Natural killer cell marker, screening device and application

Abstract

The invention discloses a natural killer cell marker, a screening device and application. The natural killer cell markers obtained through screening are RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, the AUC values of prediction models are 0.91, 0.92, 0.92 and 0.92 respectively, and the identification effect is close to or even superior to that of existing database molecular markers. Therefore, the four new markers can be used for natural killer cell identification research of single cell transcriptome data analysis. In addition, the machine learning model constructed by the invention can also be used for identifying natural killer cells in data samples.

Description

Technical Field

[0001] The present invention relates to the field of biotechnology, and in particular to a natural killer cell marker, a screening device and an application thereof. Background Art

[0002] The blood system of mammals contains more than 10 different mature cell types, all of which are differentiated from the same progenitor cell type - hematopoietic stem cells (HSC). Mature blood cell types can be roughly divided into two categories: (1) myeloid cells, mainly including megakaryocytes, erythroblasts, mast cells, eosinophils / basophils / neutrophils, monocytes, dendritic cells and macrophages; (2) lymphoid cells, mainly including B cells, T cells and natural killer cells.

[0003] Natural killer cells (NK cells) are natural key players of the immune system that automatically kill cells before they are activated by tumors or viral infections. The cytoplasm of natural killer cells is rich in small granules containing special proteins such as proteases and proteases called granzymes. It is released near the killed cells and spreads within the target cells through the formation of pores, granzymes and related molecules on the cell membrane of the target cells, leading to the destruction of the target cells. Therefore, natural killer cells are key innate effector cells that non-antigen-specifically target virus-infected cells and tumor cells.

[0004] Given the important role of natural killer cells in the body's immunity, especially in anti-tumor immunity, they have received extensive attention in scientific and clinical research. However, the primary premise for conducting relevant research is how to identify / enrich natural killer cells. The screening method of traditional biological experiments mainly utilizes the characteristics of natural killer cells that highly express CD56 cell surface marker proteins to separate natural killer cells from a group of cells by flow sorting or magnetic bead sorting. However, this method is time-consuming and labor-intensive, and there are certain limits on accuracy. In 2017, NVIGEN INC's "complex for capturing rare cells in samples" patent (patent number: WO2017035419-A1) designed beads that are operably connected to polyethylene glycol (PEG) compounds and analyte capture components connected to the beads. The analyte capture component can specifically bind to the surface markers of rare cells, thereby capturing rare immune cells or circulating cells including natural killer cells to indicate the presence of disease. This patent is based on the design of a complex of beads and analyte capture components to identify rare cells including natural killer cells, mainly for disease prediction. This technology requires the design of beads and corresponding analyte capture components, which is costly and technically difficult. After the cells are captured, the mixed cell population needs to be differentiated a second time, and the main purpose is to indicate the disease state. It is not well suited for the identification of natural killer cell types. The patent authorized in 2013, "Identification of dendritic killer cells in peripheral blood samples of cancer patients by several cell surface markers", uses several cell surface markers selected from natural killer cells, dendritic cells and human leukocyte antigens to identify and screen natural killer cells and dendritic cells in human peripheral blood samples obtained from cancer patients for the development of cancer cell immunotherapy. This technology uses 4 step-by-step screening steps to finally separate the phenotype of HLA - G - CD14 - CD19 - CD3 - CD56 + HLA - DR + This technology gradually identifies and enriches natural killer cells and dendritic cells through cell surface markers, and the screening process is relatively cumbersome. In addition, the identification method by evaluating the expression of cell surface markers is somewhat subjective, and its specificity and accuracy are also somewhat insufficient. Summary of the invention

[0005] The present invention aims to provide a natural killer cell marker, a screening device and an application, so as to provide a new natural killer cell marker.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present invention, a marker is provided for use in identifying, enriching or sorting natural killer cells, the marker comprising a gene marker and / or a protein marker, the gene marker being one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein marker being one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes.

[0007] Further, when the application is to identify natural killer cells, the application includes detecting the expression level of a gene marker, preferably, detecting the expression level of a gene marker is detecting the mRNA expression level or protein expression level of the gene marker; more preferably, the mRNA expression level of the gene marker is detected by high-throughput sequencing; optionally, the application is for the identification of natural killer cells in single-cell transcriptome studies.

[0008] Furthermore, natural killer cells from a biological sample are identified, enriched or sorted, wherein the biological sample is selected from one or more of the following: peripheral blood, bone marrow, umbilical cord blood, liver or lymph node.

