Kit and system for predicting immune age or storage medium

Abstract

The invention provides a kit and system for predicting immune age or a storage medium, and belongs to the technical field of biological detection. According to the present invention, the 60 specific genes in the peripheral blood initial T cells can be adopted as the biomarkers for immune age prediction for the first time, and the prediction result is high in accuracy, such that the immune function of the body can be evaluated, the health risk and the disease susceptibility of the subject can be predicted, and the personalized intervention guidance can be provided. According to the method, the immune age prediction model based on the initial T cell characteristics is constructed, more accurate and more reliable evaluation on the immune function of the organism is realized, and the method has important significance on health management and has a good application prospect.

Description

Technical Field

[0001] This invention belongs to the field of biological detection technology, specifically relating to a reagent kit, system, or storage medium for predicting immune age. Background Technology

[0002] A healthy immune system is crucial for maintaining bodily health and resisting disease. Immunity is the comprehensive ability of the human immune system to recognize and eliminate foreign pathogens, aging or damaged cells, and mutated cancer cells, thereby maintaining the body's stable health. T cells, as the hub of adaptive immunity, are the most direct and critical "barometer" of the body's immunity. For example, naïve T cells determine the potential to respond to antigens, memory T cells reflect the immune system's "experience" and response speed, and regulatory T cells (Tregs) are related to the balance between immune tolerance and autoimmunity.

[0003] Naïve T cells, serving as a reservoir of T cells, can differentiate into various functionally specific effector cells and memory cells, thereby mediating precise immune responses. Therefore, the quantity and quality of naïve T cells are the cornerstone of maintaining the diversity and responsiveness of the immune system. With aging, the T cell pool changes, one characteristic being a significant reduction in the number and dysfunction of naïve T cells. This decline leads to an inability to effectively respond to neoantigens, weakened immune responses, and ultimately, a decline in the overall function of the immune system and the development of related diseases. Therefore, accurate assessment of immune function is of great significance for health management, disease early warning, and evaluation of intervention effectiveness.

[0004] Currently, there is a lack of accurate and comprehensive assessment standards and tools for "immunity," and existing methods have significant limitations. There are techniques for predicting and assessing immune function based on overall peripheral blood mononuclear cells or multiple immune cell subsets (References: 1. Nat Immunol. 2025 Feb; 26(2): 308-322. doi: 10.1038 / s41590-024-02059-6; 2. NatAging. 2025 Apr; 5(4): 607-621. doi: 10.1038 / s43587-025-00819-z). However, given the extreme complexity of T cell populations and functions, these techniques suffer from insufficient predictive specificity, sensitivity, and stability, making it difficult to accurately capture the core drivers of immune system dysfunction.

[0005] Multiple cutting-edge studies have shown that the state of naive T cells is a better indicator of changes in the immune system during aging compared to other T cell subtypes. This characteristic makes it a highly promising biomarker, contributing to the development of more accurate technologies for predicting immune function.

[0006] Immune age is a biological indicator that measures the functional state of an individual's immune system, reflecting its difference from chronological age. It assesses the "aging" of the immune system by evaluating factors such as immune cell activity, inflammation levels, and disease resistance, thereby predicting health risks and disease susceptibility. This information can then guide personalized interventions to safeguard human health. Currently, there are no reports of predicting immune age based on gene expression levels in naive T cells. Summary of the Invention

[0007] The purpose of this invention is to provide a kit, system, or storage medium for predicting immune age. This invention aims to accurately predict immune function by analyzing the transcriptional state of naive T cells in peripheral blood. This is beneficial for identifying early decline in the immune system, thereby facilitating early preventative measures to reduce the risk of disease and death.

[0008] This invention provides the use of reagents for detecting gene expression levels in the preparation of kits for predicting immune age, wherein the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, and CYSLTR1. At least one of the following: LDOC1, CPQ, ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1, or SELENOM.

[0009] Further, the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, and A. NXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 and SELENOM.

