A method for predicting the killing ability of natural killer cells based on peripheral blood transcriptome sequencing data

Abstract

This invention discloses a method for predicting the cytotoxic capacity of natural killer cells (NK cells) based on peripheral blood transcriptome sequencing data, belonging to the field of NK cell cytotoxicity prediction. The method includes: acquiring peripheral blood samples from a training population to determine the true value of NK cell cytotoxicity and performing transcriptome sequencing; using single-cell transcriptome data to screen differentially expressed genes between high and low cytotoxicity groups as a feature gene set; extracting the expression levels of the feature gene set from bulk transcriptome data to form a feature matrix; training a machine learning regression model using the feature matrix as input and the true cytotoxicity value as the target; acquiring bulk transcriptome data of the sample to be tested and inputting it into the prediction model, outputting the predicted cytotoxicity value. This invention eliminates the need for in vitro NK cell expansion and functional experiments, enabling rapid and accurate prediction of NK cell cytotoxicity from peripheral blood transcriptome data, and can be used for screening superior donors for immunotherapy.

Description

Technical Field

[0001] This invention belongs to the field of cell killing ability prediction, and in particular relates to a method for predicting the killing ability of natural killer cells based on peripheral blood transcriptome sequencing data. Background Technology

[0002] Natural killer (NK) cells are an important component of the body's innate immune system, possessing the ability to directly recognize and kill tumor cells and virus-infected cells. In recent years, NK cell-based immunotherapy has been widely applied in cancer treatment and cell therapy. To obtain sufficient quantities and activity of therapeutic NK cells, it is typically necessary to isolate NK cells from donor peripheral blood and expand them in vitro before clinical application. Because NK cells from different donors exhibit significant individual differences in expansion efficiency, functional status, and ultimate killing capacity, accurately evaluating the killing capacity of expanded NK cells is crucial in cell therapy product development and donor selection. Currently, the evaluation of the killing capacity of expanded NK cells mainly relies on in vitro functional assays, such as lactate dehydrogenase release assays, flow cytometry, and fluorescent or luminescent labeled target cell killing assays. All of these methods require first completing NK cell isolation, in vitro expansion culture, and target cell co-culture, and then evaluating the killing activity of NK cells by detecting the target cell mortality rate. In addition, the development of high-throughput sequencing technology in recent years has enabled transcriptomics analysis to be gradually applied to NK cell research. Current studies mainly use single-cell transcriptome sequencing or bulk transcriptome sequencing to analyze the activation status, developmental and differentiation characteristics and functional molecular mechanisms of NK cells, and to screen candidate genes related to cytotoxicity.

[0003] However, the existing technology still has the following shortcomings: (1) The existing NK cell killing ability evaluation methods must complete cell isolation, in vitro expansion culture and killing experiment before functional evaluation results can be obtained. It is a post-evaluation and cannot predict the killing potential of NK cells in advance during the donor screening stage. (2) Due to the significant biological heterogeneity among donors, the killing activity of NK cells from different donors varies greatly after expansion. The existing technology lacks effective prediction methods. It is often necessary to complete expansion culture and functional verification of a large number of donor samples before screening out donors with high killing activity, which leads to a large consumption of cell culture resources, human resources and detection costs. (3) Although single-cell transcriptome sequencing can reveal molecular heterogeneity related to cytotoxic function within NK cells, its detection cost is high and it is difficult to use it directly for large-scale screening. While bulk transcriptome sequencing has the advantages of high throughput and easy access, it is difficult to directly identify key cell subpopulations and characteristic genes that are truly related to cytotoxic function. At present, there is a lack of technical solutions to transform functionally related features at single-cell resolution into predictive models applicable to peripheral blood bulk transcriptomes. Therefore, it is impossible to accurately predict the cytotoxic potential of NK cells using only peripheral blood transcriptome information before amplification culture. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for predicting the killing capacity of natural killer cells based on peripheral blood transcriptome sequencing data, comprising: Peripheral blood samples were obtained from the training population. These samples were used to determine the true value of the killing capacity of the corresponding natural killer cells and to perform transcriptome sequencing to obtain transcriptome expression profiles. Based on the single-cell transcriptome data in the transcriptome expression profile, the training population is divided into a high-killing-ability group and a low-killing-ability group according to the true value of the killing ability, and differentially expressed genes between the high-killing-ability group and the low-killing-ability group are screened to obtain a feature gene set. The expression levels of the characteristic gene set are extracted from the bulk transcriptome data in the transcriptome expression profile to form a feature matrix; Using the feature matrix as input features and the true value of the lethality as the prediction target, a machine learning regression model is trained to obtain a prediction model. Obtain bulk transcriptome data of the peripheral blood sample to be tested, extract the expression levels of the characteristic gene set and input them into the prediction model, and output the predicted value of the natural killer cell killing ability corresponding to the peripheral blood sample to be tested.