[0009] According to another aspect of the present invention, there is provided an application of a marker in the preparation of related products for identifying, enriching or sorting natural killer cells, the marker comprising a gene marker and / or a protein marker, the gene marker being one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein marker being one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes.

[0010] Furthermore, when the application is to prepare related products for identifying natural killer cells, the related products include using detection reagents to detect the expression of gene markers,

[0011] Preferably, the detection reagent is a detection reagent for detecting the mRNA expression level or protein expression level of a gene marker;

[0012] More preferably, the detection reagent for detecting the mRNA expression level of the gene marker is a probe and / or a primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

[0013] Further, natural killer cells derived from a biological sample are identified, enriched or sorted, wherein the biological sample is selected from one or more of the following: peripheral blood, bone marrow, umbilical cord blood, liver or lymph node;

[0014] Preferably, the product of interest comprises a kit.

[0015] According to another aspect of the present invention, a kit for identifying natural killer cells is provided. The kit includes a detection reagent, which is a detection reagent for detecting the expression of gene markers and / or protein markers, the gene marker is one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein marker is one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes; preferably, the detection reagent is a detection reagent for detecting the mRNA expression or protein expression of the gene marker; more preferably, the detection reagent for detecting the mRNA expression of the gene marker is a probe and / or primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

[0016] Furthermore, the detection reagent is a detection reagent for detecting the mRNA expression level or protein expression level of the gene marker; preferably, the detection reagent for detecting the mRNA expression level of the gene marker is a probe and / or primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

[0017] According to another aspect of the present invention, a screening device for natural killer cell markers is provided. The screening device comprises: a comparison unit configured to calculate the gene expression profile of terminally differentiated blood cells in peripheral blood under normal physiological conditions, obtain the difference characteristics of the gene expression profiles of natural killer cells and other blood cells by comparison, and screen out candidate molecular markers; a screening unit configured to combine the candidate molecular markers screened in the comparison unit with a machine learning algorithm to construct a natural killer cell identification gene model, and screen out natural killer cell markers.

[0018] Furthermore, other blood cells include B lymphocytes, T lymphocytes, dendritic cells, erythroblasts, monocytes and neutrophils; preferably, the machine learning algorithm is a logistic regression model.

[0019] Furthermore, the raw data in the comparison unit is RNA-Seq data of terminally differentiated blood cells in peripheral blood;

[0020] Preferably, calculating the gene expression profile of terminally differentiated blood cells in peripheral blood under normal physiological conditions in the comparison unit includes quantifying the gene expression profile of blood cells;

[0021] More preferably, the gene expression profile quantification includes quality control of the original blood cell RNA-Seq sequencing data, including cutting the linker, removing low-quality reads, aligning the quality-controlled sequences to the human genome, and filtering out samples with an alignment rate less than a threshold, and quantifying using the TPM method;

[0022] Preferably, the differential features of the gene expression profiles of natural killer cells and other blood cells obtained by comparison in the comparison unit include performing pairwise differential expression analysis on natural killer cells and other blood cells, and taking the average expression level of the same gene of the same cell type; for natural killer cells, there is a list of differentially upregulated genes compared with other blood cells, and the upregulated differentially upregulated genes specific to the natural killer cell type are obtained by taking the intersection, preferably after removing the marker genes reported in the CellMarker database, as candidate molecular markers for natural killer cells.

[0023] Furthermore, the screening unit is configured to take a single candidate molecular marker as input, and use a machine learning classification model to construct a model to identify natural killer cells in the data sample; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; if the AUC value is greater than a threshold, the candidate molecular marker is judged as a marker for identifying, enriching or sorting natural killer cells.

[0024] Furthermore, the screening unit further includes an identification module, which is configured to perform comparative verification on the marker identification effect; the comparative verification includes:

[0025] A control gene set was selected, and a machine learning classification model was used to build a model to identify natural killer cells in data samples; the prediction score was automatically evaluated by the model, and the model identification effect was evaluated by the area under the receiver operating characteristic curve (AUC) value; the identification effect of the marker was verified by comparing the AUC values ​​of the control gene set and the marker.

[0026] According to another aspect of the present invention, a device for identifying natural killer cell markers is provided, and the identification device is configured as follows: selecting a control gene set, using a machine learning classification model to construct a model to identify natural killer cells on a data sample; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; and the identification effect of the marker is verified by comparing the AUC value of the control gene set with that of the marker.

