is a schematic diagram comparing the AUC of four models in one implementation of serum SERS data staging of CILI mice at 1d & 3d, 3d & 6d, and 1d & 6d. Detailed Implementation

[0052] Cisplatin is an effective chemotherapy drug, but cisplatin-induced liver injury (CILI) has become one of the most serious complications. The pathophysiological mechanism of CILI is complex, and traditional techniques are difficult to obtain accurate early diagnosis in serum. There is a need to develop an innovative, simple, accurate and reliable method to identify changes in composition in order to diagnose CILI at an early stage.

[0053] Therefore, this invention proposes an innovative, simple, accurate, and reliable method to identify component changes, which can be used for the early diagnosis of liver injury (CILI). In this process, this invention proposes a microarray chip and a method for SERS serum analysis using it. The microarray chip, composed of an AuNOs array, enables high-throughput spectral acquisition and significantly amplifies the extremely weak signal intensity of biological components. The combination of SERS and the microarray chip overcomes the changes in spectral peak position and intensity caused by environmental exposure in serum SERS detection. A spectral identification and analysis model is proposed to separate serum spectra at different stages and capture several key spectral features that distinguish different stages, effectively improving the accuracy, sensitivity, and specificity in identifying similar SERS spectra, thereby ultimately enabling accurate identification of CILI at the molecular level.

[0054] To enhance understanding of the invention, it will be further described in detail with reference to embodiments and accompanying drawings.

[0055] The instruments, equipment, and testing conditions used in this invention are as follows:

[0056] The scanning electron microscope (TEM) images were obtained using an S-4800II field emission scanning electron microscope manufactured by Hitachi, Japan.

[0057] The transmission electron microscope (SEM) images were obtained using a TECNAI 10 transmission electron microscope manufactured by Philips AG of the Netherlands.

[0058] Raman spectroscopy and SERS imaging were performed using an Invia Reflex laser micro Raman spectrometer manufactured by Renishaw, UK. The test conditions were: laser wavelength 785 nm, exposure time 5 s, laser intensity 10 mW, and a 50× objective lens.

[0059] The SERS spectral data were processed using OriginPro 2021 analysis software developed by OriginLab, Inc., including Savitzky-Golay smoothing, airPLS baseline correction, and vector normalization.

[0060] The performance of the spectral recognition model was analyzed using OriginPro 2021 software developed by OriginLab, Inc.

[0061] (I) Microarray Chip

[0062] The microarray chip is obtained by adhering a polydimethylsiloxane (PDMS) layer to a plasma-treated glass slide, and then embedding a gold nanoparticle (AuNOs) array into the glass slide using laser etching. This chip utilizes the self-assembly of the gold nanoparticle (AuNOs) array at the oil-water interface. Due to its unique structure and optical properties, it exhibits strong plasma enhancement and can be used as a SERS-active substrate. It can generate high-density and uniform "hot spots" in the corners, producing a strong polarization-induced electric field, thus enhancing the reproducibility of the microarray chip. Because AuNOs can provide an open and wide hot spot region for biomolecules, it offers advantages such as fast reaction speed and high detection efficiency. This microarray chip has a simple structure and can be fabricated on a large scale. In one embodiment, the chip is 2.8 cm long, 2.8 cm wide, and 0.5 cm high, exhibiting small size and portability.

[0063] The fabrication steps of a microarray chip generally include:

[0064] 1) Preparation of gold octahedrons (AuNOs) using seed-mediated growth method;

[0065] 2) Using the liquid-liquid interface self-assembly method, the AuNOs prepared in step 1) are assembled into an AuNOs array as a SERS substrate;

[0066] 3) A microarray chip based on AuNOs arrays was designed and fabricated. Example 1: Synthesis and Characterization of AuNOs

[0067] 1) Under vigorous stirring, 1.2 mL of freshly prepared, 10 mM ice-cold NaBH4 was rapidly added to 20 mL of a mixed solution containing 0.25 mM HAuCl4 and 0.10 M CTAC. The solution was stirred for 2 minutes and aged at 30°C for 1 hour to obtain the seed solution. The temperature range of NaBH4 was 0-4°C. 30°C could also be any other temperature within the range of 25°C-30°C; in other examples, it could be 25°C, 26°C, 27°C, 28°C, 29°C, etc.

