Flower-shaped cuprous oxide-based SERS (Surface Enhanced Raman Scattering) biosensor and method for establishing tumor cell / leukocyte classification model based on same

Abstract

The invention belongs to the technical field of biological detection, and relates to a flower-shaped cuprous oxide-based SERS (Surface Enhanced Raman Scattering) biosensor and a method for establishing a tumor cell / leukocyte classification model based on the same. The flower-shaped Cu2O is successfully prepared by regulating and controlling the characteristic parameters, the raw material ratio and the like of polyvinylpyrrolidone in the preparation process, and the cuprous oxide-based SERS biosensor prepared from the flower-shaped Cu2O and precious metal has excellent SERS stability and sensitivity. The cuprous oxide-based SERS biosensor provided by the invention can be combined with tumor cells and leukocytes without any marker to obtain Raman spectrum data of different cells, a tumor cell / leukocyte classification model is established through data processing, and marker-free identification and judgment of tumor cells / leukocytes are realized.

Description

Technical Field

[0001] The invention belongs to the technical field of biological detection, and relates to a flower-shaped cuprous oxide-based SERS biosensor and a method for establishing a tumor cell / leukocyte classification model based on the flower-shaped cuprous oxide-based SERS biosensor. Background Art

[0002] Classical histopathological testing relies on obtaining tissue sections and complex immunohistochemical staining to determine the tissue type, grade, and stage of the tumor. In recent years, an emerging non-invasive liquid biopsy can detect and grade cancer by collecting blood samples, which has greatly alleviated the pain of patients and made long-term monitoring of cancer possible. However, due to the complex composition of blood samples, there is a lot of interference from blood cells (red blood cells and white blood cells) that are similar in morphology to tumor cells. Despite significant differences in size, these large numbers of blood cells (approximately 5 billion red blood cells and 4,000-11,000 white blood cells per milliliter) still make it extremely difficult to obtain and detect cancer cells. Therefore, it is necessary to develop efficient separation and enrichment technologies to further accurately identify cancer cells. Microfluidics is a highly integrated sample processing platform that combines physics, mechanical manufacturing, and biology. By controlling the flow of particles in customized microchannels, the separation and enrichment of different components in the sample can be achieved. Inertial microfluidic chips use the inertial effect of fluids to focus and separate particles or cells of different sizes. Due to its rapid separation and high-throughput processing capabilities, no external energy intervention and cost-effectiveness, it has become a powerful tool for biological sample processing. The spiral inertial microfluidic chip consists of a spiral channel and a liquid inlet, which can achieve the removal of red blood cells (6-8μm), white blood cells (10-15μm) and the enrichment of cancer cells (12-25μm) within a few minutes. However, since the size of white blood cells overlaps with that of cancer cells, the cancer cells collected by the microfluidic chip are often mixed with some interference from white blood cells. Therefore, it is very necessary to develop accurate recognition technology to distinguish cancer cells.

[0003] Non-invasive and highly sensitive surface enhanced Raman spectroscopy (SERS) is a powerful tool for material analysis. This technology can reveal the fingerprint information of biomarkers at the cellular and molecular levels. Traditional SERS substrates mainly include precious metals and semiconductors, but the signal uniformity of precious metal substrates is difficult to ensure due to the existence of the "hot spot" effect, and the signal enhancement ability of semiconductor substrates is limited. Therefore, exploring a composite SERS substrate with high sensitivity, stability, selectivity and reliability has become an urgent problem to be solved. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the purpose of the present invention is to provide a flower-shaped cuprous oxide-based SERS biosensor and a method for establishing a tumor cell / leukocyte classification model based thereon, so as to overcome the shortcomings of the prior art.

