Rapid detection and discrimination method for novel high-sensitivity coronavirus variant

Abstract

The invention relates to a high-sensitivity rapid detection and discrimination method for novel coronavirus variants. The high-sensitivity novel coronavirus variant rapid detection and discrimination method for non-disease diagnosis and treatment purposes comprises the following steps: (1) dropwise adding and pretreating a sample: dropwise adding a novel coronavirus sample to be detected onto a high-sensitivity nano-porous plasma SERS chip, and drying to obtain a prefabricated sample; (2) Raman spectrum data acquisition: acquiring Raman spectrum data of the prefabricated sample by using a confocal Raman spectrometer; and (3) automatic result discrimination: performing result identification and classification on the Raman spectrum data by using a machine learning model to complete rapid detection and discrimination of the novel high-sensitivity coronavirus variant with non-disease diagnosis and treatment purposes.

Description

Technical Field

[0001] This invention belongs to the field of virus detection and machine learning technology, specifically relating to a highly sensitive rapid detection and identification method for novel coronavirus variants. Background Technology

[0002] Many virus detection methods, such as enzyme-linked immunosorbent assay (ELISA) and polymerase chain reaction (PCR), have been greatly developed; however, these methods often require specialized personnel, equipment, and complex reagents. Surface-enhanced Raman scattering (SERS) technology, as a powerful tool for detecting and identifying molecular species at the single-molecule level, has been widely explored and applied due to its speed, sensitivity, and reliability. SERS sensing solutions have been widely used in many fields, including environmental science, medical diagnostics, food safety, and bioanalysis.

[0003] Although SERS technology can successfully obtain Raman spectra of different SARS-CoV-2 variants, it is generally difficult to accurately identify and distinguish these spectra by human visual observation alone because the spectra of different variants of the virus are highly similar, and there are interferences from differences in viral load in different patients and impurities in body fluids. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a highly sensitive rapid detection and discrimination method for novel coronavirus variants. By introducing machine learning methods, different SARS-CoV-2 variants can be accurately identified, demonstrating that with the establishment of a database, it is possible to achieve automated identification of human viruses directly through Raman spectroscopy.

[0005] Specifically, this invention provides a highly sensitive rapid detection and discrimination method for novel coronavirus variants for non-disease diagnosis and treatment purposes, the detection and discrimination method comprising the following steps: (1) Sample addition and pretreatment: The novel coronavirus sample to be tested was added onto a high-sensitivity nanoporous plasma SERS chip and dried to obtain a pre-prepared sample; (2) Raman spectral data acquisition: Raman spectral data of the pre-prepared sample were acquired using a confocal Raman spectrometer; (3) Automated result discrimination: The Raman spectral data is identified and classified using a machine learning model to complete the rapid detection and discrimination of highly sensitive novel coronavirus variants for non-disease diagnosis and treatment purposes.

[0006] Preferably, the amount of the novel coronavirus test sample is 5-20 μL, more preferably 10 μL; the drying temperature is 30-45℃, preferably drying at 37℃ for more than 15 minutes until completely dry.

[0007] Preferably, the method for fabricating the highly sensitive nanoporous plasma SERS chip includes the following steps: (1) The gold nanoparticle seed crystals are heated and stirred in colloidal gold sol to coat the surface of the gold nanoparticle seed crystals with a gold shell layer. The above heating and stirring steps are repeated to coat the gold shell layer and obtain a sol containing colloidal gold particles. (2) The sol containing colloidal gold particles was centrifuged to obtain the highly sensitive nanoporous plasma SERS chip.

[0008] Preferably, in step (1), the gold nanoparticle seed crystals are spherical; the absorption peak is located at 300-600nm, preferably 400-600nm; and the particle size is 10-30nm, preferably 10-20nm.

[0009] Preferably, in step (1), the heating and stirring temperature is 80-100℃, preferably 85℃; the thickness of the gold shell layer is 1-10nm, preferably 1-5nm.

[0010] Preferably, in step (1), the number of times the heating and stirring steps are repeated to coat the gold shell is 3-8, preferably 5-8; the size of the colloidal gold particles in the sol containing colloidal gold particles is 20-100 nm, preferably 20-50 nm.

[0011] Preferably, in step (2), the amount of sol containing colloidal gold particles used for centrifugation is 10-50 mL, more preferably 20-50 mL, and even more preferably 20-40 mL; the centrifugation process is as follows: centrifuge at a speed of 5000-12000 r / min for 5-15 minutes, preferably centrifuge at a speed of 8000-12000 r / min for 8-12 minutes.