[0010] Furthermore, the gene is a gene found in naïve T cells; And / or, the reagent for detecting gene expression levels is a qPCR detection reagent, a bulk-RNA detection reagent, or a reagent for single-cell sequencing methods; Preferably, the gene is a gene found in naive T cells of peripheral blood.

[0011] This invention also provides a kit for predicting immune age, comprising reagents for detecting gene expression levels, wherein the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, and CYSLTR1. At least one of the following: LDOC1, CPQ, ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1, or SELENOM.

[0012] Further, the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, and A. NXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 and SELENOM.

[0013] Furthermore, the gene is a gene found in naïve T cells; And / or, the reagent for detecting gene expression levels is a qPCR detection reagent, a bulk-RNA detection reagent, or a reagent for single-cell sequencing methods; Preferably, the gene is a gene found in naive T cells of peripheral blood.

[0014] The present invention also provides a system for predicting immune age, the system comprising the following parts: The input module is used to input gene expression levels, where the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, and CP. Q, ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 or SELENOM at least one of the following: The prediction module is used to input the gene expression level as a feature into the model to obtain the immune age prediction result; The output module is used to output the immune age prediction results.

[0015] Furthermore, in the input module, the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, and CP. Q, ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP 2. CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 and SELENOM; And / or, in the prediction module, the model is an elastic network regression model; Preferably, the elastic network regression model is corrected by LOESS local weighted regression.

[0016] Furthermore, the gene is a gene found in naïve T cells; And / or, the reagent for detecting gene expression levels is a qPCR detection reagent, a bulk-RNA detection reagent, or a reagent for single-cell sequencing methods; Preferably, the gene is a gene found in naive T cells of peripheral blood.

[0017] The present invention also provides a computer-readable storage medium having a computer program stored thereon for implementing the aforementioned system.

[0018] The present invention has achieved the following beneficial effects: This invention is the first to discover that 60 specific genes in naive peripheral blood T cells can be used as biomarkers for predicting immune age with high accuracy. This allows for the assessment of the body's immune function, prediction of health risks and disease susceptibility, and provision of personalized intervention guidance. This invention constructs an immune age prediction model based on naive T cell characteristics, achieving a more accurate and reliable assessment of the body's immune function. This has significant implications for health management and promising application prospects.

[0019] The key to this invention lies in identifying 60 genes (GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, ANXA1, AHNAK, RASGRP2, AP000787.1, KL) in naive peripheral blood T cells. The combination of 60 genes (F6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1, and SELENOM) is significantly associated with immune age. Therefore, the immune age of a subject can be predicted by detecting the expression levels of these 60 genes in naive T cells in peripheral blood, thus providing a more accurate and reliable assessment of the subject's immune function. Various methods currently available in the art can be used to specifically detect the expression levels of these genes.

[0020] Obviously, based on the above description of the present invention, and according to common technical knowledge and conventional methods in the field, various other modifications, substitutions or alterations can be made without departing from the basic technical concept of the present invention.

[0021] The following detailed embodiments further illustrate the above-described content of the present invention. However, this should not be construed as limiting the scope of the present invention to the following embodiments. All technologies implemented based on the above-described content of the present invention fall within the scope of the present invention. Attached Figure Description

[0022] Figure 1 A flowchart for constructing an immune age prediction model.

[0023] Figure 2 This is a graph showing the Pearson correlation analysis results between the actual age and the predicted age from the elastic network regression model.

[0024] Figure 3 The graph shows the results of the elastic network regression model with systematic errors.

[0025] Figure 4 The figure shows the results of the n-TIME model constructed using Loess's locally weighted regression to correct the error.

[0026] Figure 5 The graph shows the Pearson correlation analysis results of the n-TIME model on the training dataset for the true age and predicted age.

[0027] Figure 6 The graph shows the Pearson correlation analysis results of the n-TIME model on the test dataset for the true age and predicted age.