[0027] The technical solution of the present invention is applied, and based on a large number of public peripheral blood transcriptome data sets, a machine learning method is used to screen and develop new natural killer cell molecular markers. The new natural killer cell molecular markers RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes screened have prediction model AUC values ​​of 0.91, 0.92, 0.92, and 0.92, respectively, and the identification effect is close to or even better than the existing database molecular markers. Therefore, the four new markers can be used for natural killer cell identification research in single-cell transcriptome data analysis. In addition, the constructed machine learning model can also be used to identify natural killer cells in data samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0029] Figure 1 The multi-model receiver operating characteristic curve (ROCCurve) of the molecular marker RNA18SN1 in Example 1 is shown;

[0030] Figure 2 The multi-model receiver operating characteristic curve (ROCCurve) of the molecular marker RNA18SN2 in Example 1 is shown;

[0031] Figure 3 The multi-model receiver operating characteristic curve (ROCCurve) of the molecular marker RNA18SN3 in Example 1 is shown;

[0032] Figure 4 The multi-model receiver operating characteristic curve (ROCCurve) of the molecular marker GZMA in Example 1 is shown; and

[0033] Figure 5 A box plot of the AUC values ​​of the four control molecular marker groups in Example 1 is shown. DETAILED DESCRIPTION

[0034] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] In related single-cell transcriptome studies, researchers need to use published molecular markers in the database to identify natural killer cell populations. However, most of these molecular markers themselves come from independent experiments in different laboratories. There are many of them, their sources are unclear, and their applicability needs to be verified. In short, these identification methods based on prior knowledge are time-consuming and laborious, and there are certain inaccuracies in identification. Therefore, how to use more scientific and efficient methods to systematically screen new natural killer cell molecular markers with strong specificity is a problem that needs to be solved at present.

[0036] As sequencing costs decrease, a large number of peripheral blood transcriptome datasets have appeared in public databases. These unprecedented large-scale datasets provide a data basis for comparing gene expression differences between multiple cell types to more accurately screen natural killer cell molecular markers. At the same time, the application of new classification methods of machine learning allows computers to learn knowledge from data autonomously, which can avoid the influence of human subjective factors and enable the screening and development of new molecular markers to be carried out in a more scientific and accurate manner. Therefore, it is both necessary and feasible to develop new molecular markers for more accurate identification of natural killer cells based on large datasets through machine learning methods.

[0037] Peripheral blood contains a wide variety of blood cell types, of which natural killer cells account for a certain proportion. A large amount of public peripheral blood bulk RNA-Seq transcriptome sequencing data in the database provides an important data basis for analyzing and screening new natural killer cell molecular markers.

[0038] According to a typical embodiment of the present invention, a screening device for natural killer cell markers is provided. The screening device comprises: a comparison unit configured to calculate the gene expression profile of terminally differentiated blood cells in peripheral blood under normal physiological conditions, and to screen out candidate molecular markers by comparing the difference characteristics of the gene expression profiles of natural killer cells and other blood cells; and a screening unit configured to combine the candidate molecular markers screened in the comparison unit with a machine learning algorithm to construct a natural killer cell identification gene model, and screen out natural killer cell markers.

[0039] Here, “normal physiological conditions” are defined as the source of the cell sample being blood cells from a healthy individual.

[0040] By applying the technical solution of the present invention, a large amount of existing public peripheral blood cell data can be directly collected from the database, and new natural killer cell molecular markers with strong specificity can be mined through bioinformatics analysis, without the need for experimental materials and operations, greatly reducing costs and not consuming experimental resources; based on the comparison and integration of multi-cell type differences in large data sets, and combined with machine learning methods for systematic screening, human factors are avoided, specificity and accuracy are optimized and improved, and new molecular markers can be discovered, which can be used for experimental natural killer cell flow sorting, natural killer cell identification and annotation in bioinformatics, and cis-regulatory element mining.

[0041] In a preferred embodiment, other blood cells include B lymphocytes, T lymphocytes, dendritic cells, erythroblasts, monocytes and neutrophils, so as to screen for molecular marker genes specifically expressed by natural killer cells; the machine learning algorithm used in the present invention is preferably a logistic regression model, the results of which are more interpretable and the time cost of model training is less. Of course, similar effects may be achieved through model optimization or the use of other models.