[0005] One object of the present invention is achieved by the following technical solutions:

[0006] A method for preparing a flower-shaped cuprous oxide-based SERS biosensor comprises the following steps:

[0007] S1. Add polyvinyl pyrrolidone (PVP) to the copper salt solution, then add sodium hydroxide solution, stir for 0.5 to 3 hours, then add the first reducing agent solution, heat for reaction, and obtain flower-shaped Cu2O after centrifugation, washing and drying; the appearance of the polyvinyl pyrrolidone is white crystalline powder, the molecular weight is 1200000 to 1400000, and the density is 1.6 to 1.7 g / cm 3 , the mass ratio of polyvinyl pyrrolidone to copper salt is 30-40:1;

[0008] S2, preparing the flower-shaped Cu2O obtained in step S1 into a suspension, adding a second reducing agent solution, stirring, and then adding a noble metal salt solution, reacting, and obtaining a flower-shaped Cu2O@noble metal suspension after centrifugation, washing, and resuspending;

[0009] S3. Drop the flower-shaped Cu2O@precious metal suspension on a clean silicon wafer and obtain a flower-shaped cuprous oxide-based SERS biosensor after drying.

[0010] Preferably, the polyvinyl pyrrolidone is purchased from MacLean Reagent Company, and its appearance is white crystalline powder, the molecular weight is 1250000-1350000, and the density is 1.69g / cm 3 ; The mass ratio of polyvinyl pyrrolidone to copper salt is 33-37:1.

[0011] The characteristic parameters of PVP and the mass ratio of PVP to copper salt play a very critical role in the formation of flower-shaped Cu2O and determine whether flower-shaped Cu2O can be synthesized. The present invention adjusts the characteristic parameters of PVP to: white crystalline powder, molecular weight of 1250000-1350000, density of 1.69g / cm 3 , purchased from McLean Reagent Company, and the mass ratio of PVP to copper salt is 33-37:1, and flower-like Cu2O can be obtained.

[0012] Preferably, the copper salt is one or more of copper chloride, copper nitrate, copper sulfate and copper acetate.

[0013] Preferably, the first reducing agent is one or more of ascorbic acid, glucose and hydroxylamine.

[0014] Preferably, the molar ratio of copper salt to sodium hydroxide is 1:10-50.

[0015] Preferably, the molar ratio of the copper salt to the first reducing agent is 1:10-50.

[0016] Preferably, the second reducing agent is one or more of sodium citrate, NaBH4, hydrazine hydrate, and ascorbic acid.

[0017] Preferably, the noble metal salt is silver chloride or chloroauric acid.

[0018] Preferably, the particle size of the flower-shaped Cu2O is 0.3-3 μm.

[0019] Preferably, the mass ratio of flower-shaped Cu2O to precious metal salt is 1 to 5:1.

[0020] Preferably, the molar ratio of the noble metal salt to the second reducing agent is 1:4-50.

[0021] Preferably, the heating reaction temperature of step (1) is 45 to 65° C. and the time is 2 to 6 hours.

[0022] Preferably, the reaction in step (2) is carried out at 10-45° C. for 0.5-5 h.

[0023] The second object of the present invention is achieved by the following technical solutions:

[0024] A flower-shaped cuprous oxide-based SERS biosensor is prepared by the above preparation method.

[0025] The third object of the present invention is achieved by the following technical solutions:

[0026] A method for establishing a tumor cell / leukocyte classification model based on a flower-shaped cuprous oxide-based SERS biosensor, the method comprising the following steps:

[0027] (1) Cultivating tumor cells;

[0028] (2) adding tumor cells to diluted anticoagulated rabbit blood, injecting into the microfluidic chip and collecting the cell separation fluid at the outlet;

[0029] (3) dropping the cell separation solution on the flower-shaped cuprous oxide-based SERS biosensor, and collecting Raman spectral data of tumor cells and leukocytes on the sensor after spreading;

[0030] (4) The Raman spectral data were subjected to machine learning dimensionality reduction and feature extraction based on the linear discriminant analysis (LDA) algorithm to establish a tumor cell / leukocyte binary classification model, and the performance of the model was evaluated using confusion matrix and / or ROC curve.

[0031] The cell separation solution in step (2) includes tumor cells and leukocytes.

