A triple-coupling type SERS substrate, a preparation method and application thereof

Abstract

This invention provides a triple-coupled SERS substrate, its preparation method, and its applications. This triple-coupled SERS substrate achieves excellent signal amplification and can acquire rich SERS bands for both single and multi-component components. A robust machine learning classifier is used to identify and analyze the acquired serum metabolic Raman spectral features, demonstrating high clinical accuracy, sensitivity, and specificity in differentiating between gastrointestinal malignancies and healthy human samples. Serum metabolic analysis based on the triple-coupled SERS substrate constructed in this invention provides a promising solution to the clinical challenge of differentiating gastrointestinal malignancies and can be further expanded into clinical cancer detection based on other biological fluids.

Description

Technical Field

[0001] This invention relates to the field of surface-enhanced Raman spectroscopy, and in particular to a triple-coupled SERS substrate, its preparation method, and its application. Background Technology

[0002] Digestive system cancers are a leading cause of cancer-related morbidity and mortality worldwide, accounting for 35% of annual cancer deaths. Among them, colorectal cancer, gastric cancer, and liver cancer have persistently high incidence rates due to unhealthy diets, Helicobacter pylori infection, and infections with hepatitis B and C viruses. In clinical practice, these three types of abdominal cancer often present with similar symptoms but different underlying conditions, potentially including abdominal pain, weight loss, or gastrointestinal bleeding, making it difficult to differentiate tumor types based solely on clinical presentation. Furthermore, the treatment pathways for different digestive tract tumors vary significantly; therefore, accurate identification of digestive tract tumors is crucial for guiding treatment and improving prognosis. However, conventional diagnostic techniques such as endoscopic biopsy are highly invasive, time-consuming, and can cause physical trauma, making them extremely unsuitable for dynamic monitoring of high-risk populations and imposing a significant medical burden on patients. Therefore, developing minimally invasive, convenient, and low-cost detection technologies is of great importance for the population-based prevention and control of digestive tract tumors.

[0003] In recent years, biofluid analysis based on tumor markers (mainly protein macromolecules) has been widely used for the auxiliary diagnosis, efficacy monitoring, and recurrence prediction of tumors. However, from the current state of cancer diagnosis, most tumor markers suffer from insufficient specificity. For example, carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9), commonly used for the auxiliary diagnosis of gastric cancer, may be highly expressed in liver cancer and colorectal cancer, easily leading to misdiagnosis of cancer lesions. Specific tumors induce characteristic metabolic fingerprint changes, thus providing a reference for precise clinical tumor diagnosis. Based on this, many studies have reported the important contributions of metabolomics to cancer diagnosis and successfully distinguished different types of tumor patients from healthy controls using changed metabolite components. Therefore, unbiased detection of metabolites released in biofluids can provide opportunities for precise tumor identification.

[0004] Currently, serum-based minimally invasive biochemical and immunoassay techniques are widely used in disease detection, tumor screening, and diagnosis. These methods primarily rely on specific enzymatic reactions and immunolabeling techniques to detect specific targets, failing to provide comprehensive metabolic fingerprint analysis. Liquid chromatography-mass spectrometry (LC-MS), considered the "gold standard," can perform qualitative or quantitative analysis of complex components in serum, but generally requires complex and time-consuming sample pretreatment procedures such as derivatization or solid-phase microextraction. Therefore, developing a rapid, label-free, broad-spectrum metabolic fingerprint analysis technique is crucial for the timely screening and accurate identification of gastrointestinal tumors.

[0005] Surface-enhanced Raman scattering (SERS), a powerful molecular spectroscopy technique, provides molecular "fingerprint" information (vibrational bands), enabling simultaneous, real-time detection of multiple components (with narrow half-width at half-maximum) and offering advantages such as portability, ease of operation, and cost-effectiveness. It demonstrates high sensitivity and potential for label-free molecular detection in complex biological sample analysis. The main enhancement mechanism of SERS, electromagnetic field enhancement, requires target molecule clusters to adsorb onto or near the surface of a plasmonic nanostructure to achieve indiscriminate amplification of the Raman signal. In particular, target molecules located in nano-intervals ("hot spots") can even achieve single-molecule sensitivity. Therefore, in label-free SERS detection systems for metabolic molecule clusters, constructing a high-density "hot spot" SERS-enhanced substrate is essential to obtain more comprehensive metabolic fingerprint information. Summary of the Invention

[0006] The purpose of this invention is to provide a triple-coupled SERS substrate, its preparation method, and its application, so as to solve the technical problem that existing SERS substrates cannot achieve highly sensitive, label-free, broad-spectrum metabolic fingerprint detection of gastrointestinal tumor metabolites in human serum due to insufficient hot spot density.

[0007] To address the aforementioned technical problems, this invention first provides a method for preparing a triple-coupled SERS substrate (a hydrophobic monolayer gold nanofilm / iodide-modified silver nanoparticle coupled SERS substrate, Ag@KI / HMGF SERS substrate), comprising the following steps: S10, the substrate is purified and surface functionalized in sequence, and the modified substrate is immersed in nano gold seed sol to obtain a functionalized substrate. S20: The functionalized substrate is placed in a growth reaction system containing gold ions, a first reducing agent is added, and the growth reaction is carried out by constant temperature shaking at room temperature. After the reaction is completed, the substrate is washed and dried. The above growth, washing and drying operations are repeated multiple times to obtain a gold film SERS substrate. S30, the gold film SERS substrate is purified by iodide ion source reagent, and then the surface of the purified gold film SERS substrate is modified by hydrophobic mercapto reagent. After drying, the hydrophobic modified gold film SERS substrate is obtained. S40, Ag@KI nanosol was deposited on the surface of a hydrophobically modified gold film SERS substrate to obtain a triple-coupled SERS substrate.

[0008] Specifically, step S10 thoroughly removes impurities and active interference sites from the substrate surface, while introducing stable active binding groups to provide a uniform and robust immobilization interface for the gold nanoparticles, preventing detachment or uneven distribution. Step S20, through multiple rounds of isothermal oscillation in-situ growth reactions, prepares a dense, flat, and continuous gold film substrate on the functionalized substrate surface, constructing the first core hot spot—the electromagnetic field hot spot generated by the gold film nano-intervals—forming a stable plasma enhancement platform. Simultaneously, the regular gold film structure provides excellent interface support for subsequent nanoparticle coupling, effectively avoiding the drawbacks of disordered hot spots and poor signal repeatability in traditional substrates. Step S30 uses an iodine ion source reagent for purification, precisely removing unreacted reactants and impurities remaining on the gold film surface, significantly reducing background interference in subsequent detection processes. Then, a hydrophobic thiol reagent is used for further purification. Surface modification imparts hydrophobic enrichment capabilities to the substrate, enabling efficient targeted enrichment of gastrointestinal tumor metabolites in serum. This concentrates target molecules in hotspot regions, further enhancing detection sensitivity. Step S40 deposits Ag@KI nanosol on the hydrophobically modified gold film surface, constructing a composite structure that couples a low-background noble metal platform with enrichment capabilities to background-free noble metal nanoparticles. This simultaneously forms three types of high-density clean hotspots: substrate hotspots generated by the gold film platform itself, interparticle hotspots formed between Ag@KI nanoparticles, and cross-interfacial hotspots generated by the coupling between Ag@KI nanoparticles and the gold film platform. This synergistic superposition of three hotspots achieves extreme amplification of the Raman signal, completely eliminating background interference and acquiring more comprehensive and richer SERS spectral information of serum metabolites. This allows for the simultaneous and accurate detection of both single-component and multi-component metabolites.

[0009] Preferably, in step S10, the substrate is first cleaned and dried by alternating acid and alkali reagents, and then its surface is functionalized by an amino-based silane coupling agent; the substrate is a glass slide, and the substrate is immersed in the nano-gold seed sol for 8 to 12 hours.

[0010] Specifically, glass slides have the advantages of convenient material sourcing, low cost, and smooth, clean surface, which is beneficial for the uniform growth of gold films and SERS detection. Alternating purification and cleaning with acid and alkali reagents can thoroughly remove oil, impurities, and active interference sites from the substrate surface, resulting in an ultra-clean substrate interface. Then, functionalization modification with amino-based silane coupling agents can introduce stable amino binding sites on the substrate surface, making the gold nanoparticles bond more firmly to the substrate and less prone to detachment. The optimal soaking time for the substrate in the gold nanoparticle sol is controlled within the range of 8 to 12 hours, which ensures that the gold nanoparticles can be fully and uniformly immobilized on the substrate surface, while avoiding insufficient gold loading due to too short a soaking time or gold agglomeration and stacking due to too long a soaking time.

[0011] Preferably, the gold nanoparticle sol in step S10 is prepared by liquid-phase reduction, and the preparation process is as follows: 400-600 μL of a first soluble gold salt solution, 1-2 mL of a coordination stabilizer and 400-600 μL of a second reducing agent are added sequentially to 40-60 mL of deionized water, and the gold nanoparticle sol is obtained after in-situ reduction reaction for 1-3 h.

[0012] Specifically, the parameter range of the above-mentioned liquid-phase reduction method can ensure that the concentration of the reaction system is moderate and the reduction process is mild and controllable, and can also precisely control the nucleation and growth rate of gold seeds, thereby preparing a gold nanoparticle sol with uniform particle size, good dispersibility, no agglomeration and high stability. At the same time, the above-mentioned preferred range can avoid problems such as uneven particle size, agglomeration and inactivation of gold seeds due to imbalance of raw material ratio or improper reaction time, so that the gold nanoparticles can be uniformly and firmly immobilized on the subsequent functionalized substrate.