[0028] Figure 7 The figure shows the Pearson correlation analysis results of the n-TIME model on the validation dataset, comparing the true age and the predicted age.

[0029] Figure 8 This is a graph showing the comparison between the actual age and predicted age of the n-TIME model in a dataset of centenarians.

[0030] Figure 9 This is a graph showing the comparison between the true age and predicted age of the n-TIME model in chronic infection datasets (HIV and HBV).

[0031] Figure 10 This is a graph showing the comparison between the true age and predicted age of the n-TIME model in autoimmune disease datasets (SLE and MG).

[0032] Figure 11The figure shows the analysis results of an immune age prediction model constructed separately using the same method as immune-related genes. Detailed Implementation

[0033] Unless otherwise specified, the experimental methods used in the embodiments are conventional methods; the materials and reagents used are commercially available unless otherwise specified. Algorithms for data acquisition, transmission, storage, and processing steps not specifically described in the embodiments, as well as hardware structures and circuit connections not specifically described, can be implemented using publicly available information in the prior art.

[0034] To facilitate understanding, the terminology is explained as follows: PBMCs: Peripheral blood mononuclear cells.

[0035] Reads refer to the short nucleotide sequences that the sequencer reads from both ends of a cDNA fragment during RNA sequencing, which are used for subsequent splicing, alignment, and quantitative analysis.

[0036] scRNA-seq: Single-cell RNA sequencing is RNA sequencing performed at the single-cell level. It aims to analyze the heterogeneity in a cell population and reveal the unique gene expression characteristics of different cell types. Its advantages are that it can analyze cellular heterogeneity and is costly.

[0037] Bulk RNA-seq is a method that uses high-throughput sequencing technology to perform qualitative and quantitative analysis of all RNA molecules in a population of cells to reveal gene expression levels in different physiological or pathological states. Its advantages are high throughput and low cost.

[0038] Example 1: Construction of the Immune Age Prediction Model of the Present Invention The flowchart of the construction of the immune age prediction model of this invention is as follows: Figure 1 As shown. The method for constructing this immune age prediction model includes, but is not limited to, the following steps: I. Acquisition of peripheral blood naive T cell transcription data Venous blood samples were collected from healthy individuals of different ages with normal vital signs and no chronic diseases, and stored at room temperature using EDTA-anticoagulated tubes. Peripheral blood mononuclear cells (PBMCs) were isolated from the anticoagulated peripheral blood using density gradient centrifugation. In brief, diluted blood was carefully added to the supernatant of human lymphocyte separation medium, centrifuged, and PBMCs located in the white blood cell layer were aspirated. Red blood cells were lysed using lysis buffer, followed by multiple washes to obtain highly purified cells. This invention proposes two sequencing strategies to obtain transcriptional data of naive peripheral blood T cells, addressing the limitations of currently available commercial transcriptome sequencing methods.

[0039] 1. Obtaining transcriptional data of naive peripheral blood T cells based on scRNA-seq data 1.1 Sequencing For the obtained PBMCs, scRNA-seq was first performed using the 10× Genomics platform. First, individual cells were encapsulated in oil droplets with gel beads bearing unique barcodes, allowing for independent labeling of the mRNA in each cell. Then, the oil droplets were broken, and the cDNA libraries containing the cell barcodes and unique mRNA molecular identifiers were mixed and analyzed using high-throughput sequencing technology.

[0040] 1.2 Quality Control and Comparison (1) By using the mkfastq command of the 10×Genomics official software cellranger, the original scRNA-seq offline data (BCL file) was demultiplexed into Fastq files for each sample.

[0041] (2) The fastq file was filtered by fastp software to remove the adapters and reads containing N. Bases with a quality value lower than 5 were defined as low quality bases, and paired reads with a low quality base ratio of more than 20% were removed to obtain clean data that could be further analyzed.