[0032] Preferably, the cuprous oxide of Cu2O@precious metal in the cuprous oxide-based SERS biosensor is flower-shaped, and the precious metal is silver. Flower-shaped Cu2O@Ag has good SERS sensitivity and better spectral stability, which is of great significance for improving the accuracy of tumor cell / leukocyte classification models.

[0033] Preferably, the tumor cells are one or more of bladder cancer T24 cells, brain glioma U251 cells, and pancreatic cancer Panc02 cells.

[0034] Preferably, in step S3, cells with a diameter of ≤15 μm under a microscope are regarded as white blood cells, and cells with a diameter of ≥20 μm are regarded as tumor cells, so as to obtain Raman spectrum data of the corresponding cells.

[0035] Preferably, the Raman spectral data is in the spectral range of 600 to 1800 cm -1 SERS spectral data within.

[0036] The machine learning-assisted label-free SERS detection classification model uses various machine learning algorithms to build a classification model that includes SERS spectral data dimensionality reduction, feature extraction, training and testing. The model uses machine learning algorithms to analyze and distinguish the SERS spectral data of different types of cells, thereby achieving accurate distinction between tumor cells and white blood cells.

[0037] Preferably, the tumor cell / leukocyte binary classification model established by the linear discriminant analysis (LDA) algorithm can be used to accurately distinguish tumor cells from leukocytes with an accuracy of 90% or more.

[0038] Furthermore, the area under the ROC curve (AUC) was used to evaluate the performance of the binary classification model. The AUC value ranged from 0 to 1. The closer the AUC value was to 1, the better the classification performance of the tumor cell / leukocyte binary classification model.

[0039] When the label-free SERS detection data of bladder cancer T24 tumor cells and leukocytes were selected to establish a tumor cell / leukocyte binary classification model, the area under the ROC curve (AUC) of the classification model was 0.97.

[0040] When the label-free SERS detection data of brain glioma U251 cells and leukocytes were selected to establish the tumor cell / leukocyte binary classification model, the area under the ROC curve (AUC) of the classification model was 1.00.

[0041] When the label-free SERS detection data of pancreatic cancer Panc02 cells and leukocytes were selected to establish a tumor cell / leukocyte binary classification model, the area under the ROC curve (AUC) of the classification model was also 1.00.

[0042] The fourth object of the present invention is achieved by the following technical solutions:

[0043] A tumor cell / leukocyte binary classification model is established by the above method.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. The present invention successfully prepares a flower-like Cu2O by regulating the characteristic parameters of polyvinyl pyrrolidone, the ratio of raw materials, etc. during the preparation process.

[0046] 2. After the flower-shaped Cu2O@precious metal is formed, the optimal absorption peak is near 532nm, which makes the cuprous oxide-based SERS material have a good coupling effect with the Raman excitation light (excitation wavelength of 532nm), thereby enhancing the localized surface plasmon resonance effect of Cu2O@precious metal and promoting the improvement of the SERS performance of the material, especially improving the SERS stability of the flower-shaped Cu2O@precious metal, which is of great significance for improving the accuracy of the tumor cell / leukocyte classification model.

[0047] 3. The present invention first prepares cuprous oxide-based SERS biosensors and then combines microfluidics and machine learning to achieve accurate differentiation of tumor cells and leukocytes.

[0048] 4. The cuprous oxide-based SERS biosensor provided by the present invention can combine with tumor cells and leukocytes without any labeling to obtain Raman spectral data of different types of cells. After data processing, a tumor cell / leukocyte classification model is established to achieve label-free identification and determination of tumor cells / leukocytes.

[0049] 5. The use of cuprous oxide-based biosensors to establish a tumor cell / leukocyte classification model can effectively identify tumor cells, facilitate the establishment of an in vitro tumor cell model, and has great potential in the early screening and prognosis assessment of cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a scanning electron microscope image of the flower-shaped Cu2O particles prepared in step S1 of Example 1;

[0051] Figure 2 This is a scanning electron microscope image of the flower-shaped Cu2O@Ag particles prepared in step S2 of Example 1;

[0052] Figure 3The UV-visible absorption spectra of the flower-shaped Cu2O and Cu2O@Ag particles of Example 1;