[0013] Preferably, the first soluble gold salt solution is a 1 wt% chloroauric acid solution, the coordinating agent is a 10-15 mg / mL trisodium citrate solution, the second reducing agent is a 0.5-1 mg / mL sodium borohydride solution, and the particle size of the gold nanoparticle sol is 3-6 nm.

[0014] Specifically, a 1wt% chloroauric acid solution, as a soluble gold salt, provides a stable and high-purity gold source, ensuring the consistency of gold seed nucleation; a 10–15 mg / mL trisodium citrate solution achieves excellent dispersion stability within this concentration range, effectively inhibiting gold seed aggregation; a 0.5–1 mg / mL sodium borohydride solution, as a weak reducing agent, enables mild and controllable rapid in-situ reduction, avoiding uneven gold seed morphology caused by excessively fast or slow reduction rates; and the 3–6 nm nano-gold seeds have a small and uniform particle size range, enabling dense and uniform monodisperse immobilization on functionalized substrates, and providing an ideal seed base for the subsequent regular and smooth growth of gold films.

[0015] Preferably, step S20 specifically includes: S201, place the functionalized substrate in a reaction vessel, add 1-3 mL of deionized water, 300-400 μL of the second soluble gold salt solution and 100-200 μL of the first reducing reagent in sequence, and react at room temperature with constant temperature shaking at 200-300 r / min for 10-30 min. After taking it out, wash it and blow it dry with nitrogen. S202, add 1-3 mL of deionized water, 50-100 μL of the second soluble gold salt solution and 30-50 μL of the first reducing agent, and complete the shaking reaction, washing and drying operations under the same conditions. Repeat the above growth process 3-5 times to obtain the gold film SERS substrate.

[0016] Specifically, the S20 step adopts a stepwise growth strategy of first feeding high-concentration materials to form a film and then finely modifying it with low-concentration materials. Combined with a mild and controllable isothermal oscillation reaction at room temperature, the gold ion reduction process can be more stable and the gold film growth can be more uniform, effectively avoiding defects such as looseness, pores and rough morphology in the gold film. Then, repeating this growth process 3 to 5 times can build a dense, flat and highly continuous gold film SERS substrate layer by layer, forming a first core hot spot with uniform distribution and strong stability.

[0017] Preferably, the reaction vessel is a polypropylene centrifuge tube, the second soluble gold salt solution is a 0.2 wt% chloroauric acid solution, and the first reducing agent is a 0.02–0.06 mol / L hydroxylamine hydrochloride solution.

[0018] Specifically, polypropylene centrifuge tubes are chemically inert, do not release metal ions, and have good sealing properties, which can ensure a stable environment for the growth reaction system and prevent external contamination. At the same time, they can prevent reagent volatilization and make the gold film growth process more controllable. The 0.2wt% chloroauric acid solution has a moderate concentration and a stable gold source, which can provide sufficient gold ions for in-situ growth of gold film without causing gold particle agglomeration and rough gold film due to excessive concentration. The 0.02-0.06mol / L hydroxylamine hydrochloride solution has a mild reduction rate and matches the above gold salt concentration, which can achieve stable reduction and orderly deposition of gold ions, further ensuring the growth of a dense, flat, defect-free, and continuous gold film structure on the substrate surface.

[0019] Preferably, step S30 specifically includes: S301, immerse the gold film SERS substrate in 1-3 mL of a 10% concentration. -4 ~10 -2 After standing in a mol / L KI solution for 10–30 min, the substrate was removed and dried with nitrogen gas. This process was repeated twice to obtain the purified SERS substrate. The core mechanism of this purification process is: I with a low Raman scattering cross section. - It can strongly adsorb with Au and form Au-I bonds, effectively replacing organic ligands such as citrate ions remaining on the substrate surface, completely eliminating the background Raman signal interference caused by such substances. The background signal of the gold film after two KI treatments can be reduced to a negligible level, fully meeting the low background requirements of label-free metabolic SERS detection. S302, the purified gold film SERS substrate is immersed in 1-3 mL of 3 wt% PFDT ethanol solution, left to stand for 4-8 h, and then removed and dried to obtain the hydrophobic modified gold film SERS substrate.

[0020] Specifically, step S301 gently and thoroughly removes residual organic ligands from the gold membrane surface, cleans the active sites, and significantly reduces background interference, while optimizing the first hotspot structure of the gold membrane itself. Step S302 allows hydrophobic groups to be fully and uniformly grafted onto the gold membrane surface, giving the substrate excellent hydrophobic enrichment ability and efficiently concentrating tumor metabolites in serum to the hotspot region. The two optimization processes work together to obtain a clean, low-background, and highly enriched hydrophobic gold membrane substrate, providing an ideal interface for subsequent deposition of Ag@KI nanosols and construction of triple-coupled high-density clean hotspots.

[0021] Preferably, the preparation process of Ag@KI nanosol in step S40 is as follows: The aqueous solution of silver precursor was heated to boiling, and a water-soluble organic acid coordination reducing agent was added under stirring. The reaction was carried out at a constant temperature until complete, and silver nanoparticle sol coated with organic ligands was obtained. The silver nanoparticle sol was subjected to solid-liquid separation, and the supernatant was discarded to prepare a dispersion. The silver nanoparticles in the dispersion were subjected to surface ligand replacement using an iodine ion source reagent, followed by solid-liquid separation and discarding of the supernatant. After redispersement, Ag@KI nanosol was obtained.

[0022] Specifically, the above preparation method involves first reducing the silver precursor under boiling conditions using a water-soluble organic acid coordination reducing agent to prepare an organic ligand-coated silver nanoparticle sol with uniform particle size, excellent dispersibility, and strong stability, thus avoiding the aggregation and uneven morphology of silver nanoparticles. Then, the supernatant is discarded after solid-liquid separation for purification, which can effectively remove unreacted impurities and free ligands in the system and eliminate potential interference in subsequent detection. Finally, the surface ligands are replaced using an iodine ion source reagent to obtain Ag@KI nanosol with a clean surface and no complex organic background, which significantly reduces the background signal of SERS detection and optimizes the plasma performance of silver nanoparticles.

[0023] Accordingly, the present invention also provides a triple-coupled SERS substrate, which is prepared by the above-described method for preparing a triple-coupled SERS substrate.

[0024] Specifically, the triple-coupled SERS substrate prepared in this invention has a regular and controllable structure and excellent stable performance. Through multi-level coupling of the gold film substrate and Ag@KI nanoparticles, it constructs three high-density clean hot spots: the gold film itself, the spaces between Ag@KI nanoparticles, and the interface between Ag@KI nanoparticles and the gold film, which have extremely strong Raman signal amplification capabilities. At the same time, the above-mentioned triple-coupled SERS substrate has clean, low background and hydrophobic enrichment characteristics, which can efficiently enrich gastrointestinal tumor metabolites in serum and obtain label-free metabolic fingerprint spectra with high signal-to-noise ratio and high information content.

[0025] Accordingly, the present invention also provides an application of the above-mentioned triple-coupled SERS substrate, which is used for the detection of gastrointestinal tumor metabolites in human serum.

[0026] Specifically, the aforementioned triple-coupled SERS substrate, when applied to the detection of gastrointestinal tumor metabolites in human serum, leverages its superior signal amplification capabilities, hydrophobic enrichment properties, and low background interference due to its triple high-density clean hotspots. This enables highly sensitive, label-free, and broad-spectrum metabolic fingerprinting of tumor-related metabolites in serum, eliminating the need for complex sample pretreatment. The detection process is rapid, minimally invasive, and convenient, effectively distinguishing the metabolic characteristics of different gastrointestinal tumors. Combined with spectral analysis, it achieves high accuracy, high sensitivity, and high specificity in screening and precise identification of gastrointestinal tumors. This provides an efficient and reliable new detection method for non-invasive early diagnosis and subtyping of clinical gastrointestinal tumors, demonstrating significant clinical application value and promising prospects for wider application.

[0027] The beneficial effects of this invention are as follows: Unlike existing technologies, this invention provides a triple-coupled SERS substrate, its preparation method, and its application. The preparation method involves sequentially purifying and functionalizing the substrate, performing stepwise multi-round in-situ growth to form a dense gold film platform, iodine ion purification and impurity removal combined with hydrophobic modification to achieve low background and hydrophobic enrichment, and then depositing clean, highly active Ag@KI nanosol to complete interfacial coupling. This constructs a triple high-density clean hotspot layer by layer: the gold film itself, the spaces between Ag@KI nanoparticles, and the interface between Ag@KI nanoparticles and the gold film. This method significantly improves Raman signal amplification and detection signal-to-noise ratio. It features controllable process, stable substrate performance, and excellent metabolite enrichment with low interference. It can directly achieve highly sensitive, label-free, and broad-spectrum metabolic fingerprint detection of gastrointestinal tumor metabolites in human serum. This effectively solves the problems of insufficient substrate hotspots, large background interference, and low sensitivity in traditional SERS, as well as the highly invasive, poor specificity, and cumbersome operation of existing tumor diagnostic techniques. It provides an efficient and reliable new approach for non-invasive, rapid, and accurate screening and identification of gastrointestinal tumors in clinical practice, possessing outstanding clinical application value. Attached Figure Description

[0028] Figure 1 This invention provides a schematic diagram of the overall technical route for using a triple-coupled hotspot SERS substrate to obtain a more comprehensive serum metabolic fingerprint, and combining it with machine learning to analyze SERS spectra to achieve the identification of liver cancer (LC), gastric cancer (GC), colorectal cancer (CRC) and normal populations.