[0012] Preferably, the diameter of the highly sensitive nanoporous plasma SERS chip is 1-5 mm.

[0013] Preferably, the SARS-CoV-2 variants detected in the pre-prepared sample include: A, AlphaB.1.1.7, Beta B.1.351, Delta B.1.617.2, Omicron B.1.1.529, Omicron BA.4, or Omicron BA.5.

[0014] Preferably, the machine learning model is a residual neural network model; the construction and application of the machine learning model includes the following steps: (1) Constructing the model dataset: Raman spectra of the above 7 novel coronavirus variants were collected. The number of spectra of each variant was 100. After all the spectra data were shuffled, they were divided into training set and test set in a ratio of 1:4. (2) Training the model: The machine learning model is trained using the training set data; (3) Model performance evaluation: The performance of the machine learning model is evaluated using test set data to complete the construction of the machine learning model.

[0015] Beneficial effects This invention combines surface-enhanced Raman scattering (SERS) technology and machine learning techniques to provide a highly sensitive and rapid detection and discrimination method for novel coronavirus variants, which is expected to achieve automated identification of human viruses directly through Raman spectroscopy. On the one hand, compared to commonly used PCR nucleic acid detection, virus detection based on SERS technology is more convenient and faster; on the other hand, the machine learning-based discrimination method can realize functions such as spectral extraction, learning, classification, and prediction, greatly improving data processing efficiency and potentially enabling automated identification of human viruses. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the residual neural network deep learning model structure as an example of the present invention; Figure 2 Characterization images of the surface and internal morphology of the high-sensitivity nanoporous plasma SERS chip, which is an example of the present invention; Figure 3 This is a schematic diagram of the SARS-CoV-2 variant identification and prediction results obtained based on the RNT model in Example 1; Figure 4 Two-dimensional / three-dimensional visualizations of the SARS-CoV-2 variant identification prediction results obtained based on the SVM model in Comparative Example 1; Figure 5 (a) shows the Raman spectra of seven SARS-CoV-2 variants, and (b) and (c) show the Raman mapping spectra of the type A variant and the type B.1.1.7 variant. Detailed Implementation

[0017] The present invention will be further illustrated by the following embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the present invention.

[0018] The following is an exemplary description of the rapid detection and discrimination method for highly sensitive novel coronavirus variants for non-disease diagnosis and treatment purposes provided by the present invention. The method may include the following steps: (1) Sample addition and pretreatment: The novel coronavirus sample to be tested was added onto a high-sensitivity nanoporous plasma SERS chip and dried to obtain a pre-prepared sample; (2) Raman spectral data acquisition: Raman spectral data of the pre-prepared sample were acquired using a confocal Raman spectrometer; (3) Automated result discrimination: The Raman spectral data is identified and classified using a machine learning model to complete the rapid detection and discrimination of highly sensitive novel coronavirus variants for non-disease diagnosis and treatment purposes.

[0019] In some embodiments, the amount of the novel coronavirus sample to be tested can be 5-20 μL, preferably 10 μL; the drying temperature can be 30-45°C, preferably drying at 37°C for more than 15 minutes until completely dry.

[0020] The highly sensitive nanoporous plasma SERS chip used in this invention is fabricated through a one-step self-assembly of gold nanoparticles. In some embodiments, the fabrication method of the highly sensitive nanoporous plasma SERS chip may include the following steps: (1) The gold nanoparticle seed crystals are heated and stirred in colloidal gold sol to coat the surface of the gold nanoparticle seed crystals with a gold shell layer. The above heating and stirring steps are repeated to coat the gold shell layer and obtain a sol containing colloidal gold particles. (2) The sol containing colloidal gold particles was centrifuged to obtain the highly sensitive nanoporous plasma SERS chip.

[0021] In some embodiments, in step (1), the gold nanoparticle seed crystals are spherical; the absorption peak is located at 300-600nm, preferably 400-600nm; and the particle size is 10-30nm, preferably 10-20nm.

[0022] In some embodiments, in step (1), the heating and stirring temperature can be 80-100℃, preferably 85℃; the thickness of the gold shell layer is 1-10nm, preferably 1-5nm.

[0023] In some embodiments, in step (1), the number of times the heating and stirring steps are repeated to coat the gold shell layers is 3-8, preferably 5-8; the size of the colloidal gold particles in the sol containing colloidal gold particles is 20-100 nm, preferably 20-50 nm.