[0042] (3) The high-quality sequencing reads were aligned to the human GRCh38 reference genome using the count command of the cellranger software and the STAR aligner. Each gene in each cell was counted using cell barcodes and mRNA unique molecular identifiers, and finally a gene × cell expression matrix was generated.

[0043] 1.3 Collection of public data on scRNA-seq of PBMCs from healthy individuals To reduce the sequencing costs required to construct an immune age prediction model and to increase the robustness of the model, six independent scRNA-seq data cohorts of healthy individuals' PBMCs were collected from public databases (data cohorts were obtained from the following sources: DOI: 10.1038 / s43587-022-00198-9; DOI: 10.1126 / sciadv.abq7599; DOI: 10.1038 / s41590-024-02059-6; DOI: 10.1016 / j.cell.2021.02.018; DOI: 10.1016 / j.cell.2021.01.053; DOI: 10.1016 / j.cell.2025.02.017).

[0044] 1.4 Data Preprocessing and Initial T Cell Identification (1) The gene × cell expression matrix obtained from sequencing and public data was merged. The merged expression matrix was quality controlled to filter out low-quality cells (high proportion of mitochondrial genes or low number of genes detected) and low-expression genes. Then, standardization was performed to eliminate technical noise and differences in sequencing depth between cells. Highly variable genes were screened based on residuals and variance for downstream analysis.

[0045] (2) Principal component analysis (PCA) was used to linearly reduce the dimensionality of the data, retaining significant principal components to capture core biological variations. Then, nonlinear dimensionality reduction methods (such as t-SNE or UMAP) were used to visualize the transcriptomic similarity between cells in two-dimensional space. Based on the spatial distance after dimensionality reduction, graph clustering algorithms (such as the Louvain algorithm) were applied to divide the cells into several subgroups with similar transcriptomic characteristics.

[0046] (3) Identify the marker genes in each cell subpopulation by rank-sum test or t-test, and identify the cell subpopulation that highly expresses CD3, TCF7, CCR7, LEF1 genes and does not express or low expresses AHNAK, ANXA2, FAS, PRDM1 genes as naïve T cells.

[0047] (4) The average gene expression level of the initial T cells in each sample is obtained by averaging the normalized gene × cell expression matrix of each sample’s initial T cells.

[0048] 2. Obtaining transcriptional data of naive peripheral blood T cells based on bulk RNA-seq data 2.1 Enrichment of naive T cells PBMCs were surface-labeled and stained using a specific antibody combination (CD3, CD45RA, CD45RO, CCR7), and CD3 was precisely obtained using a flow cytometry sorting strategy. + CD45RA + CD45RO - CCR 7+ The initial T cell population in peripheral blood.

[0049] 2.2 Sequencing For the initial T cells obtained, total RNA was first extracted, and then oligo(dT) magnetic beads were used to enrich mRNA containing polyA tails to obtain complete transcriptome information. Next, the mRNA was fragmented and used as a template to synthesize double-stranded cDNA. After end repair, sequencing adapters were ligated to both ends of the cDNA fragments to construct a library suitable for high-throughput sequencing. Finally, large-scale parallel sequencing was performed on a sequencer.

[0050] 2.3 Data Preprocessing The raw BCL data files were converted to FastQ format using the bcl2fastq2 software. Sequencing quality was assessed using tools such as FASTP, and reads were filtered to remove adapter sequences, reads containing N, and reads below Q20. Subsequently, the filtered high-quality data were aligned and mapped to the GRCh38 human reference genome using STAR software. The number of reads aligned to each gene was counted using featureCounts software, generating a raw gene expression count matrix. The raw count matrix was then normalized to obtain the initial T cell gene expression levels for each sample.

[0051] II. Construction of an immune assessment model based on naive T cells (n-TIME model) After processing, the initial T-cell gene × sample expression matrix of healthy individuals contained 7 cohorts and 639 samples, with an age range of 0-97 years. Two independent cohorts were divided into two datasets: 70% for training and 30% for testing (data cohorts from the following sources: DOI:10.1016 / j.cell.2025.02.017; DOI:10.1038 / s41590-024-02059-6). The remaining cohorts were used as a validation dataset for independent healthy individuals.