[0029] Figure 2 Preparation and characterization of hydrophobic monolayer gold (HMGF) SERS substrate: (A) is a flowchart of the preparation of the HMGF SERS substrate; (B) is a flowchart of the preparation of the HMGF SERS substrate with a concentration of 5 × 10⁻⁶. -5M is 4-mercaptoaniline (4-ATP), the Raman signaling molecule. SERS spectra of monolayer gold nanofilm substrates prepared with different amounts of chloroauric acid tetrahydrate (HAuCl4) and different in-situ growth cycles are shown; (C) is... Figure 2 (B) Scanning electron microscopy (SEM) image of the HMGF substrate with the best SERS enhancement effect; (D) Image of the substrate with the best SERS enhancement effect. Figure 2 (C) shows the electromagnetic field distribution cloud map of the HMGF substrate obtained by simulating the HMGF morphology under 785nm laser excitation, which is a model; (E) is the Raman mapping image of the HMGF substrate; (F) is the static contact angle test result of the HMGF substrate after different times of hydrophobic treatment with PFDT; (G) is a comparison of the Raman spectra of the HMGF substrate before, after, and after KI treatment with PFDT.

[0030] Figure 3 Performance characterization of the triple-coupled Ag@KI / HMGF SERS substrate: (A) Schematic diagram of the evaporation process of Ag@KI nanoparticle (Ag@KINPs) sol and analyte mixture on the HMGF substrate; (B) Concentration process of Ag@KI NPs sol and analyte mixture droplets at different time points of 0 min, 20 min, 40 min and 60 min; (C) Crystal violet (CV, concentration 10) obtained by three different SERS substrates. -8 (M) is the SERS spectrum of (M); (D) is the SERS spectrum of (M) is the SERS spectrum of ... Figure 3 (C) Three different SERS substrates at 723 cm -1 The image shows a comparison of SERS intensities at the three different substrates. (E), (F), and (G) are SERS spectra of single-component solutions and mixed solutions of five small metabolic molecules (phenylalanine, cholesterol, tryptophan, glutathione, and hypoxanthine) obtained from three different SERS substrates (E: HMGF, F: Ag@KI / HG, G: Ag@KI / HMGF). (H) is a comparison of SERS spectra of the mixed solution of the five small metabolic molecules on three different SERS substrates.

[0031] Figure 4 Analysis of metabolic SERS spectral characteristics of serum samples from different groups: (A) Schematic diagram of serum metabolic SERS detection; (B) Normalized average SERS spectra of serum samples from different groups after preprocessing; (C) SERS difference spectra of serum samples from different groups; (D) Histogram of SERS characteristic peak intensities of serum samples from different groups, with annotations... P < 0.05 indicates that the difference between groups is statistically significant; (E) is the result of analyzing the SERS data of serum samples from different groups using three-dimensional principal component analysis (3D PCA); (F) is the result of analyzing the SERS data of serum samples from different groups using three-dimensional linear discriminant analysis (3D LDA).

[0032] Figure 5 To classify and identify samples from the normal group, liver cancer group, colorectal cancer group, and gastric cancer group based on machine learning: (A) is a flowchart of the machine learning workflow applicable to the classification of the four groups of samples (Normal Group, Liver Cancer Group, Colorectal Cancer Group, and Gastric Cancer Group); (B) and (C) are the training set and validation set loss curves and training set and validation set accuracy curves obtained by using the LightGBM machine learning model, respectively; (D) and (E) are the confusion matrix and ROC curve of the validation set of the four groups of samples (Normal Group, Liver Cancer Group, Colorectal Cancer Group, and Gastric Cancer Group), respectively, where AUC refers to the area under the ROC curve; (F) is an analysis chart of the SHAP values ​​corresponding to the top 10 important difference frequency bands in the LightGBM machine learning model; (G) is a comparison chart of the classification accuracy of four machine learning models: CatBoost (CB), ExtraTrees (ET), LightGBM (LGBM), and RandomForest (RF).

[0033] Figure 6 The image shows the UV-Vis absorption spectrum characterization of the gold nanoparticle sol prepared in Example 2.

[0034] Figure 7 Scanning electron microscope (SEM) images of monolayer gold nanofilms under different preparation conditions: (A) to (C) show the SEM images of monolayer gold nanofilms prepared under the condition that the number of in-situ growth cycles is kept constant, while the amount of chloroauric acid tetrahydrate (HAuCl4) added for the first time is adjusted to 320 μL, 360 μL, and 400 μL, respectively; (D) to (F) show the SEM images of monolayer gold nanofilms prepared under the condition that the amount of chloroauric acid tetrahydrate (HAuCl4) added for the first time is kept constant at 360 μL, while the number of in-situ growth cycles is adjusted to 4, 5, and 6, respectively.

[0035] Figure 8 Characterization of SERS enhancement uniformity of monolayer gold nanofilms: (A) SERS spectra of 4-mercaptoaniline (4-ATP) corresponding to 50 randomly selected detection sites on the surface of the monolayer gold nanofilm; (B) SERS spectra of the above 50 detection sites at 1076 cm⁻¹. -1 The variation of SERS intensity distribution at wavenumber.

[0036] Figure 9 SERS stability and reproducibility characterization of multiple batches of monolayer gold nanofilm substrates: (A) SERS spectra of 8 batches of parallel-prepared monolayer gold nanofilm substrates modified with 4-mercaptoaniline (4-ATP); (B) SERS spectra of the above 8 batches of different monolayer gold nanofilm substrates at 1076 cm⁻¹. -1 The SERS intensity distribution variation at wavenumber was plotted by selecting a single-layer gold nanofilm substrate for each batch and randomly selecting 6 detection sites on each substrate for testing.

[0037] Figure 10 Fourier transform infrared spectrum of hydrophobic monolayer gold film (HMGF) SERS substrate.

[0038] Figure 11 Characterization of the uniformity of hydrophobic modification on the surface of HMGF SERS substrate: (A) Infrared mapping images of different randomly selected regions on the surface of HMGF SERS substrate; (B) Bright-field topography of the substrate corresponding to the infrared mapping region.

[0039] Figure 12 A single-layer gold nanofilm with a concentration of 10 -3 Comparison of Raman signals before treatment with KI solution of M, and after treatment 1, 2, and 3 times respectively.

[0040] Figure 13 The image shows a comparison of the Raman spectra of silver nanoparticles (Ag NPs) sol before and after KI treatment.

[0041] Figure 14 This is a scanning electron microscope image of the Ag@KI nanoparticle sol.

[0042] Figure 15 The image shows a comparison of the UV-Vis absorption spectra of Ag NPs sol before and after KI treatment.

[0043] Figure 16 Dynamic light scattering particle size characterization of Ag NPs sol before and after KI treatment: (A) is the particle size distribution curve of Ag NPs sol before KI treatment measured by dynamic light scattering (DLS); (B) is the particle size distribution curve of Ag@KI nanosol after KI treatment measured by dynamic light scattering.

[0044] Figure 17 This is a comparison of the zeta potentials of Ag NPs sol before and after KI treatment.

[0045] Figure 18 Scanning electron microscope image of Ag@KI nanoparticles deposited on HMGF SERS substrate.

[0046] Figure 19Electromagnetic field simulation characterization of the substrate using the finite difference time-domain method: (A) Schematic diagram of the finite difference time-domain method (FDTD) simulation model construction; (B) Simulated electromagnetic field distribution of the HMGF substrate; (C) Simulated electromagnetic field distribution of the Ag@KI nanoparticle-HMGF coupled system; (D) Simulated electromagnetic field distribution of the coupling effect between adjacent Ag@KI nanoparticles; The structure constructed in this simulation is based on... Figure 7 (B) and Figure 18 The scanning electron microscope morphology setup shown is as follows: the constructed silver nanoparticles (Ag NPs) have a diameter of 50 nm and a width of 400 nm, and the excitation light source wavelength is set to 785 nm.

[0047] Figure 20 SERS performance factor calculation and fitting analysis for Ag@KI / HMGF substrate: (A) SERS spectra of crystal violet (CV) solutions of different concentrations on Ag@KI / HMGF substrate; (B) SERS spectra of the corresponding CV solutions in (A) at 723 cm⁻¹. -1 (C) is a graph showing the SERS intensity fitting relationship at the wavenumber, which conforms to the Langmuir adsorption isotherm model; (D) is a graph showing the linear fitting curve of CV concentration and SERS intensity corresponding to the area selected in Figure (B); (D) is a conventional Raman spectrum of crystal violet (CV) solutions of different concentrations on a hydrophobic glass (HG) substrate; (E) is a graph showing the linear fitting curve of CV concentration and Raman intensity corresponding to Figure (D); (F) is a schematic diagram of the SERS performance factor (SPF) calculation process, where the SPF value is the ratio of the slope of the linear fitting curves in Figure (C) and Figure (E).

[0048] Figure 21 The image shows a colloidal CV-Ag@KI nanoparticles deposited on an HMGF SERS substrate. The image also shows the morphology of a mixed sol of crystal violet (CV) and Ag@KI nanoparticles deposited on an HMGF SERS substrate after evaporation enrichment treatment, used to characterize the evaporation concentration and deposition effects of the mixed sol.

[0049] Figure 22 Characterization of reproducibility of Ag@KI / HMGF SERS on substrate: (A) Crystal violet (CV) SERS spectra of 20 randomly selected detection sites on the Ag@KI / HMGF SERS substrate surface; (B) Reproducible SERS spectra of the above 20 detection sites at 723 cm⁻¹. -1 The variation of SERS intensity distribution at wavenumber.