[0005] Optionally, obtaining peripheral blood samples from the training group and determining the true value of the corresponding natural killer cell killing ability includes: The peripheral blood samples of the training group were divided into a first sample portion and a second sample portion. The first sample was subjected to natural killer cell sorting and in vitro expansion culture to obtain expanded natural killer cells; The expanded natural killer cells were co-cultured with target cells, the target cell mortality rate was detected, and the specific killing rate was calculated based on the target cell mortality rate. The specific killing rate was used as the true value of the killing ability.

[0006] Optionally, obtaining peripheral blood samples from the training population and performing transcriptome sequencing to obtain transcriptome expression profiles includes: RNA was extracted from peripheral blood samples of the training population and subjected to bulk transcriptome sequencing and single-cell transcriptome sequencing to obtain bulk transcriptome expression profiles and single-cell transcriptome expression profiles.

[0007] Optionally, screening for differentially expressed genes between the high-lethality group and the low-lethality group includes: Dimensionality reduction clustering was performed on the single-cell transcriptome data of the high-killing-ability group and the low-killing-ability group to identify natural killer cell clusters; Differential expression analysis was performed on cells from the high-killing-ability group and the low-killing-ability group in the natural killer cell cluster. Genes that met the preset difference fold and significance threshold conditions were selected as the feature gene set.

[0008] Optionally, extracting the expression levels of the characteristic gene set from the bulk transcriptome data in the transcriptome expression profile includes: Obtain bulk transcriptome sequencing data for each sample in the training population and construct a whole transcriptome gene expression matrix; Based on the characteristic gene set, the expression value of each gene in the characteristic gene set for each sample is extracted from the whole transcriptome gene expression matrix to form the characteristic matrix.

[0009] Optionally, using the feature matrix as input features and the true value of the lethality as the prediction target, a machine learning regression model is trained to obtain a prediction model, including: Using the feature matrix as the independent variable and the true value of the lethality corresponding to each sample as the dependent variable, the elastic network regression algorithm is used to select the optimal regularization parameter through cross-validation and fit the model to obtain the prediction model.

[0010] Optionally, obtaining bulk transcriptome data from the peripheral blood sample to be tested includes: Peripheral blood samples were collected for testing, RNA was extracted, and bulk transcriptome sequencing was performed to obtain the gene expression profile of the samples.

[0011] Optionally, after outputting the predicted value of the natural killer cell killing ability corresponding to the peripheral blood sample to be tested, the method further includes: Based on the comparison between the predicted killing ability value and the preset threshold, it is determined whether the peripheral blood sample to be tested belongs to a sample with high natural killer cell killing ability.

[0012] On the other hand, the present invention also provides an electronic device including a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the processor implements the method when executing the computing program.

[0013] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method.