[0024] In some embodiments, in step (2), the amount of sol containing colloidal gold particles used for centrifugation is 10-50 mL, preferably 20-50 mL, more preferably 20-40 mL; the centrifugation process is: centrifuging at a speed of 5000-12000 r / min for 5-15 minutes, preferably centrifuging at a speed of 8000-12000 r / min for 8-12 minutes.

[0025] In some embodiments, the highly sensitive nanoporous plasma SERS chip has a diameter of 1-5 mm and a three-dimensional porous microstructure.

[0026] In some implementations, the SARS-CoV-2 variants contained in the pre-prepared sample may include A, Alpha B.1.1.7, Beta B.1.351, Delta B.1.617.2, Omicron B.1.1.529, Omicron BA.4, or Omicron BA.5.

[0027] The SARS-CoV-2 S protein identification technology based on SERS spectroscopy in this invention distinguishes different variants of the novel coronavirus by differentiating the main Raman characteristic peaks of the S protein. Table 1 below shows the Raman characteristic peak assignments of the SARS-CoV-2 S protein. These Raman characteristic peaks can all be attributed to protein chemical bond vibrations, including main chain conformation, side chains, and surface glycans. In addition to the strong Raman signal exhibited by specific amide vibrations of the main chain, groups with large changes in polarizability of certain amino acid conformation vibrations in the side chain of the protein usually produce high Raman intensities in the Raman spectrum. The characteristic Raman bands of the main chain are related to the vibrational modes of the CONH group and CC, CN, and CH bonds. The Raman bands of the side chains mainly originate from aromatic amino acid residues (such as benzene, benzene rings, and the indole ring of tryptophan) and sulfur-containing residues (such as CS and SS bonds). 882, 1053, and 1441 cm⁻¹ -1 These three main characteristic bands in the vicinity can be attributed to CH2 deformation of tryptophan, phenylalanine, and carbohydrates, respectively. The SARS-CoV-2S protein shows bands at 760, 882, and 1264 cm⁻¹. -1 The characteristic peak at that location can be attributed to amide V, tryptophan, and amide III (α-helix). Table 1. Assignment of Raman characteristic peaks of SARS-CoV-2S protein Note: Ala: alanine, Val: valine, Trp: tryptophan, Tyr: tyrosine, Ile: isoleucine, Ser: serine, Glu: glutamic acid, Thr: threonine, Pro: proline, Aspartic acid, Phe: phenylalanine, Arg: arginine, Gly: glycine, Met: methionine, Cys: cysteine, Fuc: fucose, Man: mannose, Gal: galactose, GlcNAc: N-acetyl-D-glucosamine, str: tensile vibration, ben: bending vibration.

[0028] In some implementations, the machine learning model is a residual neural network (RNT) model, which is used to automate the judgment of results. Compared with traditional machine learning models such as support vector machines (SVM), RNT, as a deep neural network, is particularly adept at handling complex, high-dimensional data. Compared with traditional deep learning models, by introducing residual blocks, RNT solves the gradient vanishing problem in deep neural networks, allowing the network to be deeper. This means that ResNet can have more layers, thereby extracting richer and more complex features.

[0029] The construction and application of the machine learning model may include the following steps: (1) Constructing the model dataset: Raman spectra of the above 7 novel coronavirus variants were collected. The number of spectra of each variant was 100. After all the spectra data were shuffled, they were divided into training set and test set in a ratio of 1:4. (2) Training the model: The machine learning model is trained using the training set data; (3) Model performance evaluation: The performance of the machine learning model is evaluated using test set data to complete the construction of the machine learning model.

[0030] The following combination Figure 1 This paper exemplifies the structure of the residual neural network model used in this invention. The RNT model employed in this invention relies on shortcut connections and introduces "residual blocks" to enable the network to learn identity mappings, thus preserving previous gradient information and addressing the gradient vanishing problem in deep architectures. Considering computational power and training efficiency, the RNT framework may include an initial convolutional layer (Conv Layer), a max pooling layer, and may also include batch normalization (BN), rectified linear transformation (ReLU), followed by five sequence blocks. Each sequence block includes one convolutional block and two identity blocks with shortcuts. Finally, these blocks are followed by an average pooling layer (AVG Pool), a flattening layer, and a fully connected layer (FC Layer).

[0031] To enable more accurate identification of Raman spectra of different SARS-CoV-2 variants, this invention employs a residual neural network model with a deep learning structure for spectral classification and identification. Compared to conventional machine learning methods, deep learning networks have deeper layers, enabling nonlinear transformations and automatic feature extraction and transformation for better data classification.