[0050] Figure 23Comparison of SERS spectra of CV solutions of different concentrations on three types of SERS substrates: (A) SERS spectra of crystal violet (CV) solutions of different concentrations on HMGF substrate; (B) SERS spectra of crystal violet (CV) solutions of different concentrations on Ag@KI / HG substrate; (C) SERS spectra of crystal violet (CV) solutions of different concentrations on Ag@KI / HMGF substrate.

[0051] Figure 24 The loss curves for three types of machine learning models are shown below: (A) is the loss curve for the ExtraTrees (ET) machine learning model; (B) is the loss curve for the RandomForest (RF) machine learning model; and (C) is the loss curve for the CatBoost (CB) machine learning model.

[0052] Figure 25 The following are the accuracy curves for three types of machine learning model analysis: (A) is the accuracy curve for the ExtraTrees (ET) machine learning model analysis; (B) is the accuracy curve for the RandomForest (RF) machine learning model analysis; and (C) is the accuracy curve for the CatBoost (CB) machine learning model analysis.

[0053] Figure 26 The following are the confusion matrix diagrams for four training sets of four types of machine learning models: (A) Confusion matrix diagram for ExtraTrees (ET) model for normal group (NG), liver cancer group (LC), colorectal cancer group (CRC), and gastric cancer group (GC); (B) Confusion matrix diagram for RandomForest (RF) model for the above four training sets; (C) Confusion matrix diagram for CatBoost (CB) model for the above four training sets; (D) Confusion matrix diagram for LightGBM (LGBM) model for the above four training sets.

[0054] Figure 27 The confusion matrix diagrams of four sample test sets were analyzed for three types of machine learning models: (A) The confusion matrix diagram of the ExtraTrees (ET) model for the normal group (NG), liver cancer group (LC), colorectal cancer group (CRC), and gastric cancer group (GC) test sets; (B) The confusion matrix diagram of the RandomForest (RF) model for the above four sample test sets; (C) The confusion matrix diagram of the CatBoost (CB) model for the above four sample test sets.

[0055] Figure 28The ROC curves of four training sets of four types of machine learning models were analyzed: (A) Receiver Operating Characteristic (ROC) curves of the ExtraTrees (ET) model for the normal group (NG), liver cancer group (LC), colorectal cancer group (CRC), and gastric cancer group (GC); (B) ROC curves of the RandomForest (RF) model for the above four training sets; (C) ROC curves of the CatBoost (CB) model for the above four training sets; (D) ROC curves of the LightGBM (LGBM) model for the above four training sets.

[0056] Figure 29 The ROC curves of four test sets were analyzed for three types of machine learning models: (A) Receiver operating characteristic (ROC) curves of the test sets of normal group (NG), liver cancer group (LC), colorectal cancer group (CRC), and gastric cancer group (GC) were analyzed for the ExtraTrees (ET) model; (B) ROC curves of the above four test sets were analyzed for the RandomForest (RF) model; and (C) ROC curves of the above four test sets were analyzed for the CatBoost (CB) model. Detailed Implementation

[0057] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0058] Addressing the technical shortcomings of existing SERS substrates, serum metabolic analysis based on label-free surface-enhanced Raman scattering (SERS) fingerprinting is a highly promising tumor detection technique. However, current SERS substrates generally suffer from insufficient hotspot density, resulting in incomplete molecular Raman spectral information and hindering accurate differentiation of different tumor types. Highly sensitive multi-component SERS detection is achieved primarily through two methods: first, constructing a noble metal platform with molecular concentration properties to enrich analyte molecules and enhance their signal using the SERS platform; second, preparing plasmonic nanosols, mixing them with the target analyte, and then performing SERS detection in a homogeneous system or after sample drying. For label-free metabolic fingerprinting, the most effective strategy to obtain richer molecular spectral fingerprint information is to increase the hotspot density and spatial distribution range of the substrate.

[0059] Based on this, the present invention proposes an efficient and feasible technical solution: constructing a low-background noble metal platform with molecular enrichment function, and precisely coupling it with noble metal nanoparticles free from background interference to form a triple high-density clean hot spot. Specifically, this hot spot covers three types: between noble metal nanoparticles, within the noble metal platform itself, and at the interface between noble metal nanoparticles and the noble metal platform. This eliminates detection interference and obtains more informative metabolite SERS spectra. Addressing the complex and heterogeneous characteristics of serum metabolic fingerprint Raman spectra, the present invention simultaneously incorporates machine learning technology to achieve efficient analysis of complex spectral features and accurate classification of large batches of sample data.

[0060] To meet the aforementioned requirements for highly sensitive label-free detection, this invention further constructs a high-density clean hotspot SERS substrate (Ag@KI / HMGF) based on the coupling of background-free silver nanoparticles (Ag@KINPs) and a self-concentrating low-background gold nanoparticle film. This aims to obtain more comprehensive and complete SERS characteristic bands of metabolic small molecules. To avoid the interference of the "protein crown" effect in serum on metabolic fingerprint analysis, ultrafiltration separation technology is used to achieve efficient separation of metabolic small molecules and proteins. The separated samples of various metabolic small molecules are thoroughly mixed with silver nanoparticle colloids and then dropped onto the surface of a monolayer gold nanofilm modified with perfluorosilane thiol. This coupled hydrophobic substrate can effectively avoid the "coffee ring effect" in the detection process and significantly improve the hotspot density of the substrate, thereby obtaining comprehensive and complete SERS spectra of metabolic small molecules.

[0061] The Ag@KI / HMGF SERS substrate described above was used to detect four groups of serum samples (normal group, liver cancer group, colorectal cancer group, and gastric cancer group), and a systematic data analysis of the collected SERS spectra was conducted using machine learning algorithms. The results showed that the LightGBM model can accurately distinguish normal samples from liver cancer, colorectal cancer, and gastric cancer samples, with a classification accuracy exceeding 90.4%, effectively enabling accurate identification and efficient screening of gastrointestinal tumors. These experimental results confirm that the universal Ag@KI / HMGF SERS substrate developed in this invention has extremely high application potential and promotional value in the field of serum and various biological fluid metabolic fingerprint analysis.

[0062] Please see Figure 1 , Figure 1 This invention provides a schematic diagram of the overall technical approach for using a triple-coupled hotspot SERS substrate to obtain a more comprehensive serum metabolic fingerprint, and combining it with machine learning to analyze SERS spectra to identify liver cancer (LC), gastric cancer (GC), colorectal cancer (CRC) and normal populations; wherein, Figure 1First, the hotspot and signal performance of three SERS substrates were compared. The left side shows the Ag@KI / HMGF substrate with triple-coupled hotspots, while the middle and right sides show the Ag@KI / HG and HMGF substrates with single-coupled hotspots. Under 785nm laser excitation, the triple-coupled substrate exhibited three electromagnetic field enhancement modes: gold film gap, silver particle gap, and gold-silver vertical coupling, corresponding to the output of rich Raman fingerprints. In contrast, the two single-coupled substrates could only output moderate Raman fingerprints and trace Raman fingerprints, which intuitively demonstrates that the triple-coupled substrate can obtain more comprehensive metabolic molecular characteristic bands.

[0063] Figure 1 The complete detection process is shown below: After incubating the serum sample with Ag@KI nanoparticles for 10 minutes, it was dropped onto the Ag@KI / HMGF substrate. SERS spectra were acquired using a 50x785nm laser excitation method, yielding Raman spectra for four groups of samples: GC, CRC, LC, and Normal (the horizontal axis represents Raman Shift (cm)). -1 The vertical axis represents Intensity (au); then the spectral data is input into the Machine Learning model to complete Cancer Screening, accurately distinguishing four types of samples: Normal (normal population), CRC (colorectal cancer), LC (liver cancer), and GC (gastric cancer); the bottom legend labels each core component and substrate: Ag@KI / HMGF (triple-coupled SERS substrate), Ag@KI / HG (single-coupled silver-hydrophobic glass substrate), HMGF (single-coupled hydrophobic gold film substrate), Analytes (analytes to be detected), Ag@KI (KI-modified silver nanoparticles), and Serum (serum sample).

[0064] The technical solution of the present invention will now be further described with reference to specific embodiments.

[0065] Example 1 (Experimental Reagents and Instruments): (1.1) Experimental reagents: Chloroauric acid tetrahydrate (AuCl3·HCl·4H2O), trisodium citrate (C6H5Na3O7), silver nitrate (AgNO3), sodium borohydride (NaBH4), sodium hydroxide (NaOH), tryptophan (L-Trp), pyruvic acid (96%), 3-aminopropyltrimethoxysilane (3-APTMS), crystal violet (CV), (1H,1H,2H,2H)-perfluorodecanethiol (PFDT) were all from Shanghai Aladdin Biochemical Technology Co., Ltd., anhydrous ethanol, hydroxylamine hydrochloride (NH2OH·HCl), potassium iodide (KI) and glucose (C6H5Na3O7) were all from Shanghai Aladdin Biochemical Technology Co., Ltd. 12 O6 was obtained from Sinopharm Chemical Reagent Co., Ltd., crystal violet (CV) from Shanghai Aladdin Biochemical Technology Co., Ltd., and 4-mercaptoaniline (4-ATP) from Shanghai Haohong Biomedical Technology Co., Ltd. 3 kDa ultrafiltration centrifuge tubes were purchased from Pall Corporation (Beijing, China). All purchased reagents were used directly without further purification. Ultrapure water (18.25 MΩ) was used as the solvent throughout the experiment.