[0014] Compared with the prior art, the present invention has the following advantages and technical effects: This invention ingeniously combines the functional analysis capabilities of single-cell transcriptomics with the high-throughput and readily available advantages of bulk transcriptomics. It precisely identifies characteristic genes related to cytotoxicity through single-cell sequencing, then transfers these genes to the bulk transcriptome level and integrates them with machine learning algorithms to build a predictive model. Based on this approach, only routine bulk transcriptome sequencing of peripheral blood samples is needed to accurately predict the cytotoxicity of NK cells from which the samples have been expanded in vitro. This eliminates the need for time-consuming NK cell isolation, expansion culture, and cytotoxic function testing, which can take several weeks. Therefore, prospective prediction of cytotoxic potential can be achieved during the donor selection stage, significantly saving cell culture resources and time costs. Experiments have shown that the Pearson correlation coefficient between the predicted and actual values ​​was as high as 0.93 in 73 training samples and 0.68 in 27 independent test samples, with an area under the ROC curve of 0.90. This indicates that the present invention can accurately identify samples with high cytotoxic activity and effectively solves the technical problem that existing technologies cannot predict the cytotoxic potential of NK cells in a high-throughput, accurate, and non-invasive manner before in vitro culture. It provides a rapid and accurate decision-making tool for donor selection and quality control in the field of immune cell therapy. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2 This is a single-cell cluster diagram of PBMCs from different donor sources in an embodiment of the present invention; Figure 3 This describes the differential gene distribution of NK subpopulations in PBMCs from different donor sources in this embodiment of the invention. Figure 4The figure shows the performance of the NK killing ability prediction model based on PBMC transcriptome in an embodiment of the present invention. (A) is the correlation between the actual killing value and the predicted killing value in the training set and the test set, and (B) is the subject operating characteristic curve in the training set and the test set. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0018] Example 1 like Figure 1 As shown, this embodiment provides a method for predicting the killing ability of natural killer cells based on peripheral blood transcriptome sequencing data, including: Peripheral blood samples were obtained from the training population. These samples were used to determine the true value of the killing capacity of the corresponding natural killer cells and to perform transcriptome sequencing to obtain transcriptome expression profiles. Based on the single-cell transcriptome data in the transcriptome expression profile, the training population is divided into a high-killing-ability group and a low-killing-ability group according to the true value of the killing ability, and differentially expressed genes between the high-killing-ability group and the low-killing-ability group are screened to obtain a feature gene set. The expression levels of the characteristic gene set are extracted from the bulk transcriptome data in the transcriptome expression profile to form a feature matrix; Using the feature matrix as input features and the true value of the lethality as the prediction target, a machine learning regression model is trained to obtain a prediction model. Obtain bulk transcriptome data of the peripheral blood sample to be tested, extract the expression levels of the characteristic gene set and input them into the prediction model, and output the predicted value of the natural killer cell killing ability corresponding to the peripheral blood sample to be tested.

[0019] Specifically, it includes: (1) Sample collection and NK cell amplification; Peripheral blood was collected from 73 cells, and peripheral blood mononuclear cells (PNMCs) were separated by Ficoll density gradient centrifugation. Irradiated genetically engineered K562 cells were used as feeder cells. The separated PNMCs were mixed with feeder cells at a 1:1 ratio and co-cultured in complete medium containing 5-10% fetal bovine serum and 500 IU / mL recombinant human IL-2 to expand NK cells. Cells in the culture system were counted every other day, and the same complete medium was added. On day 7, irradiated feeder cells were added again for secondary activation induction. On day 14, all suspended cells in the culture system were harvested. Flow cytometry analysis showed that the purity of CD3⁻CD56⁺ NK cells was greater than 80%.

[0020] (2) Determination of lethality value; Take 1×10⁶ NK cells after expansion in step (1) 5 Using GFP-labeled human erythroleukemia K562 cells as effector cells and GFP-labeled target cells as target cells, effector cells and target cells were mixed at an effector-to-target ratio of 1:4 and seeded into 96-well plates. A negative control group containing only target cells and a maximum release control group treated with 1% Triton X-100 were also included. The 96-well plates were co-cultured at 37°C and 5% CO2 for 6 hours, followed by staining with 7-AAD / PI staining solution. Target cell mortality was measured by flow cytometry. The specific killing rate was calculated using the following formula: (experimental group mortality rate - spontaneous mortality rate) / (maximum mortality rate - spontaneous mortality rate) × 100%. The resulting percentage represents the specific killing rate of the NK cells from which the sample was derived, and this percentage was used as the true killing capacity value of the sample.