[0032] In summary, the technical solution disclosed in this invention combines a machine learning model with SARS-CoV-2 S protein identification technology based on SERS spectroscopy. Different machine learning models have varying performance characteristics; this invention, through adjustments and improvements based on a large number of training samples, solves the compatibility problem between data types and model readings. The highly sensitive rapid detection and discrimination method for novel coronavirus variants provided by this invention can accurately identify different SARS-CoV-2 variants. It also demonstrates that, through the establishment of a database, it is possible to achieve automated identification of human viruses directly using Raman spectroscopy.

[0033] The following examples further illustrate the present invention in detail. It should also be understood that the following examples are only for further explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-essential improvements and adjustments made by those skilled in the art based on the above description of the present invention are within the scope of protection of the present invention. The specific process parameters, etc., in the following examples are merely examples within a suitable range; that is, those skilled in the art can make appropriate selections within the range based on the description herein, and are not intended to be limited to the specific values ​​in the examples below. Unless otherwise specified, the technical means used in the examples are conventional means well known to those skilled in the art.

[0034] Example 1

[0035] This embodiment utilizes a residual neural network model to distinguish the Raman spectra of seven SARS-CoV-2 variants: A, Alpha B.1.1.7, Beta B.1.351, Delta B.1.617.2, Omicron B.1.1.529, Omicron BA.4, or Omicron BA.5. The specific method steps are as follows: (1) Preparation of a highly sensitive nanoporous plasma SERS chip; (2) Model dataset construction: Seven SARS-CoV-2 variant samples were dropped onto a high-sensitivity nanoporous plasma SERS chip and dried to obtain pre-prepared samples of seven SARS-CoV-2 variants. Raman spectra of the seven novel coronavirus variants were collected using a confocal Raman spectrometer, with 100 spectra collected for each variant. All data were shuffled and divided into training and test sets at a ratio of 1:4. (3) Training the model: Use the training set data divided in step (2) to train the residual neural network model; (4) Model performance evaluation: Using the residual neural network model obtained in step (3), predict the test set divided in step (2) and output the prediction results.

[0036] Figure 2Characterization images of the surface and internal morphology of a high-sensitivity nanoporous plasma SERS chip, as an example of this invention. From... Figure 2 (a) SEM images at different scales show that the NPG SERS chip fabricated in this work has a uniform and continuous nanopore structure on its surface, and the size of the nanopores ( Figure 2 (b) It can be well matched in size to the SARS-CoV-2 S protein, making it easy to enter the nanoporous structure and confine it inside; furthermore... Figure 2 The SEM image in (c) shows the nanostructure of the cross-section of the NPG chip. It can be seen that continuous nanopores run across the entire cross-section, which indicates that the NPG chip manufactured by this invention has a highly uniform structure, both on the surface and inside, which is beneficial for producing uniform SERS enhancement.

[0037] Figure 3 This is a schematic diagram of the SARS-CoV-2 variant identification and prediction results obtained based on the RNT model in Example 1. As can be seen from the figure, the score for each variant is close to 1, indicating that each type of Raman spectrum has its own distinct characteristics.

[0038] Table 2 below shows the confusion matrix of the seven variant test sets in Example 1. According to the basic theory of machine learning, the diagonal in the table represents the number of correct predictions, and the numbers in other positions represent the number of incorrect predictions. For example, the number 4 in the second row and third column means that the number of Alpha strains predicted as Beta strains is 4. As can be seen from the table, there were only misclassification cases between Alpha and Beta variants, while all other categories were correctly identified, with an overall accuracy of 95.71%. Table 2. Confusion matrix of the test sets for the 7 variant strains in Example 1

[0039] Comparative Example 1

[0040] The specific method and steps for distinguishing the Raman spectra of seven SARS-CoV-2 variants (A, Alpha B.1.1.7, Beta B.1.351, Delta B.1.617.2, Omicron B.1.1.529, Omicron BA.4, or Omicron BA.5) in this comparative example are the same as in Example 1. The main difference is that the machine learning model used is a support vector machine model.

[0041] Figure 4 The images show two-dimensional / three-dimensional visualizations of the SARS-CoV-2 variant identification prediction results obtained from the SVM model in Comparative Example 1. As can be seen from the figures, plotting the SVM classification results on a two-dimensional coordinate system reveals that these spectra are difficult to distinguish. Figure 4(a) Even with the addition of another coordinate dimension, these spectra remain difficult to distinguish. Figure 4 (b)).