[0042] According to a typical embodiment of the present invention, the raw data in the comparison unit is RNA-Seq data of terminally differentiated blood cells in peripheral blood, for example, bulk RNA-Seq data of terminally differentiated blood cells in peripheral blood in the GEO database. This large amount of peripheral blood transcriptome data set provides a data basis for comparing gene expression differences between multiple cell types to more accurately screen natural killer cell molecular markers. Preferably, the gene expression spectrum of terminally differentiated blood cells in peripheral blood under normal physiological conditions is calculated in the comparison unit, including quantification of the gene expression spectrum of blood cells; more preferably, the gene expression spectrum quantification includes quality control of the original blood cell RNA-Seq sequencing data, including cutting joints, removing low-quality reads, aligning the quality-controlled sequence to the human genome, and filtering out samples with an alignment rate less than a threshold, and using the TPM method for quantification, so as to obtain more accurate results. Typically, the threshold can be 70%.

[0043] In one embodiment, the difference characteristics of the gene expression profiles of natural killer cells and other blood cells obtained by comparison in the comparison unit include performing a differential expression analysis between natural killer cells and other blood cells, and taking the average expression level of the same gene of the same cell type; for natural killer cells, there is a list of differentially upregulated genes compared with other blood cells, and the upregulated differentially upregulated genes specific to the natural killer cell type are obtained by taking the intersection, preferably after removing the marker genes reported in the CellMarker database, as candidate molecular markers for natural killer cells.

[0044] According to a typical embodiment of the present invention, the screening unit is configured to take a single candidate molecular marker as input, and use a machine learning classification model to construct a model to identify natural killer cells in the data sample; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; if the AUC value (AUC value is between 0-1) is greater than 0.85, if it is greater than 0.85, it is considered that the molecular marker identification effect based on the prediction model is good and can be used as a new molecular marker.

[0045] According to a typical embodiment of the present invention, the screening unit also includes an identification module, which is configured to compare and verify the effect of marker identification; the comparison and verification includes: selecting a control gene set, using a machine learning classification model to build a model to identify natural killer cells in data samples; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; by comparing the control gene set with the AUC value of the marker, the identification effect of the marker is verified. In a preferred embodiment of the present invention, the control gene set includes: (1) 10 randomly selected non-natural killer cell molecular marker genes as negative controls; (2) 51 natural killer cell molecular markers that only appear in the CellMarker database; (3) 240 candidate molecular markers after removing the intersection of the CellMarker database; and (4) 27 molecular markers whose up-regulated differential genes have an intersection with CellMarker.

[0046] In fact, the above identification module can also be used as a separate device for the identification of natural killer cell markers. According to a typical embodiment of the present invention, a natural killer cell marker identification device is provided. The identification device is configured to: select a control gene set, use a machine learning classification model to build a model to identify natural killer cells on a data sample; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; by comparing the control gene set with the AUC value of the marker, the identification effect of the marker is verified. In a preferred embodiment of the present invention, the control gene set includes: (1) 10 randomly selected non-natural killer cell molecular marker genes as negative controls; (2) 51 natural killer cell molecular markers that only appear in the CellMarker database; (3) 240 candidate molecular markers after removing the intersection of the CellMarker database; and (4) 27 molecular markers whose up-regulated differential genes have an intersection with CellMarker.

[0047] In a typical embodiment of the present invention, the gene expression profiles of seven terminally differentiated blood cells in peripheral blood under normal physiological conditions are calculated, and the difference characteristics of the gene expression profiles of natural killer cells and other blood cells (B cells, T cells, dendritic cells, erythroblasts, monocytes, neutrophils) are compared, and a natural killer cell identification gene model is constructed in combination with a machine learning algorithm (logistic regression model) to screen out new natural killer cell molecular markers. In this embodiment, it includes:

[0048] (1) Peripheral blood transcriptome data collection

[0049] First, bulk RNA-Seq data of different terminally differentiated blood cells in peripheral blood were retrieved from the GEO database. The seven cell types in this example are: natural killer cell (NK cell), B-lymphocyte (B cell), dendritic cell (DC), erythroblast, monocyte, neutrophil and T-lymphocyte (T cell), in order to obtain the molecular marker genes specifically expressed by natural killer cells. The cell culture method is a culture condition close to the normal physiological state.

[0050] (2) Quantification of gene expression profiles of blood cells

[0051] The original blood cell RNA-Seq sequencing data was quality controlled, including cutting the linker and removing the low-quality reads. The quality-controlled sequences were aligned to the human genome, and samples with an alignment rate of less than 70% were filtered out. The TPM method was used for quantification.