[0066] (1.2) Experimental Apparatus: Dynamic light scattering spectrometer (DLS, Zetasizer Nano ZSP, UK) for measuring hydrodynamic diameter and zeta potential; UV-Vis spectrometer (Duetta, HORIBA, Kyoto) for measuring the extinction spectrum of nanoparticles; Fourier transform infrared spectrometer (FT-IR, Nicoleti N10, Thermo, USA) for characterizing the chemical composition of the SERS hydrophobic substrate and performing micro-infrared imaging; Field emission scanning electron microscope (FESEM, Zeiss SIGMA) for characterizing the morphology of the SERS substrate and AgNPs@KI sol; Confocal microRaman spectrometer (XploRA plus, HORIBA Jobin) Yvon, Kyoto), used for Raman imaging characterization of the substrate; Raman imaging test conditions: 785nm, 10mW laser excitation, 1s acquisition time per integration, 50× magnification, 5μm step size; portable surface-enhanced Raman detection system (FI785E10W-Pro-WD01, Beijing, China), used to acquire Raman spectra of different samples; detection conditions: 785nm, 10mW laser excitation, 2s acquisition time per integration, 50× objective lens; constant temperature shaking chamber, used for in-situ growth of gold nanofilms; high-speed refrigerated centrifuge, used for serum pretreatment and centrifugal purification of nanosols; computer and data processing software (LabSpec6, Origin, Python), used for spectral preprocessing, data analysis, and machine learning model construction.

[0067] Example 2 (Preparation of gold nanoparticle sol): Take a 150mL round-bottom flask, add 50mL of ultrapure water, then add 500μL of 1wt% chloroauric acid tetrahydrate solution, 1.5mL of 11.4mg / mL trisodium citrate solution (trisodium citrate as a stabilizer), and finally add 500μL of freshly prepared sodium borohydride solution (0.8mg / mL). Stir the reaction at room temperature for 2 hours until the solution turns wine-red, thus obtaining a gold nanoparticle sol with a particle size of 5nm. The gold nanoparticle sol was detected using UV-Vis spectroscopy, and its maximum absorption peak was located at 516nm. Figure 6 As shown, this demonstrates the successful synthesis of gold nanoparticle sol.

[0068] Example 3 (Preparation of hydrophobically modified gold film SERS substrate): (3.1) Pretreatment and amination of glass slides: 2×1cm glass slides were cut and ultrasonically cleaned with aqua regia and alkali solution to remove surface impurities. After cleaning, the slides were rinsed with ultrapure water and dried. The pretreated glass slides were immersed in 0.1wt% APTMS (3-aminopropyltrimethoxysilane) ethanol solution for 20min. After removal, the slides were dried with nitrogen gas. The amination of the glass slides was completed by bonding the siloxy groups at the ends of the APTMS molecules with the hydroxyl groups on the surface of the glass slides.

[0069] (3.2) Electrostatic adsorption of gold nanoseed sol: The aminated glass slide was immersed in 3.5 mL of the gold nanoseed sol prepared in Example 2 and allowed to stand at room temperature for 10 h for adsorption. After removal, it was dried with nitrogen gas. Through the electrostatic adsorption between N and Au, the gold nanoseeds were fixed on the surface of the glass slide to obtain a glass slide substrate loaded with Au nanoseeds.

[0070] (3.3) In-situ growth and process optimization of Au nanofilms: A glass slide loaded with Au nanoseeds was placed in an EP tube. 2 mL of ultrapure water, 360 μL of 0.2 wt% chloroauric acid tetrahydrate solution, and 180 μL of 0.04 M hydroxylamine hydrochloride solution were added to the EP tube. The EP tube was placed in a constant temperature shaking oven and shaken at 240 r / min for 20 min at room temperature. After the reaction, the glass slide was removed, rinsed with ultrapure water, and dried with nitrogen. The glass slide was then placed in a new EP tube, and 2 mL of ultrapure water, 80 μL of 0.2 wt% chloroauric acid tetrahydrate solution, and 40 μL of 0.04 M hydroxylamine hydrochloride solution were added. The reaction was shaken at 240 r / min for 20 min at room temperature. The slide was then removed, cleaned, and dried with nitrogen. This growth operation was repeated 4 times to obtain a monolayer Au nanofilm (gold film SERS substrate).

[0071] (3.3.1) Using 4-ATP as the Raman signal mode molecule, the SERS activity of Au nanofilms was optimized by adjusting the amount of the first chloroauric acid tetrahydrate (320 μL, 360 μL, 400 μL) and the total number of in-situ growths (4, 5, 6 times). The optimization results and mechanisms were as follows: with increasing chloroauric acid concentration, the SERS enhancement effect of the gold nanofilm first increased and then decreased. This is because the gaps between the gold nanostructures decreased, which improved the SERS activity of the substrate. However, further growth of the gold nanostructures led to a decrease in the density of the nano-gap, thus weakening the SERS activity of the gold nanofilm. Similarly, with increasing the number of growths, the SERS effect of the gold nanofilm first increased and then decreased. This is because the reduced nano-gap increased the SERS activity, while further increasing the number of growths resulted in nanostructure stacking, leading to a weakening of the SERS enhancement effect. Raman spectra under different conditions are shown in [reference needed]. Figure 2 As shown in (B), the SEM morphology of Au nanofilms prepared under different conditions is shown in [Figure 1]. Figure 7 As shown in (A) to (F), the optimal process parameters were determined to be 360 ​​μL of chloroauric acid tetrahydrate for the first time and 4 total in-situ growth cycles. Under these conditions, the Au nanofilm prepared showed the best SERS enhancement effect.

[0072] (3.4) KI cleaning treatment of Au nanofilm: The optimized monolayer Au nanofilm was immersed in 2 mL of 10 -3 In a KI aqueous solution of M, let stand at room temperature for 20 minutes, then remove and dry with nitrogen gas; repeat the above KI soaking-drying operation twice, using I - The strong adsorption of Au forms Au-I bonds, replacing impurities such as citrate adsorbed on the substrate surface and eliminating background Raman signals.

[0073] (3.4.1) Verification of cleaning effect: Raman spectroscopy showed that the background signal was significantly reduced after one KI treatment, and the background signal of the gold nanofilm was almost negligible after two KI treatments. Figure 12 Since no significant additional optimization was found after 3 treatments, 2 treatments were determined to be the optimal number of treatments. This background removal step is easily overlooked for label-free metabolic SERS detection but should be given special consideration.

[0074] (3.5) Preparation of HMGF (hydrophobically modified gold film SERS substrate) by PFDT hydrophobic modification: In order to obtain coupled multiple hot spots (the distance between silver nanoparticles and gold nanofilm is less than 10 nm), and to enable the subsequent addition of liquid analyte droplets to self-concentrate to improve detection sensitivity and effectively avoid the "coffee ring" effect, small molecule PFDT was selected as a hydrophobic agent (instead of polymers such as PDMS) to construct a stable superhydrophobic plasma nanostructure surface. This modification is beneficial to achieve "top-bottom" coupling between silver nanoparticles and gold nanofilm to improve hot spot density; Au nanofilm cleaned with KI was immersed in 2 mL of ethanol solution containing 3 wt% PFDT (perfluorodecyl mercaptan) and left to stand at room temperature for 6 h to allow PFDT to form a self-assembled monolayer on the surface of Au nanofilm; after removal, the glass slide was placed in an oven to dry, and the hydrophobic monolayer gold film (HMGF) SERS substrate was obtained, as shown below. Figure 2 As shown in (A).

[0075] (3.5.1) Comprehensive verification of hydrophobic modification of the hydrophobic gold film SERS substrate: Static contact angle test: After hydrophobic treatment for 0h, 2h, 4h, 6h, 8h, and 10h, the substrate contact angle gradually increased from 88° to 104°, 117°, 130°, and 136°, respectively, and stabilized after 6h. Figure 2 (F) Therefore, 6 hours was determined to be the optimal modification time; FT-IR detection: substrate at 1140 cm⁻¹ -1 The characteristic absorption band of CF bond stretching vibration appears at this location. Figure 10 This confirmed that PFDT was successfully modified onto the surface of the gold nanofilm; micro-infrared mapping test: four areas were randomly selected on the substrate surface for detection, and the results proved that PFDT was uniformly distributed on the substrate surface. Figure 11 Raman interference verification: Comparison of HMGF Raman spectra before KI treatment, after KI treatment, and after PFDT modification showed that the PFDT-modified gold nanofilm had no broad characteristic spectrum. Figure 2 The presence of HMGF (G) indicates that the prepared HMGF does not interfere with SERS metabolic fingerprint analysis, thus solving the spectral interference problem that may be caused by hydrophobic macromolecules such as PDMS.

[0076] (3.5.2) Comprehensive characterization of the properties of the hydrophobically modified gold film SERS substrate: morphology and electromagnetic field simulation: its SEM morphology is shown in [reference needed]. Figure 2 In the middle (C), based on this morphology, a very similar drawing and modeling was performed using the FDTD (Finite-Difference Time-Domain Method), setting a 785nm laser as the excitation source and a 2×2×2nm wavelength. 3To achieve uniform mesh size, a significant enhancement of the electric field was observed near the edges of the gold nanostructures, and this high enhancement was confined to the gaps between the nanostructures. Figure 2 (D) This is attributed to the strong coupling of local surface plasmon polaritons between neighboring nanostructures, which enhances the local field and supports the SERS signal amplification effect of the prepared gold nanofilm; Raman imaging uniformity: Raman mapping imaging of 4-ATP in a 600×600μm square region showed that the SERS enhancement of the gold nanofilm was uniform. Figure 2 (E)); Quantitative uniformity: 50 sites were randomly selected on a single gold nanofilm, and the results were measured at 1076 cm⁻¹. -1 The SERS spectral data at the site demonstrate excellent uniformity of SERS enhancement on the substrate (RSD of 4.44%). Figure 8 (A) to (B)); Preparation reproducibility: Eight batches of HMGF substrates were prepared in parallel, and six points were randomly selected from each batch for detection at 1076 cm⁻¹. -1 SERS spectral results at [location missing] demonstrate that the gold nanofilm prepared by this method exhibits good stability and reproducibility (RSD of 8.30%). Figure 9 (A) to (B)); The uniformity of the enhancement effect of the gold nanofilm on the two-dimensional plane is beneficial to the stability of its SERS signal.