[0068] 2) Prepare the growth solution in parallel in two beakers (a and b). Simply put, dissolve 0.64 g of CTAC in deionized water (18.91 mL) and place in a water bath at 25°C;

[0069] 3) Gradually add 500 μL of 0.01M HAuCl4, 100 μL of 1mM KI, and 440 μL of 40mM AA to two beakers. Then, while stirring vigorously, add 50 μL of the seed solution synthesized in step 1.1) to beaker a until the solution turns light pink. Then transfer the 50 μL mixed solution from beaker a to beaker b. After thorough stirring, allow it to stand to ensure the reaction proceeds completely; the resulting AuNO3 solution is obtained. The stirring time ranges from 10 to 12 seconds, and the standing time ranges from 10 to 20 minutes, depending on the experimental environment. For example, stir for about 10 seconds and let it stand for 15 minutes.

[0070] 4) Morphology and SERS effect characterization of AuNOs: The morphology and structure of AuNOs were detected by SEM, TEM, high-resolution TEM, SAED imaging, EDS elemental mapping and UV-vis-NIR spectrophotometer.

[0071] like Figure 2 a and Figure 2 As shown in b, SEM and TEM images demonstrate the high uniformity and dispersion of AuNOs, with a statistically average edge length of 85 nm. Figure 2 The HRTEM image in c shows that the interplanar spacing at the lattice edges is This corresponds to the (111) plane of the golden octahedron (fcc) Au. Furthermore, the corresponding Fourier transform (FFT) pattern ( Figure 2 The illustration in c) confirms the single crystallinity of AuNOs with the (111) plane as the basal plane. Figure 2 Energy dispersive spectroscopy (EDS) in d indicates that AuNOs consists of only gold, while the copper peaks are caused by the copper mesh. Figure 2The UV-vis-NIR spectrum of AuNOs is shown, revealing the characteristic SPR peak of AuNOs at 588 nm. This sharp absorption peak is primarily attributed to the uniform shape and size of the AuNOs. To evaluate the SERS activity of AuNOs, DTNB (1×10⁻⁶) was measured. -6 AuNO3 modified with M) and DTNB (1×10 -2 SERS spectrum of M) Figure 2 f) shows that AuNOs can provide significant and uniform enhancement of the SERS spectral signal, which is beneficial for signal reproducibility. The synthesis steps of AuNOs demonstrate that the AuNOs array preparation method is simple and can be mass-produced.

[0072] Example 2 Assembly and characterization of AuNOs array

[0073] 1) The clean glass substrate was hydroxylated with freshly prepared piranha solution (H2SO4∶H2O2=7∶3, v / v) to make its surface hydrophilic, then rinsed with deionized water and dried with nitrogen.

[0074] 2) The prepared AuNOs solution (4 mL) was added to n-hexane (5 mL) to form an organic / water interface. Then, 4 mL of ethanol was injected into the solution as an inducer to induce the precipitation of AuNOs and form a large-area, dense, and ordered monolayer gold nanoparticle film.

[0075] 3) Insert the hydrophilic glass prepared in step 1) obliquely into the bottom of the AuNOs membrane formed in step 2), slowly lift it from the middle, and pick up the AuNOs monolayer membrane floating at the liquid-liquid interface. After the n-hexane has completely evaporated, the AuNOs array is obtained.

[0076] 4) Morphology of AuNOs array and characterization of SERS effect

[0077] The morphology of the AuNOs array substrate was detected by SEM. The uniformity of the SERS substrate was evaluated. DTNB was used as the Raman signal molecule to evaluate the uniformity of the SERS substrate. One prepared substrate was immersed in an immersion solution with a concentration of 1×10⁻⁶. -8 The substrate was placed in a DTNB solution of M and allowed to stand for 2 hours before being removed and air-dried. The dried DTNB-labeled substrate was then placed on a Raman spectrometer, and the DTNB was analyzed at 1330 cm⁻¹. -1 The intensity of the characteristic peaks was used to perform SERS imaging on the substrate. Selected areas on the substrate were scanned by laser with a point-to-point interval of 50 mm and an exposure time of 10 s.