[0052] 1. Preliminary screening of age-related genes To verify the linear association between initial T cell gene expression and subject age, a linear regression model was constructed using a linear modeling program, with gene expression level as the dependent variable and age as the independent variable. This step was implemented by calling the lm() function in the R language computing environment, and its core code is as follows: model<- lm(gene ~ age, data = data) The data contains gene expression and age information from the training dataset.

[0053] This function uses Ordinary Least Squares (OLS) as the parameter estimation criterion by default. By minimizing the sum of squared residuals, it automatically calculates the best-fit parameters of the model, namely the intercept and slope of the regression line (the coefficients of the gene variable). After the function is executed, the generated model object containing all the fitting results is stored in the variable model.

[0054] Subsequently, detailed summary information of the model can be obtained by executing the `summary(model)` command, including but not limited to: estimated values ​​of regression coefficients and their standard errors, t-statistics, and significance p-values. This invention identifies genes with absolute regression coefficient values ​​> 0.001 and multiple-correction p-values ​​< 0.01 as age-related genes.

[0055] 2. Constructing an immune age prediction model through gene combination Based on the initially screened age-related genes, an elastic regression algorithm combining L1 (Lasso) and L2 (Ridge) regularization was employed. K-fold cross-validation was used to automatically optimize the model's key hyperparameters to prevent overfitting and improve its generalization ability, thereby constructing a high-precision and robust prediction model. This process was implemented using the glmnet package in R, and the core code is as follows: # Optimize regularization parameter lambda based on cross-validation cv_model<- cv.glmnet(x = tran_x, y = age, alpha = 0.5) # Construct the final prediction model using optimal parameters en_model<- glmnet(x = train_x, y = age, alpha = 0.5, lambda = cv_model$lambda.min) Where train_x represents gene expression information from the training dataset, and age represents age information from the training dataset. The above code performs the following operations: (1) Hyperparameter optimization (cross-validation): Call the cv.glmnet() function. Here, the parameter alpha = 0.5 is a pre-defined elastic network mixing parameter used to control the ratio of L1 to L2 regularization (alpha=1 for Lasso regression, alpha=0 for ridge regression, and the intermediate value is a mixture of the two). This function uses 10-fold cross-validation by default, automatically calculating and evaluating the model performance of a series of different regularization strength parameters lambda for a given alpha value. It uses mean squared error as the evaluation criterion by default, and finally selects the optimal value lambda.min from all lambda values ​​to minimize the cross-validation error.

[0056] (2) Final model training: Call the glmnet() function, using the same mixing parameter alpha as in the previous steps, and explicitly specify the regularization parameter lambda as the optimal value obtained from cross-validation, cv_model$lambda.min. This function fits the final elastic network regression model en_model based on all training data and this optimal hyperparameter combination.

[0057] Using a resilient network regression model, the following 60 genes were selected as a key biomarker combination for predicting immune age: GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK , SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CR IP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1, SELENOM.

[0058] The immune age of a healthy population with initial T cells was predicted using a resilient network regression model. Pearson correlation analysis was performed on the true and predicted ages, and the results showed a Pearson correlation coefficient (PCC) of 0.88, indicating good model performance. Figure 2 ).