[0021] (3) Transcriptome sequencing and gene expression matrix construction; Take 5 × 10⁵ NK cells after expansion in step (1) 5 Total RNA was extracted using the TRIzol method, and RNA integrity was detected using an Agilent 2100 Bioanalyzer. Sequencing libraries were prepared using the Illumina TruSeq chain-specific mRNA library construction kit, and the prepared libraries were sequenced at 150 bp paired ends on a NovaSeq 6000 sequencing platform. Sequencing data were aligned to the human reference genome GRCh38 using STAR software, and read counts and TPM expression levels of each gene were calculated using featureCounts software. Genes with TPM values ​​greater than 1 in at least 20% of the samples were filtered out and retained. The TPM values ​​of the filtered genes were then log2(TPM+1) transformed to obtain the whole transcriptome gene expression matrix.

[0022] (4) Single-cell transcriptome sequencing and characteristic gene screening; Based on the killing power values ​​determined in step (2), the samples were ranked, and the top 10 samples with the highest killing power were selected as the high killing power group, and the bottom 10 samples were selected as the low killing power group. NK cells from the high and low killing power groups, which were amplified and frozen in step (1), were revived, and 10xGenomics single-cell transcriptome sequencing was performed on the revived cells. The sequencing data were processed using CellRanger software and subjected to dimensionality reduction clustering analysis using Seurat software to identify NK cell clusters. Differential expression analysis was performed on NK cells from the high and low killing power groups. The screening criteria were set as |avg_log2FC|>0.5 and the adjusted P value<0.05. Differentially expressed genes meeting the above conditions were selected as the feature gene set.

[0023] (5) Training the prediction model; The whole transcriptome gene expression matrix of the 73 samples in step (3) was used as the feature variable X, and the actual killing power value corresponding to each sample in step (2) was used as the target variable Y. The data were input into the glmnet package in R software. In the gene filtering stage of model construction, only the genes corresponding to the single-cell feature gene set obtained in step (4) were retained. The mixing parameter α of the elastic network was set to 0.5, the response variable family was set to "gaussian", and the regularization parameter λ that minimizes the mean square error was selected by 10-fold cross-validation. The final model was fitted using the selected optimal λ value, and the regression coefficients corresponding to each feature gene were extracted to obtain a predictive model that can predict killing power based on gene expression profile.

[0024] (6) Validation on an independent test set; Another 27 peripheral blood samples were collected and processed using the same NK cell expansion system, cytotoxicity assay method, and transcriptome sequencing procedure as steps (1) to (3) to obtain the gene expression matrix and true cytotoxicity value of the 27 independent samples. The gene expression matrix of the above 27 independent samples was input into the prediction model constructed in step (5) to output the predicted cytotoxicity value corresponding to each sample.

[0025] (7) Results analysis; like Figure 2-4 As shown, the Pearson correlation coefficient between the predicted and measured lethality values ​​of all independent test set samples in step (6) is calculated. Using the median of the measured lethality values ​​of the 27 independent test set samples as the threshold, the 27 samples are divided into a high lethality group and a low lethality group. Receiver operating characteristic curves are plotted using the predicted values ​​to evaluate the model's ability to identify highly lethal samples.

[0026] (8) Application of prediction models; In the donor selection process for NK cell immunotherapy, peripheral blood samples from donors to be evaluated are collected and bulk transcriptome sequencing is performed to obtain the gene expression profile of the sample. Then, the gene expression values ​​corresponding to the characteristic gene set obtained in step (4) are extracted from the gene expression profile. These expression values ​​are input into the prediction model established in step (5), and the model can directly output the predicted killing capacity of NK cells from the donor after standard in vitro expansion. Based on this predicted value, superior donors with killing capacity exceeding a preset threshold are selected for use in the subsequent preparation of NK cell therapy products.

[0027] It should be noted that the detection target of this method is natural killer cells that have been expanded and cultured in vitro, not blood or tissue samples directly collected from the human body for pathological analysis. The entire procedure is completed in vitro (in culture dishes and multi-well plates) and does not involve any interventional treatment on the human or animal body.

[0028] The purpose of this invention is to predict the killing capacity of NK cells for screening superior donors and quality control of cell products. Its direct output is a "cell function score," not a judgment or diagnosis of whether the donor has a certain disease (such as tumor or infection). It belongs to the category of in vitro assessment of cell biological function.