[0053] Figure 4 The SERS detection performance of the flower-shaped Cu2O@Ag in Example 1 for different signal molecules, where A is the detection limit of the flower-shaped Cu2O@Ag-AR (Alizarin Red), BC is the detection limit of the flower-shaped Cu2O@Ag-AR (10 -5 M) 20-point intensity distribution heat map and spectrum; D: detection limit of flower-like Cu2O@Ag-CV (crystal violet), EF: flower-like Cu2O@Ag-CV (10 -4 M) 20-point intensity distribution heat map and spectrum; G: Detection limit of flower-like Cu2O@Ag-MB (methylene blue), HI: Flower-like Cu2O@Ag-MB (10 -4 M) 20-point intensity distribution heat map and spectrum;

[0054] Figure 5 The confusion matrix of linear discriminant analysis of three tumors and leukocytes by flower-shaped Cu2O@Ag in Example 4;

[0055] Figure 6 The ROC curves of the linear discriminant analysis of the flower-shaped Cu2O@Ag on three tumors and leukocytes in Example 4 are shown. DETAILED DESCRIPTION

[0056] The technical scheme of the present invention is further described below by specific examples and accompanying drawings. It should be understood that the specific examples described herein are only used to help understand the present invention and are not intended for specific limitations of the present invention. The accompanying drawings used herein are only for better illustrating the disclosure of the present invention and do not have a limiting effect on the scope of protection. If not otherwise specified, the raw materials used in the embodiments of the present invention are all commonly used raw materials in the art, and the methods used in the embodiments are all conventional methods in the art.

[0057] In the following examples, the polyvinyl pyrrolidone used was purchased from MacLean Reagent Company, which has the appearance of white crystalline powder, a molecular weight of 1250000-1350000, and a density of 1.69 g / cm 3 .

[0058] Example 1

[0059] The flower-shaped cuprous oxide-based SERS biosensor of this embodiment is prepared by the following steps:

[0060] S1. Weigh 3 g of polyvinyl pyrrolidone and dissolve it in 100 mL of CuCl2·2H2O (0.005 mol / L) aqueous solution; then, slowly add 10 mL of NaOH aqueous solution (2.0 mol / L) into the copper salt solution, and the color of the reaction solution changes from light blue to light black; after magnetic stirring at 500 rpm for 1 hour, further add ascorbic acid aqueous solution (10 mL, 2.0 mol / L) dropwise into the mixed solution, and heat it in a 50°C water bath for 3 h; collect the obtained precipitate by centrifugation, and repeatedly wash it with deionized water and anhydrous ethanol to remove residual polymer and inorganic ions; dry the obtained ethanol suspension in an oven at 50°C for 3 h to obtain flower-like Cu2O.

[0061] S2. Take 4.5 mg of flower-like Cu2O powder and ultrasonically disperse it in 3 mL of anhydrous ethanol to obtain a Cu2O suspension. Under vigorous stirring, the above suspension is dissolved in a mixture of ethanol / deionized water (20 mL / 80 mL); then sodium citrate aqueous solution (1.2 mL, 30 mM) and freshly prepared NaBH4 aqueous solution (1.5 mL, 100 mM) are sequentially dropped into the solution containing Cu2O; after stirring at 500 rpm for 5 minutes, AgNO3 solution (1.5 mL, 10 mM) is added dropwise to the stirred mixture, and the color of the solution turns light yellow, indicating that small Ag nanoparticles are formed; after reacting at room temperature (25°C) for 2 hours, the reaction solution is centrifuged and washed with deionized water and ethanol; finally, the collected particles (Cu2O@Ag) are suspended in ethanol to obtain a flower-like Cu2O@Ag ethanol suspension for further use and characterization.

[0062] S3. Drop the flower-shaped Cu2O@Ag ethanol suspension on a clean silicon wafer (10×10 mm) and obtain the flower-shaped cuprous oxide-based SERS biosensor after drying.

[0063] Figure 1 This is a scanning electron microscope image of the Cu2O prepared in step S1. It can be seen that the morphology of the obtained Cu2O is flower-like.