[0059] 3. System error correction for elastic network regression models To further improve the accuracy of the prediction model, the systematic prediction bias exhibited by the aforementioned elastic network regression model (en_model) on the training dataset is addressed. Figure 3(That is, for samples with small true values, the predicted values ​​tend to overestimate; for samples with large true values, the predicted values ​​tend to underestimate.) Therefore, a post-processing correction method based on Locally Weighted Regression (LOESS) is introduced. This method can effectively identify and fit the nonlinear relationship between prediction bias and true values, thereby accurately compensating for the initial predicted values. The specific correction process is implemented using the following R language code: # Calculate the difference between the initial predicted age and the actual age (i.e., prediction bias). age_df$age_diff<- age_df$pred_age - age_df$true_age # Modeling the relationship between prediction bias and true age using LOESS regression loess_m<- loess(age_diff ~ true_age, data = age_df, span = 0.75,control = loess.control(surface = "direct")) # Predict the bias value for each sample using the fitted LOESS model and correct the initial predicted age. age_df$pred_adj<- age_df$pred_age- predict(loess_m, age_df$true_age) Here, `age_df` is a DataFrame object containing information about the predicted age (`pred_age`) and the true age (`true_age`). `age_diff` is the difference between the predicted age and the true age. `pred_adj` is the predicted age corrected using the LOESS method.

[0060] The above code performs the following core operations: (1) Deviation calculation: First, create a new variable age_diff in the data frame age_df. Its value is the difference between the predicted value pred_age of the initial model and the true observation value true_age. This difference directly quantifies the prediction deviation of the model.

[0061] (2) Deviation Relationship Modeling: The `loess()` function is called to build a nonparametric regression model to explore and fit the potential relationship between the predicted deviation `age_diff` and the true value `true_age`. Key parameters are set here: span = 0.75: This parameter controls the smoothness of the regression and determines the proportion of data points used when fitting the curve. The default value is usually 0.75, which is used to balance the flexibility and smoothness of the model.

[0062] `control = loess.control(surface = "direct")`: This parameter controls how the fitted surface is calculated. The default method is "interpolate", while this invention explicitly specifies "direct" to ensure accurate calculations are performed directly on all data points. Although this increases the computational load, it yields a more accurate fitted model, suitable for subsequent accurate predictions.

[0063] (3) Bias correction: Using the trained LOESS model loess_m, predict the corresponding systematic bias based on the true value true_age of each sample. Finally, subtract the predicted bias from the initial predicted value pred_age to obtain the corrected and more accurate predicted value pred_adj.

[0064] Through this processing step, the present invention successfully eliminates the systematic bias in the prediction results of the elastic network regression model, and significantly improves the prediction accuracy and consistency of the model across the entire data range. Figure 4 This invention refers to the elastic network regression model corrected by LOESS locally weighted regression as the n-TIME model.

[0065] Example 2: Performance Evaluation of the n-TIME Immune Age Prediction Model 1. Evaluation of the generalization ability of the n-TIME model To evaluate the performance of the n-TIME model, Pearson correlation analysis was performed on the true and predicted ages of the initial T cell datasets from healthy individuals, the test datasets, and the independent validation datasets. The results showed that the Pearson correlation coefficients (PCCs) were 0.94, 0.92, and 0.93, respectively, indicating that the model has good generalization ability. Figure 5-7 ).

[0066] In addition, we combined our PBMCs scRNA-seq data from healthy centenarians with data collected from public databases, as well as PBMCs scRNA-seq data from healthy centenarians, those with chronic infections (including HIV and HBV), and those with autoimmune diseases (including SLE and MG), to serve as a validation cohort for the n-TIME model. (The validation cohorts were obtained from the following publications: DOI: 10.1073 / pnas.1907883116; DOI: 10.1038 / s43587-022-00198-9; DOI: 10.1172 / jci.insight.176208; DOI: 10.1016 / j.scib.2025.03.038; DOI: 10.1038 / s41590-020-0743-0; DOI: 10.3389 / fimmu.2024.1382320). The results showed that the immune age of centenarians was younger than their actual age, while the predicted age of patients with chronic infections and autoimmune diseases was older than their actual age. Figure 8-10 The results are consistent with the immune function status of the corresponding cohort population, indicating that n-TIME can accurately predict the acceleration and delay of immune age and the status of immune function in different populations.

