A physiological age prediction model based on immune repertoire and transcriptome characteristics and a construction method thereof

Abstract

The application provides a physiological age prediction model based on an immune repertoire and a transcriptome feature and a construction method, and belongs to the technical field of bioinformatics. The physiological age prediction model with high precision and high robustness is constructed by fusing an immune repertoire diversity index and an age-related gene expression profile. The physiological age prediction model can not only accurately evaluate the biological aging degree of an individual, but also effectively distinguish the aging difference under the health and disease states, provides a reliable tool for personalized health management, early warning of disease risk and evaluation of the effect of anti-aging measures, and has important clinical application and health management values.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics technology, and in particular to a physiological age prediction model and construction method based on immune repertoire and transcriptome features. Background Technology

[0002] Biological age, also known as chronological age, is an indicator reflecting an individual's actual degree of aging from the perspective of changes in bodily function, metabolism, and structure. It often differs from chronological age calculated based on birth date. Accurate assessment of biological age is crucial for understanding aging mechanisms, predicting age-related disease risk, and evaluating the effectiveness of interventions. Currently, most mainstream international methods for predicting biological age are based on epigenetic markers, particularly DNA methylation levels. These methods construct "epigenetic clocks" by detecting the methylation status of specific genomic loci, such as the Horvath clock, Hannum clock, and the second-generation clock GrimAge, enabling high-precision age prediction at the population level. However, these methods rely on bisulfite treatment and high-depth sequencing or microarray detection, resulting in high costs. Furthermore, they primarily reflect aging information at the level of genomic epigenetic modifications, failing to fully capture functional changes in the immune system, a core system of aging. Immunosenogenesis is a key component of aging, characterized by decreased T-cell and B-cell receptor diversity, immune cell dysfunction, and chronic low-grade inflammation. Therefore, developing a new method for assessing physiological age that can integrate immune system-specific information, is relatively low-cost, and is easily obtained from routine testing data has become a worthwhile direction to explore in this field. Summary of the Invention

[0003] The purpose of this invention is to provide a physiological age prediction model and construction method based on immune repertoire and transcriptome features, which solves the technical problems of existing physiological age prediction methods that mainly rely on DNA methylation data, are costly, and cannot fully reflect the aging dynamics of the immune system.

[0004] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0005] This invention provides a physiological age prediction model, which is obtained by training a machine learning regression algorithm based on an integrated feature dataset; The integrated feature dataset is composed of immune repertoire diversity indicators and age-related gene expression levels; The immune repertoire diversity indicators include clone count, clone diversity, average clone frequency, geometric mean clone frequency, non-convergent clone diversity, non-convergent clone frequency, average nucleotide length of CDR3 region, average insert length, average NDN region length, convergence coefficient, Shannon diversity index, clone richness, clone evenness, clonality, percentage of IGH-type receptor chains, percentage of IGK-type receptor chains, percentage of IGL-type receptor chains, percentage of TRA-type receptor chains, percentage of TRB-type receptor chains, percentage of TRD-type receptor chains, percentage of TRG-type receptor chains, IGH_gsva score, IGK_gsva score, IGL_gsva score, TRA_gsva score, TRB_gsva score, TRD_gsva score, and TRG_gsva score. The age-related genes include ENSG00000233952, ENSG00000185168, GREM2, GTSCR1, GLB1, HNRNPA1P21, ENSG00000254275, CDCA7L, NRCAM, MTA3, CACHD1, EFNA1, TARBP1, PTK7, PTPRK, SCML1, FBLN2, SATB1, FLNB, ROBO1, TMEM263, PLAG1, ADGRA3, ABLIM1, ZNF69, and N4BP3.

[0006] The above concepts in this invention are explained as follows: Clone count: The total number of clones detected in a sample that have a unique T cell receptor (TCR) or B cell receptor (BCR) sequence.

[0007] Clonal diversity: a comprehensive indicator reflecting the richness of immune cell clones in a sample, usually considering the number of clones and their respective frequencies. A higher value indicates a stronger potential ability of the immune system to recognize antigens.

[0008] Average clone frequency / geometric mean clone frequency: These refer to the arithmetic mean and geometric mean of the proportions of all clones in the sample, respectively, and are used to describe the central tendency of the clone distribution.