[0052] (3) Screening of candidate molecular markers for natural killer cells

[0053] The differential expression analysis between natural killer cells and other cell types was performed, and the expression of the same gene in the same cell type was averaged. For natural killer cells, there are 6 lists of upregulated differentially expressed genes compared with other cell types. The upregulated differentially expressed genes specific to natural killer cell types were obtained by taking the intersection, and after removing the marker genes reported in the CellMarker database, they were used as candidate molecular markers for natural killer cells.

[0054] (4) Construction of a single candidate marker prediction model

[0055] A single candidate molecular marker was used as input, and a machine learning classification model (logistic regression (LR)) was used to construct models for natural killer cell identification of data samples. The prediction score was automatically evaluated by the model, and the model identification effect was evaluated by the area under the receiver operating characteristic curve (AUC) value. The AUC value ranged from 0 to 1. If it was greater than 0.85, it was considered that the molecular marker identification effect based on the prediction model was good and could be used as a new molecular marker.

[0056] (5) Comparative verification of the identification effect of new markers

[0057] Four control gene sets were selected, and the prediction model was constructed in the same way as in step (4). The identification effect of the new marker was verified by comparing the AUC values ​​of the four gene sets with those of the new marker. The four gene sets selected were: (1) 10 randomly selected non-natural killer cell molecular marker genes as negative controls; (2) 51 natural killer cell molecular markers that only appeared in the CellMarker database; (3) 240 candidate molecular markers after removing the intersection of the CellMarker database; (4) 27 molecular markers that had intersections with CellMarker in the up-regulated differentially expressed genes.

[0058] It can be known from the description of the above implementation mode that a person skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary hardware devices such as detection instruments. Based on such an understanding, the data processing part of the technical solution of the present application can be embodied in the form of a software product, and the computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application or some parts of the embodiments.

[0059] The present application can be used in many general or special computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0060] Obviously, those skilled in the art should understand that some modules or steps of the present application described above can be implemented in a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented with program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0061] In the present invention, new natural killer cell molecular markers mined based on machine learning methods can be used to develop detection products or be applied to cell type annotation research to assist in the identification of natural killer cell types. In the present invention, the new natural killer cell molecular markers RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes screened out have prediction model AUC values ​​of 0.91, 0.92, 0.92, and 0.92, respectively, and the identification effect is close to or even better than the existing database molecular markers. Therefore, the new markers RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes can be used for natural killer cell identification research in single-cell transcriptome data analysis. In addition, the constructed machine learning model can also be used to identify natural killer cells in data samples.

[0062] According to a typical embodiment of the present invention, a marker is provided for use in identifying, enriching or sorting natural killer cells, the marker comprising a gene marker and / or a protein marker, the gene marker being one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein marker being one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes.

[0063] When the application is to identify natural killer cells, the application includes detecting the expression of gene markers, preferably, detecting the expression of gene markers is detecting the mRNA expression or protein expression of gene markers; more preferably, detecting the mRNA expression of gene markers by high-throughput sequencing; optionally, the application is to identify natural killer cells in single-cell transcriptome research. In one embodiment of the present invention, when applied for identification, the expression can be used as an indicator of two-fold difference; for example, in single-cell transcriptome cell annotation analysis, if the average expression of the gene in the candidate cell population is at least greater than or equal to twice the average expression of the gene in other cell populations, the candidate cell population can be identified as natural killer cells.

[0064] In the present invention, natural killer cells derived from a biological sample are identified, enriched or sorted, wherein the biological sample may be selected from one or more of the following: peripheral blood, bone marrow, umbilical cord blood, liver or lymph node.

[0065] Based on the above application, the marker can also be used to prepare various related products. According to a typical embodiment of the present invention, a marker is provided for use in preparing related products for identifying, enriching or sorting natural killer cells. The marker includes a gene marker and / or a protein marker. The gene marker is one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes. The protein marker is one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes.

[0066] When used for preparing related products for identifying natural killer cells, the related products include using a detection reagent to detect the expression level of a gene marker. Preferably, the detection reagent is a detection reagent for detecting the mRNA expression level or protein expression level of the gene marker; more preferably, the detection reagent for detecting the mRNA expression level of the gene marker is a probe and / or primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

[0067] Typically, the product of interest will be a kit.