[0078] Figure 3SEM images in figure a confirm that AuNOs are tightly and orderly assembled on the glass substrate. The AuNOs array was prepared by self-assembly at the organic / water interface. After the addition of ethanol, the AuNOs in the organic / water system can be compressed into a tightly packed, ordered film. Figure 3 b(i)). The prepared AuNOs film can be transferred onto a hydrophilic glass substrate ( Figure 3 b(ii)); Therefore, after bonding the PDMS layer to a glass slide with 16 embedded AuNOs substrates via laser etching, a microarray chip can be successfully fabricated. Figure 3 b(iii)). Wherein: the PDMS layer, in one preparation method, includes the following steps:

[0079] 3.1) Create a 2D drawing template using AutoCAD software. Spin-coat the negative photoresist SU-82025 onto a clean silicon wafer (low speed 500 rpm for 10 seconds, high speed 1000 rpm for 50 seconds), then perform a soft baking process. The soft baking temperature range can be selected as 60-70℃ or 90-95℃, and the baking time can be selected as 2-5 minutes or 8-10 minutes, respectively. For example, soft baking is performed at 65℃ and 95℃ for 3 minutes and 9 minutes, respectively. Next, simultaneously perform photolithography on the SU-8 thin film using a photolithography machine for 8 seconds, and bake at 60-70℃ and 90-95℃ for 2-5 minutes and 6-8 minutes, respectively, to obtain the prepared mold. For example, baking is performed at 65℃ and 95℃ for 2 minutes and 7 minutes, respectively.

[0080] 3.2) Subsequently, the polydimethylsiloxane (PDMS) prepolymer and curing agent are mixed in a 10:1 ratio, poured onto a prepared mold, and then vacuum-sealed for curing. The curing temperature range is 80-90°C, and the curing time range is 15-20 minutes. For example, curing is performed at 80°C for 20 minutes. After the PDMS layer cools, the cleaned glass surface is plasma-treated. After cooling, the PDMS layer is peeled off from the silicon wafer and then plasma-treated again.

[0081] 3.3) The PDMS layer prepared in 3.2) is adhered to a glass slide, and the glass slide is embedded with the prepared AuNOs array by laser etching; thus, the proposed microarray chip is obtained.

[0082] SERS plots were performed on the AuNOs substrate to estimate its uniformity, such as... Figure 3 As shown in c. Each pixel in the diagram represents the DTNB at 1330cm. -1The peak intensity at the point (attributed to the symmetric nitro extension of (NO2)) was determined using a color scheme from blue (minimum intensity) to red (maximum intensity). The calculated relative standard deviation (RSD) was 6.14%, reflecting the excellent homogeneity of the AuNOs substrates. Furthermore, reproducibility was verified by measuring the SERS spectra of 10 randomly selected points from 5 AuNOs substrates prepared from different batches. Figure 3 d) Relatively consistent SERS intensity was observed at all points, with an RSD of 7.47% ( Figure 3 e). In the presence of DTNB, SERS spectra were recorded on the substrate, ranging from 10. -8 M to 10 -13 M, to explore its sensitivity ( Figure 3 f). The spectra of different concentrations of DTNB showed a clear degree of differentiation. Figure 3 g showed 1330cm -1 A large linear relationship exists between the SERS intensity and the logarithm of the DTNB concentration. The calibration curve is y = 10975.94x + 142840.33, with a coefficient of determination (R²). 2 The value is 0.9954. Therefore, the detection limit (LOD) can be calculated to be 83.2 fM, revealing the excellent signal amplification performance of the AuNOs substrate. FDTD confirms the enhancement mechanism of the AuNOs array ( Figure 5 This clearly demonstrates that more abundant "hot spots" can be generated between adjacent AuNOs. Furthermore, the prepared microarray chip, after being stored at room temperature for different times (0 days, 1 day, 5 days, 10 days, and 50 days), showed only a slight decrease in SERS spectrum intensity, indicating satisfactory stability. Figure 3 h). For unlabeled SERS measurements, interference from the background signal from the substrate is unacceptable; therefore, we applied 0.1 M KI to remove residual chemicals, and the SERS spectra recorded on the substrate indicate the cleanliness of the AuNOs (h). Figure 3 i). The above results indicate that the prepared AuNOs array exhibits good SERS enhancement and uniformity.