[0009] Non-convergent clonal diversity / frequency: "Convergence" refers to the phenomenon where different individuals or clones produce the same or similar antigen-specific receptors. "Non-convergent clones" are the collection of clones after excluding these convergent sequences. This indicator reflects the unique, non-shared immune repertoire characteristics of individuals.

[0010] Average nucleotide length of the CDR3 region: CDR3 (complementarity-determining region 3) is a highly variable core region in the TCR / BCR variable region, directly determining antigen recognition specificity. This indicator refers to the average nucleotide length of the CDR3 coding sequence across all clones.

[0011] Average Insertion Fragment Length / Average NDN Region Length: During TCR / BCR gene rearrangement, nucleotides are randomly inserted or deleted between V, (D), and J gene fragments, forming "N regions" and "D regions" (some have them). This indicator calculates the average length of these non-template inserted nucleotides, reflecting the randomness of the rearrangement process.

[0012] Convergence coefficient / clonal convergence: quantifies the degree to which the same or highly similar CDR3 amino acid sequences appear between different individuals or clones, reflecting the commonality or shared nature of immune responses.

[0013] The Shannon Diversity Index (SDI) is derived from ecology and is a comprehensive indicator used to quantify the richness (number of clones) and evenness (balanced distribution of clones) of clonal species in an immune repertoire. A higher SDI indicates better diversity.

[0014] Clonal richness: Simply put, it refers to the number of unique clones in a sample, without considering the frequency of each clone.

[0015] Clonal evenness: An indicator describing whether different clones are evenly distributed in a sample. High evenness indicates that there are no a few clones over-amplified, and the immune repertoire is more balanced.

[0016] Clonality: In contrast to uniformity, it describes whether the immune repertoire is dominated by a few high-frequency clones. High clonality may indicate recent antigen-driven specific clonal amplification.

[0017] IGH / IGK / IGL receptor chain proportion: B cell receptors (BCRs) consist of a heavy chain (IGH) and a light chain (IGK or IGL). This set of indicators calculates the proportion of receptor clones using IGH, IGK, and IGL genes in the total BCR clones.

[0018] Proportion of TRA / TRB / TRD / TRG receptor chains: T cell receptors are mainly of the αβ type (TRA+TRB chain) and γδ type (TRD+TRG chain). This set of indicators calculates the proportion of each type of chain in the total TCR clone, reflecting the composition of T cell subsets.

[0019] IGH_gsva score / IGK_gsva score / IGL_gsva score / TRA_gsva score / TRB_gsva score / TRD_gsva score / TRG_gsva score: GSVA (Gene Set Variation Analysis) is a method for analyzing gene set variation. These indicators are scores calculated using the GSVA algorithm based on the expression profile activity of genes associated with a specific receptor chain (such as IGH), and are used to quantify the overall activity status of immune pathways or cell populations related to that receptor chain at the transcriptome level.

[0020] The present invention also provides a method for constructing the above-mentioned physiological age prediction model, comprising the following steps: The data acquisition and preprocessing steps involve quality control and adapter removal of the raw RNA sequencing data from peripheral blood to obtain clean data. The immune repertoire analysis and index extraction steps involve assembling and annotating immune receptors in the clean data, and calculating and extracting the diversity index of the immune repertoire. The transcriptome analysis and characteristic gene extraction steps involve comparing the clean data with a reference genome to calculate gene expression levels and extracting the expression levels of age-related genes. The feature fusion and model building steps involve standardizing and fusing the extracted immune repertoire diversity indicators and the expression levels of the extracted age-related genes to form the integrated feature dataset, and using the individual's true age as the response variable, training the physiological age prediction model using a machine learning regression algorithm.

[0021] Preferably, in the immune repertoire analysis and indicator extraction steps, the immune receptors include T cell receptors and B cell receptors.

[0022] Preferably, the machine learning regression algorithm is the random forest algorithm.

[0023] Preferably, the model parameters of the random forest algorithm include a decision tree count of 500 and a feature count of 7 randomly selected when splitting nodes.

[0024] Preferably, in the data acquisition and preprocessing steps, the raw RNA sequencing data comes from the GSE231409 and GSE186507 datasets in the public database GEO.