[0042] Comparative Example 2

[0043] The specific method and steps for distinguishing the Raman spectra of seven SARS-CoV-2 variants (A, Alpha B.1.1.7, Beta B.1.351, Delta B.1.617.2, Omicron B.1.1.529, Omicron BA.4, or Omicron BA.5) in this comparative example are the same as in Example 1. The main difference is that the distinction is made directly by observing the differences between the spectra.

[0044] Figure 5 (a) shows the Raman spectra of seven SARS-CoV-2 variants; (b) and (c) show the Raman mapping spectra of variant A and variant B.1.1.7. Figure 5 As can be seen in (a), due to the similarity of the S protein structure in different variants, their Raman spectra are likely to be highly similar; from Figure 5 As can be seen from (b) and (c), there can be significant differences in the spectra of the same strain, making it very difficult to identify these spectra with the human eye.

[0045] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A highly sensitive rapid detection and discrimination method for novel coronavirus variants for non-disease diagnosis and treatment purposes, characterized in that, The detection and discrimination method includes the following steps: (1) Sample addition and pretreatment: The novel coronavirus test sample was added to the high-sensitivity nanoporous plasma SERS chip and dried to obtain the pre-prepared sample; (2) Raman spectral data acquisition: Raman spectral data of the pre-prepared sample were acquired using a confocal Raman spectrometer; (3) Automated result discrimination: The Raman spectral data is identified and classified using a machine learning model to complete the rapid detection and discrimination of highly sensitive novel coronavirus variants for non-disease diagnosis and treatment purposes.

2. The detection and discrimination method according to claim 1, characterized in that, The amount of the novel coronavirus sample to be tested is 5-20 μL, preferably 10 μL; the drying temperature is 30-45℃, preferably at 37℃ for more than 15 minutes until completely dry.

3. The detection and discrimination method according to claim 1 or 2, characterized in that, The fabrication method of the highly sensitive nanoporous plasma SERS chip includes the following steps: (1) The gold nanoparticle seed crystals are heated and stirred in colloidal gold sol to coat the surface of the gold nanoparticle seed crystals with a gold shell layer. The above heating and stirring steps are repeated to coat the gold shell layer and obtain a sol containing colloidal gold particles. (2) The sol containing colloidal gold particles was centrifuged to obtain the highly sensitive nanoporous plasma SERS chip.

4. The detection and discrimination method according to any one of claims 1-3, characterized in that, In step (1), the gold nanoparticle seed crystals are spherical; the absorption peak is located at 300-600nm, preferably 400-600nm; the particle size is 10-30nm, preferably 10-20nm.

5. The detection and discrimination method according to any one of claims 1-4, characterized in that, In step (1), the heating and stirring temperature is 80-100℃, preferably 85℃; the thickness of the gold shell layer is 1-10nm, preferably 1-5nm.

6. The detection and discrimination method according to any one of claims 1-5, characterized in that, In step (1), the number of times the heating and stirring steps are repeated to coat the gold shell is 3-8 times, preferably 5-8 times; the size of the colloidal gold particles in the sol containing colloidal gold particles is 20-100nm, preferably 20-50nm.

7. The detection and discrimination method according to any one of claims 1-6, characterized in that, In step (2), the amount of sol containing colloidal gold particles used for centrifugation is 10-50 mL, preferably 20-50 mL, and more preferably 20-40 mL; the centrifugation process is as follows: centrifuge at a speed of 5000-12000 r / min for 5-15 minutes, preferably centrifuge at a speed of 8000-12000 r / min for 8-12 minutes.

8. The detection and discrimination method according to any one of claims 1-7, characterized in that, The diameter of the highly sensitive nanoporous plasma SERS chip is 1-5 mm.

9. The detection and discrimination method according to any one of claims 1-8, characterized in that, The SARS-CoV-2 variants detected in the pre-prepared samples include: A, Alpha B.1.1.7, Beta B.1.351, Delta B.1.617.2, Omicron B.1.1.529, Omicron BA.4, or Omicron BA.

5.

10. The detection and discrimination method according to any one of claims 1-9, characterized in that, The machine learning model is a residual neural network model; the construction and application of the machine learning model includes the following steps: (1) Constructing the model dataset: Raman spectra of the above 7 novel coronavirus variants were collected. The number of spectra of each variant was 100. After all the spectra data were shuffled, they were divided into training set and test set in a ratio of 1:

4. (2) Training the model: The machine learning model is trained using the training set data; (3) Model performance evaluation: The performance of the machine learning model is evaluated using test set data to complete the construction of the machine learning model.

Images