[0083] In existing technologies, while Raman signals can be significantly enhanced, even down to the single-molecule level, through surface plasmon resonance (SPR) on rough metallic nanostructures, SERS, in principle, relies on electromagnetic "hot spots" on light-generated plasmonic nanostructures where weak Raman signals can be enhanced. While discrete clusters formed by the aggregation of monodisperse colloidal solutions produce significant enhancements, these enhancements are rather uncontrolled and non-uniform, leading to poor reproducibility. Conversely, large-scale nanostructured substrates exhibit uniform "hot spots" that can be reproducibly enhanced; therefore, efforts have been made to fabricate plasmonic arrays to maximize field enhancement. However, due to the need for expensive and complex fabrication equipment, most of these SERS-active substrates are not cost-effective for large-scale production in clinical applications.

[0084] As can be seen from the above embodiments, the microarray chip utilizes the unique structure and optical properties of gold nano octahedrons to generate a high-density and uniform "hot spot" with a strong polarized light-induced electric field at the corner, exhibiting good repeatability. The preparation method of gold nano octahedron arrays is simple and can be scaled up, resulting in microarray chips that are small in size and portable. Furthermore, since gold nano octahedron arrays can provide an open and wide hot spot region for biomolecules, they offer advantages such as fast reaction speed and high detection efficiency during detection.

[0085] (II) Application of microarray chips for SERS serum analysis

[0086] Using the microarray chip obtained in any of the above embodiments, surface-enhanced Raman scattering (SERS) spectra of sample serum are acquired. The changes in the composition of the sample serum lead to changes in the corresponding SERS spectral signal intensity, and the changes in serum composition are determined by the changes in the characteristic peaks of the detected SERS spectral signal.

[0087] As can be seen, Raman spectroscopy for serum analysis is a non-destructive, reagent-free vibrational spectroscopy technique and a convenient label-free detection method. Raman spectroscopy can analyze the Raman-active functional groups of substances such as nucleic acids, proteins, lipids, and carbohydrates in human biochemical samples, providing objective and quantifiable molecular information for diagnosis and efficacy evaluation, and revealing differences in expression at the molecular level.

[0088] However, typical Raman spectra are very weak. In principle, SERS relies on electromagnetic "hot spots" on light-generated plasmonic nanostructures, where the weak Raman signal can be enhanced. While discrete clusters formed by the aggregation of monodisperse colloidal solutions produce significant enhancements, these enhancements are quite uncontrolled and non-uniform, leading to poor reproducibility. By combining nanotechnology with Raman spectroscopy, the Raman signal can be significantly improved, even down to the single-molecule level, through surface plasmon resonance (SPR) of rough metal nanostructures. However, fabricating plasmonic arrays to maximize field enhancement requires expensive and complex processing equipment, making the large-scale production of most of these SERS-active substrates uneconomical for clinical applications. The microarray chip proposed in this invention overcomes these problems.

[0089] While SERS-based research has yielded significant results for early disease diagnosis, directly detecting vibrational fingerprints generated by disease-related biomolecules is often unimportant. Serum components are complex and heterogeneous, making them difficult to distinguish through visual inspection in a short time. Therefore, this invention designs a spectral recognition and analysis model capable of large-scale on-site SERS identification and detection. Details are as follows:

[0090] PCA is used to project the SERS spectrum of the sample onto the principal components (PCs) for data analysis, reducing high-dimensional data to generate low-dimensional data and dividing the data into groups. The top N PCs that account for more than 95% of the total variance are selected as spectral features. The spectral features are linear combinations of the K nearest centroid neighbors found by the nearest centroid neighborhood (NCN) criterion, and the values ​​of N and K can be set.