[0029] On the other hand, this embodiment also provides an electronic device, including a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the processor implements the method when executing the computing program.

[0030] On the other hand, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method.

[0031] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting the killing capacity of natural killer cells based on peripheral blood transcriptome sequencing data, characterized in that, include: Peripheral blood samples were obtained from the training population. These samples were used to determine the true value of the killing capacity of the corresponding natural killer cells and to perform transcriptome sequencing to obtain transcriptome expression profiles. Based on the single-cell transcriptome data in the transcriptome expression profile, the training population is divided into a high-killing-ability group and a low-killing-ability group according to the true value of the killing ability, and differentially expressed genes between the high-killing-ability group and the low-killing-ability group are screened to obtain a feature gene set. The expression levels of the characteristic gene set are extracted from the bulk transcriptome data in the transcriptome expression profile to form a feature matrix; Using the feature matrix as input features and the true value of the lethality as the prediction target, a machine learning regression model is trained to obtain a prediction model. Obtain bulk transcriptome data of the peripheral blood sample to be tested, extract the expression levels of the characteristic gene set and input them into the prediction model, and output the predicted value of the natural killer cell killing ability corresponding to the peripheral blood sample to be tested.

2. The method according to claim 1, characterized in that, The process of obtaining peripheral blood samples from the training population and determining the true value of the corresponding natural killer cell killing capacity includes: The peripheral blood samples of the training group were divided into a first sample portion and a second sample portion. The first sample was subjected to natural killer cell sorting and in vitro expansion culture to obtain expanded natural killer cells; The expanded natural killer cells were co-cultured with target cells, the target cell mortality rate was detected, and the specific killing rate was calculated based on the target cell mortality rate. The specific killing rate was used as the true value of the killing ability.

3. The method according to claim 1, characterized in that, The process of obtaining peripheral blood samples from the training population and performing transcriptome sequencing to obtain transcriptome expression profiles includes: RNA was extracted from peripheral blood samples of the training population and subjected to bulk transcriptome sequencing and single-cell transcriptome sequencing to obtain bulk transcriptome expression profiles and single-cell transcriptome expression profiles.

4. The method according to claim 1, characterized in that, Screening for differentially expressed genes between the high-lethality group and the low-lethality group includes: Dimensionality reduction clustering was performed on the single-cell transcriptome data of the high-killing-ability group and the low-killing-ability group to identify natural killer cell clusters; Differential expression analysis was performed on cells from the high-killing-ability group and the low-killing-ability group in the natural killer cell cluster. Genes that met the preset difference fold and significance threshold conditions were selected as the feature gene set.

5. The method according to claim 1, characterized in that, Extracting the expression levels of the characteristic gene set from the bulk transcriptome data in the transcriptome expression profile includes: Obtain bulk transcriptome sequencing data for each sample in the training population and construct a whole transcriptome gene expression matrix; Based on the characteristic gene set, the expression value of each gene in the characteristic gene set for each sample is extracted from the whole transcriptome gene expression matrix to form the characteristic matrix.

6. The method according to claim 1, characterized in that, Using the feature matrix as input features and the true value of the lethality as the prediction target, a machine learning regression model is trained to obtain a prediction model, including: Using the feature matrix as the independent variable and the true value of the lethality corresponding to each sample as the dependent variable, the elastic network regression algorithm is used to select the optimal regularization parameter through cross-validation and fit the model to obtain the prediction model.

7. The method according to claim 1, characterized in that, The acquisition of bulk transcriptome data from the peripheral blood sample to be tested includes: Peripheral blood samples were collected for testing, RNA was extracted, and bulk transcriptome sequencing was performed to obtain the gene expression profile of the samples.

8. The method according to claim 1, characterized in that, After outputting the predicted value of the natural killer cell killing ability corresponding to the peripheral blood sample to be tested, the method further includes: Based on the comparison between the predicted killing ability value and the preset threshold, it is determined whether the peripheral blood sample to be tested belongs to a sample with high natural killer cell killing ability.

9. An electronic device comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, characterized in that, When the processor executes the computing program, it implements the method of any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-8.

Images