[0068] Correspondingly, according to a typical embodiment of the present invention, a kit for identifying natural killer cells is provided. The kit includes a detection reagent, which is a detection reagent for detecting the expression of gene markers and / or protein markers, the gene marker is one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein marker is one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes. The detection reagent is a detection reagent for detecting the mRNA expression amount or protein expression amount of the gene marker; preferably, the detection reagent for detecting the mRNA expression amount of the gene marker is a probe and / or a primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

[0069] The beneficial effects of the present invention will be further described below in conjunction with embodiments.

[0070] Example 1

[0071] (1) Peripheral blood transcriptome data collection

[0072] First, bulk RNA-Seq data of seven mature blood cells in peripheral blood were retrieved from the GEO database. The seven cell types in this embodiment are: B lymphocytes, dendritic cells, erythroblasts, monocytes, natural killer cells, neutrophils, and T lymphocytes [Table 1]. The cell culture method is a culture condition close to the normal physiological state.

[0073] (2) Quantification of gene expression profiles of blood cells

[0074] The original blood cell sequencing data was quality controlled, including cutting the junction and removing low-quality reads. The quality-controlled sequences were aligned to the human genome (GRCh38.p13), and samples with higher data quality with an alignment rate greater than 70% were retained. The TPM method was used for quantification, and the formula is as follows:

[0075] TPM=(Ni / Li)*1000000 / (sum(N1 / L1+N2 / L2+N3 / L3+…+Nn / Ln))

[0076] Ni is the number of reads aligned to the i-th gene; Li is the length of the i-th gene; sum(N1 / L1+N2 / L2+...+Nn / Ln) is the sum of the values ​​of all (n) genes after being standardized by length.

[0077] (3) Screening of candidate molecular markers for natural killer cells

[0078] The DESeq2 package (R software package) was used to perform differential expression analysis between natural killer cells and other cell types. The expression of the same gene in the same cell type was averaged and logarithmically transformed. 2 (TPM+1) was used to calculate the fold change (FC). The screening conditions for differentially expressed genes were |FC|>2, p value less than 0.05, and corrected p value less than 0.1. For natural killer cells, there were 6 lists of differentially upregulated genes compared with the other 6 cell types. The upregulated differentially regulated genes specific to natural killer cell types were obtained by taking the intersection. After removing the marker genes reported in the CellMarker database, they were used as candidate molecular markers for natural killer cells.

[0079] (4) Construction of a single candidate marker prediction model

[0080] First, the number of natural killer cell samples was expanded to the same number as the total number of samples of other cell types by random sampling with replacement, and the number of natural killer cell and other cell type samples was 341. Then, the two groups of samples were randomly split into training set and test set at a ratio of 9:1. At this time, the training set included 307 natural killer cell samples and 307 other cell type samples, and the test set included 34 natural killer cell samples and 34 other cell type samples. The prediction model of a single candidate natural killer cell molecular marker was built in the training set. The logistic regression model was randomly repeated 10 times through ten-fold cross validation. The model prediction effect was evaluated in the test set, and the average of the 10 test results was taken. The final AUC value reflects the molecular marker identification effect. The AUC value is between 0-1. If it is greater than 0.85, it is considered that the molecular marker identification effect based on the prediction model is good and can be used as a new molecular marker.

[0081] (5) Four new molecular markers for natural killer cells Based on step (4), the final screening obtained new molecular markers for identifying natural killer cells, RNA18SN1, RNA18SN2, RNA18SN3 and one or more of the GZMA genes [Table 2], with AUC values ​​of 0.91 [ Figure 1 】、0.92

Figure 2

Figure 3

【 Figure 4 Based on the horizontal comparison of the identification effect in step (6), the AUC values ​​of the four markers were higher than the median AUC values ​​of the four control groups, indicating that the four markers RNA18SN1, RNA18SN2, RNA18SN3 and GZMA gene were close to or even higher than the existing marker identification performance.

[0082] (6) Comparative verification of the identification effect of new markers

[0083] Four control gene sets were selected and the prediction model was constructed in the same way as step (4). The identification effect of the new marker was verified by comparing the AUC values ​​of the four gene sets with those of the new marker. Figure 5】. The four gene sets selected are: (1) 10 randomly selected non-natural killer cell molecular markers (Others); (2) 51 natural killer cell molecular markers that only appear in the CellMarker database (CellMarker); (3) 240 candidate molecular markers after removing the intersection of the CellMarker database (Candidate); (4) 27 molecular markers that have intersections with CellMarker (Common). The median AUC values ​​of the four control gene sets were 0.51, 0.59, 0.64, and 0.86, respectively, while the AUC values ​​of the markers RNA18SN1, RNA18SN2, RNA18SN3, and GZMA genes were 0.91, 0.92, 0.92, and 0.92, respectively. Therefore, the four new markers can more accurately identify natural killer cells.