[0064] Figure 2 This is a scanning electron microscope image of the Cu2O@Ag particles prepared in step S2. It can be seen that the silver particles are coated on the outer layer of the flower-shaped Cu2O particles.

[0065] Example 2: UV-Vis absorption spectrometry

[0066] The flower-like Cu2O prepared in step S1 of Example 1 was dispersed in anhydrous ethanol, and 1 mL of the flower-like Cu2O ethanol suspension and the flower-like Cu2O@Ag ethanol suspension prepared in step S2 were respectively placed in a cuvette, and anhydrous ethanol was used as a reference solution to measure the light absorption properties of the flower-like Cu2O and the flower-like Cu2O@Ag in the range of 300-850 nm. Figure 3 Figure 3 is the UV-visible absorption spectra of flower-like Cu2O and flower-like Cu2O@Ag. It can be seen from the figure that flower-like Cu2O has certain absorption performance in the visible light region, and the absorption peaks are near 500nm and 600nm. After the flower-like Cu2O@Ag is formed, the optimal absorption peak is successfully adjusted to near 532nm. This change enables the cuprous oxide-based SERS material to have a good coupling effect with the Raman excitation light (excitation wavelength of 532nm), thereby enhancing the localized surface plasmon resonance effect of Cu2O@Ag and promoting the improvement of the SERS performance of the material, especially improving the SERS stability of the flower-like Cu2O@Ag.

[0067] Example 3: SERS detection performance of flower-shaped Cu2O@Ag for various Raman signal molecules

[0068] In order to illustrate the SERS performance of Cu2O@Ag, the Raman signal molecules Alizarin Red (AR), Crystal Violet (CV) and Methylene Blue (MB) powders were dissolved in ethanol and diluted at different concentrations. The resulting solution was immersed in the Cu2O@Ag ethanol suspension of Example 1 for 8 hours in the dark to form Cu2O@Ag-AR / CV / MB. After that, the mixture was centrifuged to remove excess signal molecules. Then, after drying at room temperature, the resulting suspension was dropped onto a clean silicon wafer with a micropipette for Raman detection. The SERS spectrum was collected by a confocal micro-Raman spectrometer under 532nm excitation light. Figure 4 As can be seen from (A), the LOD value of flower-shaped Cu2O@Ag as SERS substrate for detecting AR reaches 10 -9 In order to verify the repeatability of SERS spectra, the Cu2O@Ag-AR was tested at 10 -5 mol / L Raman spectrum, from Figure 4 It can be seen from (B) and (C) that the 20 spectra shown have characteristic peaks at 1256 and 1440 cm -1 The relative standard deviation of the intensity distribution is 6.13% and 5.81%, which is better than the stability of hexahedral Cu2O@Ag (hexahedral Cu2O@Ag in Example 2 of CN116930143A) in detecting AR (1256 and 1440 cm -1 The relative standard deviations of the strength at the two locations were 15.33% and 14.68% respectively, which was significantly improved. Figure 4(D) and (G) show the SERS spectra of Cu2O@Ag detecting CV and MB molecules with different concentrations. It can be seen that the detection limit of Cu2O@Ag for CV and MB molecules is as low as 10 -8 mol / L, and showed satisfactory spectral stability (e.g. Figure 4 (F), (I)). The above results show that Cu2O@Ag has good SERS stability and sensitivity.