[0077] Example 4 (Preparation of Ag@KI nanosol): (4.1) Preparation of Ag nanosol: Take a 250mL round bottom flask, add 100mL of 1mM silver nitrate aqueous solution, heat the flask to 120℃, adjust the stirring rate to 1080r / min, and when the solution is in a state of slight boiling, quickly add 1.8mL of 1wt% trisodium citrate solution. After a few seconds, the solution turns yellow-brown. Continue the slight boiling reaction for 1h, and the solution eventually turns gray-green, thus obtaining AgNPs sol (Ag nanosol). The AgNPs sol shows a strong background Raman signal.

[0078] (4.2) Preparation of Ag@KI nanosol by KI modification: Take three 4.5 mL portions of the above AgNPs sol and place them in three centrifuge tubes respectively. Centrifuge at 3000 r / min for 15 min, remove the supernatant, concentrate the precipitate to 1 / 3 of the original volume, and then ultrasonically disperse. Add 50 μL of 1 mM KI aqueous solution to each centrifuge tube, mix thoroughly, and let stand at 25 °C for 20 min. Centrifuge again at 3000 r / min for 15 min, remove the supernatant, and redisperse the precipitate to the volume before centrifugation to obtain I - Modified Ag@KI nanosol, which has almost no background interference.

[0079] (4.2.1) Comprehensive verification of modification effect: Raman background: Raman spectroscopy showed that the background Raman signal of Ag@KI nanosol was almost negligible compared with the original Ag NPs sol. Figure 13 Morphology and dispersibility: SEM characterization showed that Ag@KI nanoparticles exhibited good monodisperse morphology. Figure 14 Optical and particle size properties: The UV-Vis absorption spectra of Ag NPs before and after KI treatment remained essentially unchanged. Figure 15 DLS testing showed that the hydrated particle size changed slightly from 50.5 nm to 48.7 nm. Figure 16 (A) to (B) indicate that the surface ligand replacement process did not cause AgNPs to aggregate, and they remained in a monodisperse state, which is beneficial to the stability of the label-free SERS signal; zeta potential: zeta potential testing showed that the zeta potential of the Ag@KI NPs sol (-114.5) was significantly lower than that of the original Ag NPs sol (-87.8). Figure 17 This indicates that the citrate ions on the surface were successfully replaced by iodide ions.

[0080] Example 5 (Construction and performance characterization of a triple-coupled Ag@KI / HMGF SERS substrate): (5.1) Substrate construction: The Ag@KI nanosol prepared in Example 4 was mixed with the analyte to be detected at a volume ratio of 1:1. 5 μL of the mixed solution was dropped onto the surface of the HMGF SERS substrate prepared in Example 3 and allowed to dry naturally at room temperature to obtain the Ag@KI / HMGF SERS substrate with triple coupling hot spots.

[0081] (5.2) Verification of Plasmon Coupling Effect (FDTD Simulation): To verify whether the designed SERS substrate has multiple plasmon coupling effects, the electromagnetic field distribution and its SERS enhancement mechanism were studied using the FDTD method; to more realistically simulate the electromagnetic field, the structures of gold nanofilms and silver nanoparticles were constructed based on SEM results. Figure 18 The laser source was set to 785nm, and a 2×2×2nm laser was used. 3 Uniform mesh size ( Figure 19 (A)); The electromagnetic field distribution monitored by the monitors on different planes indicates that there are three electromagnetic field enhancement modes in the designed SERS substrate ( Figure 19 Middle (A) to (D): The first is the strong electromagnetic field existing in the gaps between gold nanofilms (monitor 1, Figure 19 (B)); The second type is a strong electromagnetic field located in the gaps between silver nanoparticles (monitor 3, Figure 19 The third type is the enhancement of the electromagnetic field generated by the plasma coupling between the surface silver nanoparticles and the underlying gold nanofilm (monitor 2, Figure 19(C)); The simulation method of FDTD confirms the multiple plasmon coupling effect, providing a theoretical basis for the generation of high-density hotspots, and can endow the substrate with good SERS enhancement performance for label-free SERS detection.

[0082] (5.3) SERS Performance Factor (SPF) Determination: The SERS performance of Ag@KI / HMGF substrates was studied using different concentrations of CV. Based on the SERS performance factor calculation method developed by Liu et al., SERS performance factors were obtained at 723 cm⁻¹. -1 The SERS and intrinsic Raman spectra at the location were obtained, and the SPF was calculated to be (7.2±0.7)×10⁻⁶ using the slope of the linear fitting curve. 6 This value has a SERS performance factor comparable to that of gold dimer, indicating that the proposed coupled SERS substrate has a good enhancement effect. Specific detection and calculation process: SERS spectra of different concentrations of CV on this substrate are shown in [reference needed]. Figure 20 Middle (A), 723cm -1 The SERS intensity and CV concentration at the site conform to the Langmuir adsorption model ( Figure 20 (B) ), linear fit correlation coefficient R 2 =0.9793 ( Figure 20 (C)); Simultaneously, the intrinsic Raman spectra of different concentrations of CV on a hydrophobic glass slide (HG) were measured and linearly fitted (C). Figure 20 From (D) to (E), the SPF is calculated by the ratio of the slopes of their linear fitting curves. Figure 20 (F)

[0083] (5.4) Anti-coffee ring effect and droplet concentration effect: On the surface of the hydrophobic gold nanofilm, the evaporation process of the mixed solution of silver nanoparticles and analytes is mainly dominated by the dynamic retraction of the contact line and the internal circulation driven by the Marangoni effect, which effectively counteracts the coffee ring effect, thereby guiding the particles to aggregate towards the center of the droplet and forming obvious and uniform deposition spots. Specific characterization: 5 μL of a CV and Ag@KI NPs mixed solution was dropped onto the hydrophobic gold nanofilm ( Figure 3 (A) The concentration process of Ag@KINPs and analyte mixture droplets at 0 min, 20 min, 40 min, and 60 min is shown in [reference]. Figure 3 As shown in (B), the contact area between the droplet and the gold film increased from the initial 1.91 mm. 2 0.16mm when dried 2 This forms a spot with a uniformly distributed composition, approximately 450 μm in diameter. Figure 21 ).

[0084] (5.5) The reproducibility of SERS signals on Ag@KI / HMGF coupled enhanced substrates was studied: CV solution was mixed with silver nanoparticles, deposited, and dried on a hydrophobic gold nanofilm. SERS spectra were then collected from 20 randomly selected points within the completely dried spot, and the characteristic CV bands (723 cm⁻¹) were analyzed. -1 A comparative analysis of the SERS signal strength of the samples showed that RSD = 5.30%. Figure 22 (A) to (B) indicate that this coupled SERS substrate has good signal reproducibility, which will greatly shorten the test time and improve the efficiency of high-throughput serum metabolic SERS detection.

[0085] (5.6) To verify the signal enhancement advantage of the prepared Ag@KI / HMGF, the following comparative experiments were conducted using a hydrophobic gold nanolayer film (HMGF) SERS substrate and a hydrophobic glass sheet combined with silver nanosol (Ag@KI / HG) SERS substrate as controls: (5.6.1) CV detection comparison: 10 tests were performed. -8 The SERS spectra of M on three substrates showed that Ag@KI / HMGF exhibited the highest absolute signal intensity and the most comprehensive spectral fingerprint information. Figure 3 (C) to (D)); CV at 723cm -1 The characteristic peak spectrum at the location also indicates that the detection limit of the Ag@KI / HMGF SERS substrate is lower (10). -12 M)( Figure 23 (A) to (C)), which is due to the high-density hotspots constructed by the Ag@KI / HMGF SERS substrate, thus promising to obtain richer and more sensitive label-free metabolic SERS fingerprints; (5.6.2) Comparison of small molecule metabolism detection in tumors: Five representative small molecules of gastrointestinal tumors (phenylalanine, cholesterol, tryptophan, glutathione, and hypoxanthine) were used as models for detection using the three substrates mentioned above. The results showed that for any small molecule at the same concentration, compared with the HMGF SERS substrate and the Ag@KI / HG SERS substrate, the Ag@KI / HMGF SERS substrate showed more Raman characteristic bands and a better signal-to-noise ratio. Figure 3 (E) to (G)); Similarly, for the above five small molecule mixed solutions, the Ag@KI / HMG FSERS substrate showed characteristic peaks of all five components, and for each component, at least three different Raman bands were shown corresponding to it, while the HMGF SERS substrate and Ag@KI / HG SERS substrate only showed Raman characteristic peaks of specific components and the number of corresponding bands was limited. Figure 3(See Table 1 for H); Therefore, the Ag@KI / HMGF SERS substrate can display the most comprehensive Raman characteristic bands of small molecule mixtures due to its high-density hot spots, making it more suitable for sensitive detection of small metabolic molecules in serum.