[0091] The spectral identification and analysis model described herein, denoted as PCA-RCKNCN, improves upon existing machine learning (ML) methods, such as Principal Component Analysis (PCA), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), PCA-KNN, and PCA-SVM, in classifying large and complex spectral data and distinguishing SERS spectra of biological samples with complex compositions. This is evident from the experimental comparison results that follow.

[0092] The following example illustrates how to perform SERS serum analysis using a microarray chip, specifically focusing on the label-free detection and identification of cisplatin-induced liver injury.

[0093] Example 3: Establishment of the CILI rat model

[0094] 1) Rats were orally administered cisplatin (dissolved in 40% polyethylene glycol 400) at a dose of 250 mg / kg / day for 6 days. Blood and liver samples were obtained on days 1, 3, and 6 (n=20 samples). Serum was separated by centrifuging blood samples at 2000 rpm for 20 min.

[0095] 2) Sections stained with hematoxylin and eosin (HE) showed hepatocellular damage on days 3 and 6. Figure 4 af). Meanwhile, ALT and AST both increased on days 3 and 6. Figure 4 These results indicate that cisplatin treatment caused liver damage in rats, suggesting that the model was successfully established.

[0096] Example 4: SERS Characterization of Serum at Different Stages

[0097] 1) SERS measurements were performed using a 50× objective confocal Raman microscope under 785nm laser excitation. The exposure time for the SERS spectra was 5s. The spot size and laser power were set to 2mm and 10mW, respectively. Spectra were measured at five different points to ensure representativeness and reliability. All SERS spectral data were processed using Origin software (Savitzky-Golay smoothing, airPLS baseline correction, and vector normalization). After normalization, the average of 20 spectra at each stage was taken to obtain representative spectra for 1 day, 3 days, and 6 days. Characteristic peaks of the representative SERS spectral data were analyzed using Origin software. All processed serum sample SERS spectral data were further analyzed using a spectral identification analysis model to improve classification performance.

[0098] 2) Serum samples were analyzed using the prepared microarray chip. The mean normalized SERS spectra of CILI mouse model serum at different time points (1, 3, and 6 days) after noise reduction and baseline correction were obtained, as shown below. Figure 6 As shown. The standard deviations do not overlap in the spectral regions, indicating the significance and reproducibility of the differences. The main SERS peak is 637 cm⁻¹. -1 760cm -1 854cm -1 936cm -1 1004cm -1 1087cm -1 1120cm -1 1210cm -1 1252cm -1 1320cm -1 1357cm -1 1434cm -1 1534cm -1and 1655cm -1 Three serum groups were observed consistently. These spectra revealed various biological components, such as lipids, proteins, and nucleic acids. The positions of significant Raman peaks and provisional vibrational mode assignments observed in this invention are shown in Table 1 (Table 2 shows their corresponding Chinese versions). Variations in composition lead to corresponding SERS spectral signal intensities; therefore, characteristic peak analysis helps to explain the different serum component variations observed between different groups. For example, the 1655 cm⁻¹ peak of the α-helix belonging to protein amides. -1 A significant increase in peak intensity was observed at 1087 cm⁻¹, revealing changes in the protein's secondary structure. Furthermore, the peak intensity was at 1087 cm⁻¹. -1 This is attributed to the phosphate group PO in the DNA backbone. 2- The symmetric stretching vibrations increased with increasing cisplatin-induced oxidative stress, indicating a relative increase in nucleic acid content and a change in the DNA double helix structure. However, it is difficult to classify time-differentiated groups based solely on the morphology of the SERS spectral signal.