[0069] Example 4: Establishment of a tumor cell / leukocyte typing model

[0070] Before establishing the tumor cell / leukocyte typing model, the tumor cell and leukocyte suspension were distinguished from each other to accurately collect the SERS data of tumor cells and leukocytes: bladder cancer T24 cells, brain glioma U251 cells, and pancreatic cancer Panc02 cells were cultured according to standard cell culture conditions; fluorescently labeled bladder cancer T24 cells, fluorescently labeled brain glioma U251 cells, and fluorescently labeled pancreatic cancer Panc02 cells were added to diluted anticoagulated rabbit blood (5 ml of blood was collected from the ear vein of healthy rabbits, and 0. 5ml, prepare fresh anticoagulated rabbit blood for standby use; take 100μL of fresh anticoagulated rabbit blood, add 0.01M phosphate buffer (PBS) 9.9ml dilution), inject into the microfluidic chip respectively and collect the cell separation liquid at the outlet, collect the suspension including 200 leukocytes and 200 fluorescently labeled T24 cells, the suspension including 200 leukocytes and 200 fluorescently labeled U251 cells, and the suspension including 200 leukocytes and 200 fluorescently labeled Panc02 cells, and observe, mark, measure and record the collected suspension using a fluorescence microscope. In the suspension including 200 leukocytes and 200 fluorescently labeled T24 cells, the results showed that the average diameter of the fluorescently labeled T24 tumor cells was 22.22μm, while the average diameter of the leukocytes was 15.8μm. In a suspension of 200 leukocytes and 200 fluorescently labeled U251 cells, the average diameter of U251 cells was 22.79 μm, and the average diameter of leukocytes was 16.2 μm. In a suspension of 200 leukocytes and 200 fluorescently labeled Panc02 cells, the average diameter of Panc02 cells was 20.92 μm, and the average diameter of leukocytes was 15.4 μm. After analysis, it was found that the size distribution of leukocytes and tumor cells conformed to normality and the sizes of the two cells were significantly different. Therefore, in the subsequent label-free SERS detection, cells with a diameter of ≤15 μm under the microscope were regarded as leukocytes, while cells with a diameter of ≥20 μm were regarded as tumor cells, so as to obtain label-free SERS detection data of the corresponding cells.

[0071] The steps to establish the tumor cell / leukocyte classification model are:

[0072] Bladder cancer T24 cells, brain glioma U251 cells, and pancreatic cancer Panc02 cells were cultured according to standard cell culture conditions;

[0073] Bladder cancer T24 cells, brain glioma U251 cells, and pancreatic cancer Panc02 cells were added to diluted anticoagulated rabbit blood, respectively, injected into the microfluidic chip, and the cell separation fluid at the outlet was collected to form a mixed sample of T24 cells and leukocytes, a mixed sample of U251 cells and leukocytes, and a mixed sample of Panc02 cells and leukocytes.

[0074] The mixed samples of T24 cells and leukocytes, the mixed samples of U251 cells and leukocytes, and the mixed samples of Panc02 cells and leukocytes were respectively coated on the silicon wafer surface in the flower-shaped cuprous oxide-based SERS biosensor of Example 1 and detected by confocal micro-Raman spectrometer with 532 nm laser as the excitation source.

[0075] LDA (Linear Discriminant Analysis) is a powerful supervised machine learning algorithm, mainly used for dimensionality reduction and classification, which can effectively classify samples by maximizing inter-class differences and minimizing intra-class differences. In the application of Raman spectroscopy, LDA can process high-dimensional spectral data and extract important features that help distinguish different categories by reducing dimensionality. In this protocol, T24 cells and white blood cells were firstly detected at 600-1800 cm -1 50 SERS spectra within the spectral range were analyzed. With the assistance of machine learning, the data set was divided into a training set and a test set in a ratio of 7:3. In order to prevent overfitting of the model and evaluate its stability, a 5-fold cross validation was used. Finally, a binary classification model capable of distinguishing tumor cells from leukocytes was established. The results showed that the accuracy of the model in distinguishing T24 tumor cells from leukocytes was 96.48%. Subsequently, the same method was used to process and model the other two sets of SERS data. The results showed that the accuracy of the other two classification models in distinguishing U251 tumor cells and Panc02 tumor cells reached 90.23% and 93.33%, respectively. These results show that the classification model established based on this algorithm has good stability and generalization.