[0086] Table 1 shows the Raman spectroscopy detection of five small metabolic molecules (phenylalanine, cholesterol, tryptophan, glutathione, and hypoxanthine) on different substrates (Ag@KI / HMGF, Ag@KI / HG, and HMGF), and the assignment of Raman characteristic peaks.

[0087]

[0088] Example 6 (Detection of serum samples based on Ag@KI / HMGF SERS substrate): (6.1) Serum sample collection: The serum samples used in this embodiment were provided by Wuhan University People's Hospital. The experimental protocol was approved by the Clinical Research Ethics Committee of the hospital (approval number: WDRY2025-K212). A total of 368 human serum samples were collected, including 50 serum samples from healthy volunteers, 101 serum samples from gastric cancer patients, 95 serum samples from liver cancer patients, and 122 serum samples from colorectal cancer patients. All subjects signed informed consent forms.

[0089] (6.2) Serum sample pretreatment: Since serum contains a large number of protein macromolecules, in serum-based detection of low-content metabolic small molecules, the surface of Ag@KI NPs nanoparticles will form a "protein crown" structure, which will hinder the entry of metabolic small molecules into the hot spot area, thus causing the loss of metabolic fingerprint information. Therefore, in order to obtain more comprehensive SERS spectral information of metabolic small molecules, a 3kDa ultrafiltration centrifuge tube was selected to perform simple pretreatment of serum to remove biomacromolecules. Specific operation: 200μL of serum sample was placed in an ultrafiltration centrifuge tube with a 3kDa cutoff, and ultrafiltration centrifuged at 4℃ and 12000g for 50min. The supernatant at the bottom of the ultrafiltration centrifuge tube was collected and stored in an ultra-low temperature freezer at -80℃ for later use. Unfiltered serum was also stored at -80℃ for later use.

[0090] (6.3) SERS detection of serum samples: Take 10 μL of the above serum ultrafiltrate and mix it thoroughly with 10 μL of Ag@KI nanosol prepared in Example 4. Take 5 μL of the mixture and drop it onto the surface of the HMGF substrate prepared in Example 3. Let it air dry at room temperature. Figure 4 (A) SERS spectra of the dried substrate were acquired using a confocal micro Raman spectroscopy system. The detection conditions were: 785 nm laser excitation, 10 mW laser power, integration time 2 s, and a spectral detection range of 400-1800 cm⁻¹. -¹, ×50 objective lens, at least 20 SERS spectra were acquired for each sample; before detection, a 520cm² silicon wafer was used. - ¹Use Raman spectral peaks for instrument calibration; The collected SERS spectra were subjected to background correction using LabSpec 6 software, and then normalized using Origin software to obtain the average SERS spectra of the four groups of serum samples. Figure 4 The SERS characteristic peak assignments of serum analytes are shown in Table 2 (Assignment of SERS characteristic peaks of serum analytes). Table 2

[0091] Specifically, the difference spectra and characteristic peak intensity histograms of the four groups of samples are shown below. Figure 4 Middle (C) to (D): 482cm -1 616cm -1 852cm -1 1046cm -1 1225cm -1 and 1636cm -1 The Raman characteristic peak intensities at these locations show significant differences, which can serve as key sites for differential analysis of metabolic fingerprints.

[0092] After comparing the four sets of average spectra, it was found that 482 cm⁻¹ -1 616cm -1 852cm -1 1046cm -1 1225cm -1 and 1636cm -1 The changes in the intensity of the Raman characteristic peaks are due to the differences in SERS signals of metabolites in the serum of different groups. Therefore, analyzing the SERS characteristic peaks is expected to elucidate the differences in metabolic fingerprints among different cancer groups; the samples from each group ranged from 400 to 1800 cm⁻¹. -1 The spectra within the range differed. To identify the differences in key characteristic peaks of SERS between different groups, the Kruskal–Wallis test was used with Bonferroni correction. Screening was performed using P < 0.05; results showed that the serum SERS band in the Normal group was at 852 cm⁻¹. -1 (Tryptophan and purine / pyrimidine bases) and 1046 cm -1The normalized peak intensity at (L-serine, threonine, creatine, and adenine) was higher than that in the LC, CRC, and GC groups. This may be because the tumor overexpresses IDO1 / TDO, consuming tryptophan; and activates β-catenin, accelerating the consumption of amino acids such as tyrosine and purines / pyrimidines, leading to a decrease in the serum levels of these metabolites; while at 895 cm⁻¹... -1 (β-D-galactose and mannose) and 1675cm -1 The peak intensities (cholesterol, cholesterol esters, LDL-phospholipids and free unsaturated fatty acids) in the normal group were lower than those in the cancer group. This is a typical metabolic reprogramming feature of tumor cells, providing energy, raw materials for biomembrane synthesis and precursors for signaling molecules for rapid proliferation.

[0093] Traditional Algorithm Classification Attempts: To classify the above-mentioned Normal, LC, CRC, and GC serum samples, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) were used to analyze the SERS data respectively; the results showed that three-dimensional PCA analysis was difficult to classify the above four types of samples. Figure 4 (E)), while three-dimensional LDA can only distinguish between healthy people and cancer patients, and it is difficult to completely distinguish between different cancers (E). Figure 4 (F)); This indicates that the traditional linear dimensionality reduction and discrimination algorithm has limited discrimination efficiency for multi-class classification tasks of four types of serum SERS data. Therefore, it is urgent to introduce a nonlinear ensemble learning strategy to capture the high-order interaction relationship between spectral features and improve the anti-interference performance of the model, thereby achieving accurate discrimination of serum SERS spectra in various gastrointestinal malignancies.

[0094] Example 7 (Identification of Gastrointestinal Malignant Tumors from Serum Samples Based on Machine Learning): (7.1) Data preparation and model building: Machine learning can extract deep features from the high-dimensional and complex data of serum metabolic SERS fingerprints, thus effectively addressing the challenge of difficult discrimination of serum metabolic SERS signals in different tumors; a high-throughput analysis platform based on Ag@KI / HMGF SERS was used to obtain serum metabolite SERS spectra of the Normal group, LC, CRC and GC populations, and the spectral data were processed by four machine learning models (CatBoost (CB), ExtraTrees (ET), LightGBM (LGBM), RandomForest (RF)) to achieve screening of LC, CRC and GC cancers in a large population; Data partitioning and labeling: 80% of the Normal group, LC, CRC and GC data were randomly selected as the training set and 20% as the test set for identification analysis; the metabolic SERS spectra of serum of the Normal group, LC, CRC and GC groups were labeled as 0, 1, 2 and 3 respectively for four-level classification.

[0095] (7.2) Construction and Training of Machine Learning Models: Four machine learning models were constructed using Python: CatBoost (CB), ExtraTrees (ET), LightGBM (LGBM), and RandomForest (RF). The models were trained on the training set data, and the changes in training loss, validation loss, and accuracy with the number of iterations were monitored until the models converged and stabilized. Figure 5 As shown in (A); the loss curve and accuracy curve of the LGBM model are shown in [reference needed]. Figure 5 The loss curves and accuracy curves of the ET, RF, and CB models are shown in (B) to (C). Figure 24 (A) to (C) Figure 25 (A) to (C); the results show that the training and validation losses of the four models (CB, ET, RF, and LGBM) decrease with iteration, while the accuracy increases and tends to stabilize. Figure 5 From (B) to (C), Figure 24 (A) to (C) Figure 25 The values ​​(A) to (C) indicate that these models fit well, do not exhibit overfitting, and have good fitting performance.

[0096] (7.3) Model performance evaluation: The model performance is comprehensively evaluated using metrics such as confusion matrix, ROC curve and area under the curve (AUC), sensitivity, specificity, and accuracy. (7.3.1) Confusion matrix analysis: The LGBM model showed the best discrimination effect among the four groups of samples. Figure 5 (D) The confusion matrices of the training sets for the four models are shown in the table. Figure 26 The confusion matrices for the test sets of ET, RF, and CB models (A through D) are shown below. Figure 27 The confusion matrices of the training and test sets (A to C) show that, compared to the CB, ET, and RF models, the LGBM model achieves the best differentiation of serum metabolic SERS profiles for the Normal group, LC, CRC, and GC groups. Figure 5 (D) Figure 26 (A) to (D) Figure 27 (A) to (C)); (7.3.2) ROC curve and AUC value analysis: The LGBM model had the best macro-averaging AUC on the test set. The AUC values ​​for the healthy group, liver cancer group, colorectal cancer group, and gastric cancer group were 0.998, 0.975, 0.988, and 0.967, respectively. Figure 5 (E) , ROC curves of the training sets for the four models are shown in the figure. Figure 28The ROC curves for the ET, RF, and CB models on the test set are shown in (A) to (D). Figure 29 Figures (A) to (C) demonstrate that the model has good predictive ability for the above four types of tumors; (7.3.3) Accuracy, sensitivity, and specificity: The four machine learning models (ET, RF, CB, LGBM) all achieved good differentiation of the four groups of serum SERS metabolic fingerprints, with the LGBM model showing the best differentiation performance, achieving an accuracy of 98.6% on the training set and 90.4% on the test set. Figure 5 The sensitivity and specificity of the four models for detecting four groups of samples are summarized in Table 3 (Sensitivity and specificity of four machine learning models (A) ExtraTrees, (B) RandomForest, (C) CatBoost, (D) LightGBM analysis of (NG, LC, CRC, GC)). The LGBM model showed the best performance in all indicators. The sensitivity of the healthy group, liver cancer group, colorectal cancer group and gastric cancer group were 0.855, 0.911, 0.954 and 0.903, respectively, and the specificity was 1.000, 0.916, 0.998 and 0.944, respectively.