[0025] This invention also provides a method for predicting physiological age, comprising: For the blood RNA sequencing sample of the individual to be tested, the data acquisition and preprocessing steps, immune repertoire analysis and index extraction steps, and transcriptome analysis and characteristic gene extraction steps in the above physiological age prediction model construction method are performed to extract the corresponding immune repertoire diversity index and the expression level of the age-related gene. The extracted immune repertoire diversity indicators and the expression levels of age-related genes are standardized and fused to form the characteristic data of the individual to be tested. The feature data is input into the physiological age prediction model described above to output the predicted physiological age.

[0026] Preferably, the method further includes a health status assessment step, which calculates the aging acceleration value of the individual to be tested based on the predicted physiological age, wherein the aging acceleration value is the difference between the predicted physiological age and the actual age of the individual to be tested. When the aging acceleration value is significantly greater than zero, the individual being tested is determined to be in a state of accelerated aging. When the method is applied to population assessment, the average predicted physiological age of the disease population is significantly higher than that of the average predicted physiological age of the healthy population in the same time-series age range. The disease population includes individuals with autoimmune diseases, cancer, or chronic inflammatory diseases.

[0027] The present invention also provides a system for predicting physiological age, comprising: The data preprocessing module is used to perform quality control and adapter removal on raw blood RNA sequencing data to obtain clean data; The immune repertoire analysis module is used to perform immune receptor analysis on the clean data and extract the aforementioned immune repertoire diversity indicators. The transcriptome analysis module is used to analyze gene expression in the clean data and extract the expression levels of the aforementioned age-related genes. The feature fusion and model building module is used to fuse the immune repertoire diversity index and the expression data of the age-related genes and train a machine learning regression model to build the physiological age prediction model provided by the present invention. The prediction module is used to call the physiological age prediction model based on the feature data of the new sample to output the predicted physiological age; The machine learning regression model is a random forest model.

[0028] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described physiological age prediction method.

[0029] The beneficial effects of this invention are: This invention constructs a high-precision and robust physiological age prediction model by integrating immune repertoire diversity indicators with age-related gene expression profiles. It can not only accurately assess the biological aging degree of an individual, but also effectively distinguish the aging differences between healthy and disease states. It provides a reliable tool for personalized health management, early warning of disease risks, and evaluation of the effectiveness of anti-aging measures, and has important clinical application and health management value. Attached Figure Description

[0030] Figure 1 This is a heatmap showing the correlation between the 28 immune repertoire diversity indicators used in this invention and the individual's actual time-series age. Figure 2 This is a heatmap showing the correlation between the expression levels of 26 age-related genes used in this invention and the individual's actual chronological age. Figure 3 This is a scatter plot showing the results of the prediction model of this invention predicting physiological age on a test set containing 386 samples. Figure 4 Box plot comparing the predicted physiological age of the disease group and the healthy control group after applying the method of the present invention to predict the physiological age of the two groups. Detailed Implementation

[0031] The technical solutions provided by the present invention will be described in detail below with reference to the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0032] Example This embodiment includes the following steps: 1) Data acquisition and preprocessing: Obtain raw sequencing data of peripheral blood RNA sequencing from GSE231409 and GSE186507 in the public database GEO; perform quality control and adapter removal on the raw data to obtain high-quality clean data.

[0033] 2) Immune repertoire analysis and index extraction: The clean data obtained in step 1) were assembled and annotated using immune repertoire analysis software to assemble T-cell receptors and / or B-cell receptors, and then 28 immune repertoire diversity indicators were calculated and extracted; these 28 indicators are shown in Table 1. Table 1. Detailed information on the indicators

[0034] 3) Transcriptome analysis and feature gene extraction: The clean data obtained in 1) were compared with the reference genome to calculate gene expression levels; gene sets that were significantly related to age in the two datasets were calculated respectively. After threshold screening (pvalue < 0.05) and intersection, the 26 genes with the highest expression levels and temporal age were selected, as shown in Table 2 below, and their standardized expression data were extracted.

[0035] Table 2 Specific Gene Information

[0036] 4) Feature fusion and model building: The expression levels of 28 immune repertoire indicators extracted from S2 and 26 genes extracted from S3 are standardized and fused to form an integrated dataset containing 54 features; the physiological age prediction model is trained using machine learning regression algorithm with the individual's real age as the response variable.

[0037] 5) Physiological age prediction: For a new blood RNA sequencing sample, repeat steps 1) to 4), extract its 54 features, input them into the prediction model trained in step 4), and the output is the predicted physiological age of the individual.