[0084] Table 1. Sources of peripheral blood cell transcriptome datasets

[0085]

[0086] Table 2. Gene information of four new natural killer cell molecular markers

[0087]

[0088] The specific four gene sets are as follows:

[0089] 1) Entrez gene IDs of 10 randomly selected non-natural killer cell molecular markers (Others): 105372114, 5154, 105370627, 100033808, 390081, 27154, 127064, 257160, 57412, 439965;

[0090] 2) Entrez gene IDs of 51 natural killer cell molecular markers (CellMarker) that only appear in the CellMarker database: 483,682,914,915,916,920,924,925,930,960,965,969,974,1043,1524,1604,2207,2215,2353,3003,3133,3559, 3575,3577,3586,3627,3672,3678,3725,3732,3815,4288,5133,5788,6402,6967,7305,7538,8743,8807,9435,9612,10225,10663,10859,11266,22914,27087,51523,126259,445347;

[0091] 3) Entrez gene IDs of 240 candidate molecular markers after removing the intersection of CellMarker database: 154,374,631,894,1015,1339,1390,1521,1946,2149,2793,2999,3001,3324,3337,3384,3458,3802,3804,3805,3809,3811,3812,3823,3887,3892,3904,3982,3990,4603,5148,5243,5436,5521,5551,5645,5729,5740,5874,592 0,6691,6966,6971,6982,6988,7049,7118,7294,7314,8277,8292,8320,8530,8644,8740,8784,9022,9079,9172,9263,9289,9324,9338,9436,951 0,9638,9760,10316,10609,10769,11228,23270,23566,23632,23666,26226,26232,26785,27197,28520,28521,28522,28627,28631,29799,51744 ,53637,54796,54843,54941,55112,55501,55605,55687,56253,57713,57823,60489,64282,79037,79156,79633,79870,79899,80310,80312,8156 3,83875,83930,83988,84099,84525,84632,84969,85390,90011,90102,92211,93010,113802,114801,115653,117157,124221,127435,131583,13 9189,147727,147968,149628,152100,152687,154075,165186,165679,221830,257093,257203,267010,284367,338098,338099,339789,339988,3 43413,374443,374946,389170,390345,391045,391267,391656,392522,400002,401124,402644,431705,432369,494470,642280,643733,643831,644068,646819,648987,653390,653658,728509,728953,729362,729505,767558,100129243,100130581,100130872,100131689,100147744,100151683,100188953,100189123,100287551,100288615,100288798,1002891 33,100379345,100419077,100420062,100505501,100506451,100506569,100506736,100507195,100528032,100652801,101927202,101927369,101927486,101954277,102723099,102724146,105221694,105369656,10537 0276,105370660,105370910,105371291,105371677,105372267,105373467,105374842,105374933,105375130,105377006,105377782,105378347,105379316,106480254,106480437,106631777,106631781,107075115,107 181291,107984023,107984204,107984667,107984889,107985056,107986188,107986676,107986709,107986784,109617003,109864279,109864280,109910379,109910380,112267867,112268042,112268155,114515514;,

[0092] 4) Entrez gene IDs of 27 molecular markers (Common) that overlap with CellMarker: 919,952,3002,3398,3560,3820,3821,3822,3824,4684,4818,6348,6351,6352,6375,6846,6983,7704,9437,10417,10578,11098,11126,28526,51348,83888,259197.

[0093] From the above description, it can be seen that the above embodiments of the present invention achieve the following technical effects:

[0094] 1) Collect a large amount of existing public peripheral blood cell data directly from the database, and mine new and highly specific natural killer cell molecular markers through bioinformatics analysis, without the need for experimental materials and operations, greatly reducing costs and consuming experimental resources;

[0095] 2) Based on the comparison and integration of multi-cell type differences in large data sets, combined with machine learning methods for systematic screening, it avoids human factors, optimizes and improves specificity and accuracy, and can discover new molecular markers. It can be used for experimental natural killer cell flow cytometry sorting, natural killer cell identification and annotation in bioinformatics, and cis-regulatory element mining.