[0076] In the classification model of machine learning, the performance evaluation of the model is very important. Confusion matrix and ROC curve (receiver operating characteristic curve) are common methods to evaluate the effectiveness of classification models. Confusion matrix is ​​a table used to visualize the classification results of classification models in machine learning. It shows the relationship between the model's predicted results and the actual results, and is a common method to evaluate the performance of classification models. The purpose of the confusion matrix is ​​to help us understand the performance of classification models in different categories. By dividing the classification results of the model into true categories (True) and false categories (False), we can calculate a series of performance indicators. These indicators help us quantify the classification accuracy, reliability, and comprehensiveness of the model. From Figure 5 The confusion matrix shows that the binary classification model constructed by the LDA algorithm can effectively distinguish tumor cells from white blood cells. The overall performance of the classification model was evaluated in detail through the confusion matrix. The results showed that in three different groups of samples, the classification model had an overall distinction accuracy of 94%, 94% and 100% for T24, U251 and Panc02 cells, respectively. At the same time, the distinction accuracy of white blood cells in the three groups of samples also reached 89%, 94% and 100%, respectively. The above results all confirm that the classification model has excellent distinction efficiency.

[0077] The ROC curve is another method to evaluate the performance of a classification model by plotting the relationship between the true positive rate (TPR) and the false positive rate (FPR). In the ROC curve, the horizontal axis represents the FPR, or false positive rate, and the vertical axis represents the TPR, or true positive rate. For example, in medicine, the ROC curve can be used to evaluate the effectiveness of a diagnostic model in classifying cancer patients and healthy people. If the model can correctly distinguish cancer patients from healthy people and has a lower FPR and a higher TPR, the ROC curve should show a pattern close to the upper left corner. In addition, the ROC curve can calculate the area under it (AUC), which represents the overall quality of the model's performance. The AUC ranges from 0 to 1, and the closer the value is to 1, the better the model's performance. Therefore, the ROC curve can be used as a useful method for evaluating machine learning models.

[0078] In this paper, ROC curve ( Figure 6) The performance of the classification model was verified by ROC curves. The ROC curve shows the excellent overall performance of the model in multiple mixed samples. When distinguishing mixed samples of T24 cells and leukocytes, the classification model showed a sensitivity of up to 94.44% and a specificity of 94.10%, and the AUC value reached 0.97. In the face of mixed samples of U251 cells and leukocytes, the sensitivity of the classification model reached 94.44%, while the specificity reached 100%, and the AUC value also reached the optimal 1.00. For mixed samples of Panc02 cells and leukocytes, the model also showed 100% sensitivity and specificity, and the AUC value reached 1.00. This reflects its consistent and accurate recognition ability between different combinations of tumor cells and leukocytes. These results once again verify that the classification model is efficient and reliable in distinguishing tumor cells from leukocytes.

[0079] The various aspects, embodiments, and features of the present invention should be considered to be illustrative in all aspects and not limiting of the present invention, the scope of which is defined solely by the claims. Other embodiments, modifications, and uses will be apparent to those skilled in the art without departing from the spirit and scope of the claimed invention.

[0080] In the preparation method of the present invention, the order of each step is not limited to the order listed. For those skilled in the art, without creative work, the order of each step is also within the protection scope of the present invention. In addition, two or more steps or actions can be performed simultaneously.

[0081] Finally, it should be noted that the specific embodiments described herein are merely examples of the present invention, and are not intended to limit the implementation methods of the present invention. A person skilled in the art of the present invention may make various modifications or supplements to the specific embodiments described, or replace them in a similar manner. It is not necessary and impossible to provide all examples of all implementation methods here. However, these obvious changes or modifications derived from the essential spirit of the present invention still fall within the scope of protection of the present invention, and interpreting them as any additional limitation is contrary to the spirit of the present invention.

Claims

1. A method for preparing a flower-shaped cuprous oxide-based SERS biosensor, characterized in that: The following steps are involved: S1. Add polyvinyl pyrrolidone to the copper salt solution, then add sodium hydroxide solution, stir for 0.5 to 3 hours, then add the first reducing agent solution, heat for reaction, and obtain flower-shaped Cu2O after centrifugation, washing and drying; the polyvinyl pyrrolidone has the appearance of white crystalline powder, a molecular weight of 1200000 to 1400000, and a density of 1.6 to 1.7 g / cm 3 , the mass ratio of polyvinyl pyrrolidone to copper salt is 30-40:1; S2, preparing the flower-shaped Cu2O obtained in step S1 into a suspension, adding a second reducing agent solution, stirring, and then adding a noble metal salt solution, reacting, and obtaining a flower-shaped Cu2O@noble metal suspension after centrifugation, washing, and resuspending; S3. Drop the flower-shaped Cu2O@precious metal suspension on a clean silicon wafer and obtain a flower-shaped cuprous oxide-based SERS biosensor after drying.