[0038] 6) Health status assessment application: Based on the predicted physiological age obtained in 5), calculate the "accelerated aging value", which is the difference between the predicted physiological age and the actual age; when the accelerated aging value is significantly greater than zero, the individual is determined to be in an accelerated aging state; at the population level, the average predicted physiological age of the disease group (such as patients with autoimmune diseases, cancer or chronic inflammatory diseases) is significantly higher than that of the healthy group in the same time period.

[0039] The machine learning regression algorithm used is random forest.

[0040] The parameters to use are: rfModel<- randomForest(trainclass[["age"]] ~., data = trainMatrix,ntree = 500, mtry = 7, replace = TRUE, nodesize = 5, splitrule = “variance”,importance = FALSE) Based on the above scheme, this embodiment uses GEO public database (https: / / www.ncbi.nlm.nih.gov / geo) data GSE231409 and GSE186507. These data include peripheral blood mRNA sequencing data from cohorts of inflammatory bowel disease (884) and novel coronavirus infection (124), including 654 disease patients and 234 healthy individuals.

[0041] Immunological repertoire data extraction and analysis using raw sequencing data: [The following steps were performed sequentially...] 1) Trimmomatic uses the following parameters: LEADING:3 TRAILING:3 SLIDINGWINDOW:4:20MINLEN:50 TOPHRED33 to remove connectors and filter out unpaired data from the raw data; 2) MixCR extracts CDR3 sequences with the following parameters: -s hsa –-starting-matrial rna 3) After converting the MixCR data using the Convert function of vdjtools, the CalcBasicStats function is used to count the immune repertoire data of each sample, totaling 28 parameters (as mentioned above). Simultaneously, correlation analysis was used to analyze the gene expression data provided by the database. Based on the degree of correlation with age, genes significantly related to age in both datasets were calculated, and their intersection was taken to identify the aforementioned 26 genes highly related to age. Combined with immune repertoire parameters, a total of 54 feature parameters were obtained, such as... Figures 1-2 As shown.

[0042] The 1008 samples were randomly split into a training cohort and a test cohort in a 6:4 ratio. A prediction model was then built using the training cohort dataset and the aforementioned 54 features, employing a random forest algorithm. Predictions were then made on the test cohort data, and the model's stability was evaluated by calculating the error between the predicted and actual ages (RMSE). The overall error was ±9.7, and within the more concentrated 40-60 age range, the error was ±7.4. Figure 3 As shown. We also found that the predicted physiological age of the disease group was significantly higher than that of the healthy control group (as shown). Figure 4 The result (p(t.test) = 0.0087) indicates that this method can effectively reveal an individual's immune aging status and related disease risks.

[0043] This invention constructs a physiological age prediction model based on blood bulk-seq immune repertoire features, representing a pioneering approach in this field. Compared to mainstream DNA methylation clock models, its core advantages lie in: firstly expanding the immunological dimension of physiological age prediction, enabling simultaneous assessment of age prediction with the body's immune function status (clonal diversity, clonal richness, etc.) and gene expression levels; secondly, the detection can directly reuse existing bulk-seq data without requiring special experimental steps such as bisulfite treatment, resulting in strong sample compatibility and lower clinical implementation costs.

[0044] Therefore, it can be concluded that the predictive model constructed based on the above 54 characteristic biomarkers provided by this invention can predict physiological age and effectively reveal an individual's immune aging status and related disease risks, and has broad application prospects in health management, disease early warning and evaluation of anti-aging intervention effects.

[0045] As demonstrated by the above embodiments, this invention provides a feasible scheme for constructing a physiological age prediction model based on blood RNA sequencing data. This model exhibits reliable predictive performance on an independent test set, accurately estimating an individual's physiological age. Further analysis shows that the predicted age calculated using this model can effectively distinguish the differences in aging status between healthy individuals and those with specific diseases; the disease-affected individuals show a significant accelerated aging trend. This confirms that the model of this invention can not only be used for quantitative assessment of physiological age but also has potential application value in identifying disease-related abnormal aging.