[0099] Table 1

[0100]

[0101]

[0102] Table 2

[0103]

[0104]

[0105] Example 5: Establishment of the PCA-RNKNCN Model

[0106] In this invention, the PCA-RNKNCN model (spectral identification and analysis model) is an automatic identification and judgment model for on-site SERS detection. It utilizes PCA to project the SERS spectrum of the sample onto principal components (PCs) for data analysis, selecting the top N PCs accounting for >95% of the total variance as spectral features. Specifically, the PCA-RNKNCN model finds the K nearest centroid neighbors using the nearest centroid neighborhood (NCN) criterion, and the test sample is a linear combination of these K nearest centroid neighbors. In this linear combination, the representation coefficients and classification contribution have the following relationship: comparing the sum of the representation coefficients for each class among the centroid neighbors, the larger the sum of the representation coefficients, the greater the classification contribution, and the closer the test sample is to that class.

[0107] 1) PCA-RCKNCN was analyzed using Origin software. The built-in "PCA" function was used to obtain scree plots, scoring plots, and loading plots. The scree plots showed the change in the number of components in the serum SERS spectrum with eigenvalues. The first and second principal components were used for further data analysis. The scoring plots composed of the first two principal components (PCs) (groups 1 & 3, 3 & 6, and 1 & 6, respectively) easily extracted key features from the SERS spectrum. By evaluating the PC1 and PC2 loading plots, key spectral features distinguishing different stages of CILI mice were captured. After PCA processing, the top 10 PCs, accounting for 95% of the total variance, were used as features of PCA-RCKNCN. To analyze the performance of PCA-RCKNCN, we compared this model with traditional PCA-KNN, PCA-SVM, and PCA-LDA in groups 1 & 3, 3 & 6, and 1 & 6, respectively. Leave-one-out cross-validation (LOOCV) was used to evaluate and optimize the performance of the discrimination model in terms of accuracy, AUC, sensitivity, and specificity. For the PCA-RCKNCN model, using the nearest centroid neighborhood (NCN) criterion based on representation coefficients and k = 1–8, and using the top 10 centroids as input, the model achieved optimal performance by optimizing the parameters PCs and K. For the PCA-KNN model, using the Euclidean distance function and k = 1–8, and using the top 10 centroids as input, the model achieved optimal performance by optimizing the parameters PCs and K. For the PCA-SVM model, using a linear kernel function and the top 10 centroids as input, the model achieved optimal performance by optimizing the parameters PCs. For the PCA-LDA model, using the top 10 centroids as input, the model achieved optimal performance by optimizing the parameters PCs.

[0108] 2) RCKNCK classification principle as follows Figure 10 As shown, firstly, we use the nearest centroid neighborhood (NCN) criterion to find the nearest centroid neighbor of the test sample. Figure 10 a) Then the test samples are linearly represented by their nearest centroid neighbors (K=6), and the representation coefficients of these nearest centroid neighbors are as follows: Figure 10 b. Finally, compare the sum of the representation coefficients of each class in the centroid neighbors. The larger the sum of the representation coefficients, the greater the contribution and the closer the test sample is to this class. Conversely, the smaller the sum of the representation coefficients, the smaller the contribution and the more the test sample escapes this class. The test sample ultimately selects the correct class.

[0109] Example 6 Performance Analysis of the PCA-RNKNCN Model

[0110] 1) Principal Component Analysis (PCA) is a multivariate data analysis method that reduces high-dimensional data to low-dimensional data and distinguishes groups by projecting SERS spectra onto principal components (PCs), thus avoiding the loss of valuable information. The PCA process can extract the most meaningful information from the SERS spectra obtained from the CILI mouse model and concentrate it into a limited number of PCs, which are then distinguished by the constructed classification model. Figure 7 a, Figure 8 a and Figure 9 a shows the contribution of the eigenvalues ​​of each PC to the total variance of all SERS spectra in groups 0&3d, 3&6d, and 0&6. This clearly shows that the eigenvalues ​​decrease rapidly with increasing number of PCs, with only the first few PCs representing the largest variance. It can be observed that the first 10 PCs describe over 95% of the variance in groups 0&3d, 3&6d, and 0&6. Based on the obtained dataset, PCA scoring plots using PC1 and PC2 reveal the separation among groups 0&3d, 3&6d, and 0&6, which each account for 55.4% of the total variance. Figure 7 b) 65.1% Figure 8 b) and 58% Figure 9 (b) The score plots of PC1 and PC2 clearly exhibit a unique and distinct clustering pattern, with ellipses representing 95% confidence intervals. Although some overlapping areas were observed, the distinct clusters allowed for precise identification of the key features of time-dependent component changes in serum samples.