[0097] Table 3

[0098] (7.4) Key Feature Band Analysis: Based on the LGBM model, the contribution of different Raman feature bands to the prediction results is obtained. The SHAP values ​​of the top 10 important frequency bands are then used to analyze the results. Figure 5 In the middle (F), the features that contribute the most to the overall decision of the model can be quickly identified, and these SERS bands are assigned accordingly, as shown in Table 4 (Assignment of the top ten important SERS bands of serum analytes); Table 4

[0099] (7.4.1) Mechanism of characteristic differences: Due to the elevated levels of ketone bodies and ceramides in the serum of CRC patients, the elevated levels of sulfur amino acids and phospholipids in the serum of LC patients, and the elevated levels of purines and glycolytic intermediates in the serum of GC patients, the serum metabolic SERS fingerprints are different; specifically, CRC patients have elevated levels of β-hydroxybutyrate (1051-1054 cm⁻¹) in their serum. -1 ), Riboflavin (1592-1595cm) -1 ) and ceramides (1590-1594cm) -1 The levels of sulfur-containing amino acids (cysteine: 510-515 cm⁻¹) in serum were significantly elevated; -1 Methionine: 635-638 cmcm -1) and phospholipids (1050-1055cm) -1 Elevated levels of GC in serum lactate (1043-1047 cm⁻¹) -1 ), glutamic acid (1052-1055cm) -1 1585-1587cm -1 ), glycine (1044-1046cm) -1 ) and guanine (636-639cm) -1 1583-1587cm -1 () significantly increased.

[0100] Therefore, all four machine learning models can effectively distinguish between healthy individuals and patients with gastrointestinal tumors. Among them, the LGBM model has the best identification effect, which proves that the method based on Ag@KI / HMGF SERS combined with machine learning has good accuracy and reliability in the screening of gastrointestinal malignancies and has high clinical application potential.

[0101] In summary, this invention, based on the principle of triple-coupling hotspot enhancement, innovatively designs and constructs a clean SERS detection platform, aiming to obtain more comprehensive metabolic SERS fingerprint spectra and achieve accurate detection of gastrointestinal malignancies. Specifically, this invention uses KI to treat the surface of silver nanoparticles (Ag NPs) and hydrophobic monolayer gold film (HMGF) substrates, which can efficiently remove the background Raman signal on the SERS substrate surface, thereby avoiding interference with the metabolite SERS fingerprint signal and ensuring detection accuracy. Furthermore, by using PFDT to hydrophobically modify the HMGF substrate surface, the substrate acquires the self-concentration and enrichment characteristics of analyte droplets, providing a structural basis for rapid, sensitive, and high-throughput acquisition of metabolite SERS spectra.

[0102] The core advantage of the SERS detection platform lies in its triple plasmon coupling effect: enhanced electromagnetic field within the monolayer gold nanofilm, enhanced electromagnetic field within the aggregates of potassium iodide-modified silver nanoparticles (Ag@KI NPs), and enhanced vertical electromagnetic field coupling between the monolayer gold nanofilm and Ag@KI NPs. This triple coupling effect significantly improves the SERS hotspot density, giving the Ag@KI / HMGF coupled SERS substrate excellent SERS enhancement performance, with a SERS performance factor (SPF) as high as 7.9 × 10⁻⁶. 6 .

[0103] Compared with traditional single-coupled solid-phase SERS substrates and homogeneous nanosol particles, the Ag@KI / HMGF SERS detection system provided by this invention can capture richer Raman characteristic bands of serum metabolic molecules, providing highly specific metabolic molecular fingerprint information and providing key support for the accurate identification of metabolites in gastrointestinal tumors.

[0104] Furthermore, combining the serum metabolic SERS spectrum obtained by the method of this invention with the LightGBM (LGBM) machine learning algorithm demonstrated excellent performance in the detection of gastrointestinal malignancies: the test set accuracy reached 90.4%, the test set Macro-AUC value reached 98.4%, and it also exhibited good sensitivity and specificity. These results indicate that the SERS detection system proposed in this invention can serve as a candidate point-of-care testing (POCT) technology for the differential diagnosis of gastrointestinal malignancies in clinical settings, possessing broad clinical application prospects.

[0105] Furthermore, the coupled SERS substrate prepared by this invention can be further extended to the detection of other types of cancer or various body fluid samples (such as plasma, urine, cerebrospinal fluid, etc.), providing core support for the construction of efficient and reliable label-free multiplex SERS liquid biopsy technology, thereby realizing early screening and accurate diagnosis of tumors, and has important technical promotion value.

[0106] It should be noted that all the above embodiments belong to the same inventive concept, and the descriptions of each embodiment have different focuses. Where the description in a particular embodiment is not detailed, please refer to the description in other embodiments.

[0107] The above embodiments merely illustrate implementation methods of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for fabricating a triple-coupled SERS substrate, characterized in that, Includes the following steps: S10, the substrate is sequentially purified and surface functionalized, and the modified substrate is immersed in nano gold seed sol to obtain a functionalized substrate. S20, the functionalized substrate is placed in a growth reaction system containing gold ions, a first reducing agent is added, and the growth reaction is carried out by constant temperature shaking at room temperature. After the reaction is completed, the substrate is washed and dried. The above growth, washing and drying operations are repeated multiple times to obtain a gold film SERS substrate. S30, the gold film SERS substrate is purified using an iodine ion source reagent, and then the surface of the purified gold film SERS substrate is modified using a hydrophobic thiol reagent. After drying, a hydrophobic modified gold film SERS substrate is obtained. S40, Ag@KI nanosol is deposited on the surface of the hydrophobically modified gold film SERS substrate to obtain a triple-coupled SERS substrate.

2. The method for fabricating a triple-coupled SERS substrate according to claim 1, characterized in that, In step S10, the substrate is first purified and dried using aqua regia solution, and then its surface is functionalized using an amino-based silane coupling agent. The substrate is a glass slide, and the substrate is immersed in the gold nanoparticle sol for 8 to 12 hours.

3. The method for fabricating a triple-coupled SERS substrate according to claim 1, characterized in that, The gold nanoparticle sol in step S10 is prepared by liquid-phase reduction. The preparation process is as follows: 400-600 μL of a first soluble gold salt solution, 1-2 mL of a coordination stabilizer, and 400-600 μL of a second reducing agent are added sequentially to 40-60 mL of deionized water. After in-situ reduction reaction for 1-3 hours, the gold nanoparticle sol is obtained.

4. The method for preparing a triple-coupled SERS substrate according to claim 3, characterized in that, The first soluble gold salt solution is a 1 wt% chloroauric acid solution, the coordination reagent is a 10-15 mg / mL trisodium citrate solution, the second reducing reagent is a 0.5-1 mg / mL sodium borohydride solution, and the particle size of the gold nanoparticle sol is 3-6 nm.

5. The method for fabricating a triple-coupled SERS substrate according to claim 1, characterized in that, The S20 step specifically includes: S201, the functionalized substrate is placed in a reaction vessel, and 1-3 mL of deionized water, 300-400 μL of the second soluble gold salt solution and 100-200 μL of the first reducing reagent are added in sequence. The reaction is carried out at room temperature with constant temperature shaking at 200-300 r / min for 10-30 min. After removal, it is washed and dried with nitrogen. S202, add 1-3 mL of deionized water, 50-100 μL of the second soluble gold salt solution and 30-50 μL of the first reducing agent, and complete the shaking reaction, washing and drying operations under the same conditions. Repeat the above growth process 3-5 times to obtain the gold film SERS substrate.

6. The method for fabricating a triple-coupled SERS substrate according to claim 5, characterized in that, The reaction vessel is a polypropylene centrifuge tube, the second soluble gold salt solution is a 0.2 wt% chloroauric acid solution, and the first reducing reagent is a 0.02–0.06 mol / L hydroxylamine hydrochloride solution.

7. The method for fabricating a triple-coupled SERS substrate according to claim 1, characterized in that, The S30 step specifically includes: S301, immerse the gold film SERS substrate in 1-3 mL of KI solution with a concentration of 10 -4 -10 -2 mol / L, remove after standing for 10-30 min, dry with nitrogen, repeat the above operation twice to obtain the purified SERS substrate; S302, the purified gold film SERS substrate is then immersed in 1-3 mL of 3 wt% PFDT ethanol solution, left to stand for 4-8 hours, and then removed and dried to obtain the hydrophobic modified gold film SERS substrate.

8. The method for fabricating a triple-coupled SERS substrate according to claim 1, characterized in that, The preparation process of the Ag@KI nanosol in step S40 is as follows: The aqueous solution of silver precursor was heated to boiling, and a water-soluble organic acid coordination reducing agent was added under stirring. The reaction was carried out at a constant temperature until complete, and silver nanoparticle sol coated with organic ligands was obtained. The silver nanoparticle sol was subjected to solid-liquid separation, and the supernatant was discarded to prepare a dispersion. The silver nanoparticles in the dispersion were subjected to surface ligand replacement using an iodine ion source reagent, followed by solid-liquid separation and discarding of the supernatant. After redispersement, the Ag@KI nanosol was obtained.

9. A triple-coupled SERS substrate, characterized in that, It is prepared by the preparation method of the triple-coupled SERS substrate as described in any one of claims 1 to 8.

10. An application of the triple-coupled SERS substrate as described in claim 9, characterized in that, The triple-coupled SERS substrate is used for the detection of gastrointestinal tumor metabolites in human serum.

Images