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. Use of a marker in identifying, enriching or sorting natural killer cells, characterized in that: The markers include gene markers and / or protein markers, wherein the gene markers are one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein markers are one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes.

2. The use according to claim 1, characterized in that: When the application is to identify natural killer cells, the application includes detecting the expression level of the gene marker, Preferably, detecting the expression level of the gene marker is detecting the mRNA expression level or protein expression level of the gene marker; More preferably, the mRNA expression level of the gene marker is detected by high-throughput sequencing; Optionally, the application is the identification of natural killer cells in single cell transcriptome studies.

3. Use of the marker in the preparation of products related to identification, enrichment or sorting of natural killer cells, characterized in that: The markers include gene markers and / or protein markers, wherein the gene markers are one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein markers are one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes.

4. The use according to claim 3, characterized in that: When the application is to prepare a product related to identifying natural killer cells, the product includes using a detection reagent to detect the expression level of the gene marker, Preferably, the detection reagent is a detection reagent for detecting the mRNA expression level or protein expression level of the gene marker; More preferably, the detection reagent for detecting the mRNA expression level of the gene marker is a probe and / or a primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

5. A kit for identifying natural killer cells, characterized in that: The kit includes a detection reagent, which is a detection reagent for detecting the expression amount of a gene marker and / or a protein marker, wherein the gene marker is one or more of RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes, and the protein marker is one or more of the protein products expressed by RNA18SN1, RNA18SN2, RNA18SN3 and GZMA genes; Preferably, the detection reagent is a detection reagent for detecting the mRNA expression level or protein expression level of the gene marker; More preferably, the detection reagent for detecting the mRNA expression level of the gene marker is a probe and / or a primer, and further preferably, it is a related reagent for preparing the mRNA of the gene marker into a high-throughput sequencing library.

6. A natural killer cell marker screening device, characterized in that: include: a comparison unit configured to calculate the gene expression profile of terminally differentiated blood cells in peripheral blood under normal physiological conditions, obtain the difference characteristics of the gene expression profiles of natural killer cells and other blood cells by comparison, and screen out candidate molecular markers; The screening unit is configured to combine the candidate molecular markers screened out in the comparison unit with a machine learning algorithm to construct a natural killer cell identification gene model to screen out natural killer cell markers.

7. The device according to claim 6, characterized in that The raw data in the comparison unit is RNA-Seq data of terminally differentiated blood cells in peripheral blood; Preferably, the calculating of the gene expression profile of the terminally differentiated blood cells in the peripheral blood under normal physiological conditions in the comparison unit includes quantifying the gene expression profile of the blood cells; More preferably, the gene expression profile quantification includes quality control of the original blood cell RNA-Seq sequencing data, including cutting the linker, removing low-quality reads, aligning the quality-controlled sequence to the human genome, and filtering out samples with an alignment rate less than a threshold, and quantifying using the TPM method; Preferably, the difference characteristics of the gene expression profiles of natural killer cells and other blood cells obtained by comparison in the comparison unit include performing a differential expression analysis between the natural killer cells and the other blood cells, and taking the average of the expression levels of the same gene of the same cell type; for the natural killer cells, there is a list of differentially upregulated genes compared with the other blood cells, and the natural killer cell type-specific upregulated differentially upregulated genes are obtained by taking the intersection as candidate molecular markers for natural killer cells.

8. The device according to claim 7, characterized in that The screening unit is configured to take a single candidate molecular marker as input and use a machine learning classification model to respectively construct a model to perform natural killer cell identification on the data sample; The prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; if the AUC value is greater than the threshold, the candidate molecular marker is judged as a marker for identifying, enriching or sorting natural killer cells.

9. The device according to claim 7, characterized in that The screening unit further comprises an identification module, wherein the identification module is configured to perform comparative verification on the marker identification effect; the comparative verification comprises: A control gene set is selected, and a model is constructed using a machine learning classification model to identify natural killer cells in data samples; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; and the identification effect of the marker is verified by comparing the AUC value of the control gene set with that of the marker.

10. A natural killer cell marker identification device, characterized in that: The identification device is configured as A control gene set is selected, and a model is constructed using a machine learning classification model to identify natural killer cells in data samples; the prediction score is automatically evaluated by the model, and the model identification effect is evaluated by the area under the receiver operating characteristic curve AUC value; and the identification effect of the marker is verified by comparing the AUC value of the control gene set with that of the marker.

Images