[0111] 2) Figure 7 cd, Figure 8 cd and Figure 9 The PCA loading plots of PC1 and PC2 in the CD showed significant vibrational peaks, which significantly contributed to identification. A t-test was used (*p < 0.05, **p < 0.01, ***p < 0.001) to analyze the characteristic peak intensities of serum from time-differentiated GILI mice in the PC1 and PC2 loading plots. Figure 7 eh、 Figure 8 eh and Figure 9 Quantitative analysis was performed to confirm that seven different key characteristic peaks under PC1 and PC2 loads can serve as key recognition factors for CILI mice at different stages.

[0112] 3) At 854cm -1 The characteristic peak intensity of the (COC) skeletal pattern belonging to α-anomers (polysaccharides, pectin) decreased in groups 1 & 3. Figure 7 e) Add (to groups 3 & 6) Figure 8 e), while at 1534cm -1 The peak intensities of the amide carbon group vibration and aromatic hydrogen increase in groups 1 & 3. Figure 7f), in groups 3 & 6, the decrease ( Figure 8 f). This can be explained by the increased secretion of antioxidants such as carotenoids to compensate for the oxidative losses caused by cisplatin-stimulated oxidative reactive substances. As time progresses, the content of reactive oxygen species continues to rise, and the function and conformation of biomolecules such as nucleic acids and proteins become abnormal. Figure 9 f) This leads to decreased antioxidant capacity and inhibited carotenoid secretion. Researchers found that beta-carotene can inhibit collagen synthesis in hepatic stellate cells, activation of hepatic stellate cells (HSCs), and lipid loss; therefore, the two peaks show opposite trends. From Figure 7 h、 Figure 8 g and Figure 9 g shows that at 1210cm -1 The peak values ​​for tyrosine and phenylalanine were observed at 936 cm⁻¹, with varying degrees of increase in each group. Furthermore, the peak value at 936 cm⁻¹ was also observed. -1 The peak intensity change at that point is caused by the CC main chain. Figure 7 e) indicates that changes in amino acid residues in the side chain of the protein molecule may lead to protein structural disorder. In the 3&6d group, at 1434cm -1 The peak value at this point belongs to the CH2 scissoring of lipid molecules and shows an increasing trend, indicating that lipid accumulation occurs in the later stages of liver injury due to a decrease in liver metabolism. Figure 8 h). During CILI, abnormal methionine metabolism occurs, leading to weakened methylation and sulfidation, such as... Figure 7 As shown in g.

[0113] 4) To analyze the performance of PCA-RCKNCN, the model was compared with traditional PCA-k nearest centroid neighbor (KNN), PCA-Support Vector Machine (SVM), and PCA-Linear Discriminant Analysis (LDA). The model's performance in terms of accuracy, AUC, sensitivity, and specificity was evaluated using LOOCV. Figure 11The diagnostic accuracy of the PCA-RCKNCN model was studied, and it was found that the highest accuracy (97.5%) was achieved in groups 1 & 3, while the accuracy of PCA-SVM, PCA-KNN, and PCA-LDA was slightly lower, at 92.5%, 90%, and 90%, respectively. Similarly, compared with traditional models, the PCA-RCKNCN model also showed better performance in AUC (97.5%), sensitivity (100% at 1 day, 95% at 3 days), and specificity (95% at 1 day, 100% at 3 days). Furthermore, the performance of the PCA-RCKNCN model remained consistent in groups 3 & 6 and 1 & 6, indicating its good ability to analyze SERS spectral data. The AUC of the four models is shown by the area under the ROC curve values ​​for groups 1 & 3, 3 & 6, and 1 & 6. Figure 12 The results showed that the PCA-RCKNCN model had the highest AUC value in groups 1 and 3, while the AUC values ​​of groups 3 and 6 and groups 1 and 6 reached 1, indicating that the PCA-RCKNCN model has satisfactory staging analysis capabilities. Therefore, the established PCA-RCKNCN model achieved satisfactory identification results in groups 1 and 3, 3 and 6, and 1 and 6.