[0067] 2. Constructing an immune age prediction model based on immune-related genes In this invention, the genes LRRN3, S100A4, GBP5, and ANXA1 in the gene combination are highly correlated with immunity. Following the model construction method described above, an immune age prediction model was constructed using these single genes, and then used to predict the immune age of healthy individuals. The results showed that the immune age prediction model constructed using single genes performed poorly, with a Pearson correlation coefficient (PCC) of only 0.37-0.53 for predicting the immune age of healthy individuals. Figure 11 This demonstrates that a single gene cannot accurately and completely capture changes in immune function during the aging process, thus failing to construct an effective immune age prediction model.

[0068] In summary, this invention is the first to discover that 60 specific genes in naive peripheral blood T cells can be used as biomarkers for predicting immune age with high accuracy. This allows for the assessment of the body's immune function, prediction of health risks and disease susceptibility in individuals, and provision of personalized intervention guidance. This invention is of great significance for health management and has promising application prospects.

Claims

1. The use of a reagent for detecting gene expression levels in the preparation of a kit for predicting immune age, characterized in that: The genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP , IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, ANXA1, At least one of the following: AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1, or SELENOM.

2. The use according to claim 1, characterized in that: The genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NRE P, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, ANXA 1. AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, C RIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 and SELENOM.

3. The use according to claim 1 or 2, characterized in that: The gene in question is a gene found in naïve T cells; And / or, the reagent for detecting gene expression levels is a qPCR detection reagent, a bulk-RNA detection reagent, or a reagent for single-cell sequencing methods; Preferably, the gene is a gene found in naive T cells of peripheral blood.

4. A kit for predicting immune age, characterized in that: It includes reagents for detecting gene expression levels, wherein the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, and CPQ. At least one of the following: ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1, or SELENOM.

5. The reagent kit according to claim 4, characterized in that: The genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NRE P, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, ANXA 1. AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, C RIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 and SELENOM.

6. The kit according to claim 4 or 5, characterized in that: The gene in question is a gene found in naïve T cells; And / or, the reagent for detecting gene expression levels is a qPCR detection reagent, a bulk-RNA detection reagent, or a reagent for single-cell sequencing methods; Preferably, the gene is a gene found in naive T cells of peripheral blood.

7. A system for predicting immune age, characterized in that: The system includes the following components: The input module is used to input gene expression levels, where the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, and CP. Q, ANXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 or SELENOM at least one of the following: The prediction module is used to input the gene expression level as a feature into the model to obtain the immune age prediction result; The output module is used to output the immune age prediction results.

8. The system according to claim 7, characterized in that: In the input module, the genes are GBP5, S100A4, TPR, AL355472.1, ID2, CFAP36, UGP2, AC133644.2, ARL6IP5, OCIAD2, NAAA, FAM13A, PIK3R1, NREP, IK, SQSTM1, PI16, CCND3, C7orf50, BZW2, NRCAM, LRRN3, HIPK2, CA5B, DYNLT3, PLP2, EDA, CYSLTR1, LDOC1, CPQ, and A. NXA1, AHNAK, RASGRP2, AP000787.1, KLF6, EIF4EBP2, RGS10, PTGES3, C12orf75, TRAV8-4, PTGER2, DHRS7, GPHN, CRIP2, CRIP1, GCHFR, BCL2A1, MAZ, CIAPIN1, WBP2, CD7, GTSCR1, ERGIC3, TSHZ2, ZBP1, CNN2, ADAMTS10, FXYD5, XBP1 and SELENOM; And / or, in the prediction module, the model is an elastic network regression model; Preferably, the elastic network regression model is corrected by LOESS local weighted regression.

9. The system according to claim 7 or 8, characterized in that: The gene in question is a gene found in naïve T cells; And / or, the reagent for detecting gene expression levels is a qPCR detection reagent, a bulk-RNA detection reagent, or a reagent for single-cell sequencing methods; Preferably, the gene is a gene found in naive T cells of peripheral blood.

10. A computer-readable storage medium, characterized in that: It stores a computer program for implementing the system described in any one of claims 7 to 9.

Images