2. The preparation method according to claim 1, characterized in that: The polyvinyl pyrrolidone was purchased from MacLean Reagent Company and has the appearance of white crystalline powder, a molecular weight of 1,250,000 to 1,350,000, and a density of 1.69 g / cm 3 ; The mass ratio of polyvinyl pyrrolidone to copper salt is 33-37:

1.

3. The preparation method according to claim 1, characterized in that: The copper salt is one or more of copper chloride, copper nitrate, copper sulfate and copper acetate; and / or, the first reducing agent is one or more of ascorbic acid, glucose, and hydroxylamine; and / or, the second reducing agent is one or more of sodium citrate, NaBH4, hydrazine hydrate, and ascorbic acid; And / or, the noble metal salt is silver chloride or chloroauric acid. and / or, the molar ratio of copper salt to sodium hydroxide is 1:10 to 50; and / or, the molar ratio of the copper salt to the first reducing agent is 1:10 to 50; and / or, the mass ratio of flower-shaped Cu2O to noble metal salt is 1 to 5:1; and / or, the molar ratio of the noble metal salt to the second reducing agent is 1:4 to 50; And / or, the heating reaction temperature of step (1) is 45-65° C. and the time is 2-6 hours; And / or, the reaction in step (2) is carried out at 10-45° C. for 0.5-5 h.

4. A flower-shaped cuprous oxide-based SERS biosensor prepared by the preparation method according to claim 1.

5. A method for establishing a tumor cell / leukocyte classification model based on a flower-shaped cuprous oxide-based SERS biosensor according to claim 4, characterized in that: The method comprises the following steps: (1) Cultivating tumor cells; (2) adding tumor cells to diluted anticoagulated rabbit blood, injecting into the microfluidic chip and collecting the cell separation fluid at the outlet; (3) dropping the cell separation solution on the flower-shaped cuprous oxide-based SERS biosensor, and collecting Raman spectral data of tumor cells and leukocytes on the sensor after spreading; (4) The Raman spectral data were subjected to machine learning dimensionality reduction and feature extraction based on the linear discriminant analysis (LDA) algorithm to establish a tumor cell / leukocyte binary classification model, and the performance of the model was evaluated using confusion matrix and / or ROC curve.

6. The method according to claim 5, characterized in that The tumor cells are one or more of bladder cancer T24 cells, brain glioma U251 cells, and pancreatic cancer Panc02 cells.

7. The method according to claim 5, characterized in that In step S3, cells with a diameter of ≤15 μm under a microscope are regarded as white blood cells, and cells with a diameter of ≥20 μm are regarded as tumor cells, thereby obtaining Raman spectral data of the corresponding cells.

8. The method according to claim 5, characterized in that Raman spectroscopy data is in the spectral range of 600-1800cm -1 SERS spectral data within.

9. The method according to claim 5, characterized in that The area under the ROC curve was used to evaluate the performance of the binary classification model.

10. The method according to claim 5, characterized in that When the label-free SERS detection data of bladder cancer T24 tumor cells and leukocytes were selected to establish the tumor cell / leukocyte binary classification model, the area under the ROC curve of the classification model was 0.97; When the label-free SERS detection data of brain glioma U251 cells and leukocytes were selected to establish the tumor cell / leukocyte binary classification model, the area under the ROC curve of the classification model was 1.00; When the label-free SERS detection data of pancreatic cancer Panc02 cells and leukocytes were selected to establish a tumor cell / leukocyte binary classification model, the area under the ROC curve of the classification model was also 1.00.

Images