[0114] Table 3 below lists the English abbreviations and corresponding Chinese explanations of the units involved in this invention.

[0115] Table 3

[0116]

[0117] As can be seen from the above embodiments, SERS spectral analysis driven by the PCA-RCKNCN model can accurately identify CILI at the molecular level. The fabricated microarray chip composed of AuNOs arrays enables high-throughput spectral acquisition and significantly amplifies the extremely weak signal intensity of biological components. The PCA-RCKNCN algorithm is used to construct a SERS spectral recognition model, separating serum spectra at different stages and capturing several key spectral features that distinguish these stages. This effectively improves the accuracy, sensitivity, and specificity in identifying similar SERS spectra. The PCA-RCKNCN model can be considered a good alternative to traditional discriminant models. Combining SERS with the PCA-RCKNCN model presents a powerful and promising label-free serum detection tool with great application potential for the early diagnosis of CILI.

[0118] Through the above description of the embodiments, those skilled in the art can clearly understand that the method of this disclosure can be implemented by means of software plus necessary general-purpose hardware, and of course, it can also be implemented by special-purpose hardware, including application-specific integrated circuits, dedicated CPUs, dedicated memory, dedicated components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or dedicated circuits. However, for the purposes of this disclosure, software program implementation is more often a preferred implementation method.

[0119] It should be noted that the capacity data in this application is exemplary and should not be considered limiting. It can be changed depending on the number of products and application requirements.

[0120] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A method for preparing a microarray chip, characterized in that: the microarray chip is used for surface enhanced Raman scattering (SERS) detection, which is obtained by adhering a polydimethylsiloxane (PDMS) layer on a plasma-treated glass sheet, and then embedding the glass sheet in a gold nanooctahedron (AuNOs) array by laser etching; wherein the nanooctahedron (AuNOs) array, the obtaining step comprises: adding the prepared AuNOs solution into n-hexane to form an organic / water interface; injecting ethanol as an inducer into the solution obtained in the above step to induce the precipitation of AuNOs and form a dense and ordered monolayer of gold nanoparticles; slowly inserting the prepared hydrophilic glass into the bottom of the formed AuNOs film, slowly lifting it from the middle, and catching the AuNOs monolayer film floating on the liquid-liquid interface, and after the n-hexane is completely volatilized, the AuNOs array is obtained; the AuNOs solution, the preparation step comprises: under stirring, adding freshly prepared NaBH4 with a temperature range of 0-4℃ into a mixed solution containing HAuCl4 and CTAC, stirring for 2 minutes and aging at a set temperature for 1 hour, then a seed solution is obtained, and the set temperature belongs to a temperature range of 25-30℃; preparing the growth solution in two beakers in parallel; adding HAuCl4, KI and AA into the two beakers step by step; under stirring, adding the synthesized seed solution into the mixed solution in one beaker until the color of the solution changes to light pink, then transferring the mixed solution in the beaker to the other beaker, and after about 8-12 seconds of thorough stirring, standing to ensure that the reaction is completely carried out, to obtain the AuNOs solution. the preparation method of the PDMS layer comprises:

2. The method of claim 1, wherein: making a two-dimensional drawing template, spin coating negative photoresist on a silicon wafer, then soft baking at 60-70℃ and 90-95℃ for 2-5 minutes and 8-10 minutes respectively, photoetching the photoresist film through a photoetching machine with an etching time of 8 seconds, and baking at 60-70℃ and 90-95℃ for 2-5 minutes and 6-8 minutes respectively to obtain a prepared mold; mixing polydimethylsiloxane (PDMS) prepolymer and curing agent at a ratio of 10:1, then pouring them into the prepared mold, then vacuumizing, and curing at 80-90℃ for 15-20 min; after the PDMS layer cools down, peeling the PDMS layer from the silicon wafer, then performing plasma treatment to obtain the prepared PDMS layer. ​

Images