[0046] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A physiological age prediction model, characterized in that, The model was obtained by training a machine learning regression algorithm based on an integrated feature dataset. The integrated feature dataset is composed of immune repertoire diversity indicators and age-related gene expression levels; The immune repertoire diversity indicators include clone count, clone diversity, average clone frequency, geometric mean clone frequency, non-convergent clone diversity, non-convergent clone frequency, average nucleotide length of CDR3 region, average insert length, average NDN region length, convergence coefficient, Shannon diversity index, clone richness, clone evenness, clonality, percentage of IGH-type receptor chains, percentage of IGK-type receptor chains, percentage of IGL-type receptor chains, percentage of TRA-type receptor chains, percentage of TRB-type receptor chains, percentage of TRD-type receptor chains, percentage of TRG-type receptor chains, IGH_gsva score, IGK_gsva score, IGL_gsva score, TRA_gsva score, TRB_gsva score, TRD_gsva score, and TRG_gsva score. The age-related genes include ENSG00000233952, ENSG00000185168, GREM2, GTSCR1, GLB1, HNRNPA1P21, ENSG00000254275, CDCA7L, NRCAM, MTA3, CACHD1, EFNA1, TARBP1, PTK7, PTPRK, SCML1, FBLN2, SATB1, FLNB, ROBO1, TMEM263, PLAG1, ADGRA3, ABLIM1, ZNF69, and N4BP3.

2. A method for constructing a physiological age prediction model as described in claim 1, characterized in that, Includes the following steps: The data acquisition and preprocessing steps involve quality control and adapter removal of the raw RNA sequencing data from peripheral blood to obtain clean data. The immune repertoire analysis and index extraction steps include assembling and annotating the clean data with immune receptors, and calculating and extracting the immune repertoire diversity index in the physiological age prediction model of claim 1. The transcriptome analysis and characteristic gene extraction steps involve comparing the clean data with a reference genome to calculate gene expression levels and extracting the expression levels of age-related genes in the physiological age prediction model of claim 1. The feature fusion and model building steps involve standardizing and fusing the extracted immune repertoire diversity indicators and the expression levels of the extracted age-related genes to form the integrated feature dataset. Using the individual's true age as the response variable, a machine learning regression algorithm is used to train and obtain the physiological age prediction model.

3. The construction method according to claim 2, characterized in that, In the immune repertoire analysis and indicator extraction steps, the immune receptors include T cell receptors and B cell receptors.

4. The construction method according to claim 2, characterized in that, The machine learning regression algorithm mentioned is the random forest algorithm.

5. The construction method according to claim 4, characterized in that, The model parameters of the random forest algorithm include 500 decision trees and 7 features randomly selected when splitting nodes.

6. The construction method according to claim 2, characterized in that, In the data acquisition and preprocessing steps, the raw RNA sequencing data comes from the GSE231409 and GSE186507 datasets in the public database GEO.

7. A method for predicting physiological age, characterized in that, include: For the blood RNA sequencing sample of the individual to be tested, the data acquisition and preprocessing steps, immune repertoire analysis and index extraction steps, and transcriptome analysis and characteristic gene extraction steps described in the method for constructing the physiological age prediction model as described in claim 2 are performed to extract the corresponding immune repertoire diversity indicators and the expression levels of the age-related genes. The extracted immune repertoire diversity indicators and the expression levels of age-related genes are standardized and fused to form the characteristic data of the individual to be tested. The feature data is input into the physiological age prediction model as described in claim 1 to output the predicted physiological age.

8. The method according to claim 7, characterized in that, The method further includes a health status assessment step, which calculates the aging acceleration value of the individual to be tested based on the predicted physiological age, wherein the aging acceleration value is the difference between the predicted physiological age and the actual age of the individual to be tested. When the aging acceleration value is significantly greater than zero, the individual being tested is determined to be in a state of accelerated aging. When the method is applied to population assessment, the average predicted physiological age of the disease population is significantly higher than that of the average predicted physiological age of the healthy population in the same time-series age range. The disease population includes individuals with autoimmune diseases, cancer, or chronic inflammatory diseases.

9. A system for predicting physiological age, characterized in that, include: The data preprocessing module is used to perform quality control and adapter removal on raw blood RNA sequencing data to obtain clean data; An immune repertoire analysis module is used to perform immune receptor analysis on the clean data and extract immune repertoire diversity indicators from the physiological age prediction model of claim 1. The transcriptome analysis module is used to analyze gene expression in the clean data and extract the expression levels of age-related genes in the physiological age prediction model of claim 1. The feature fusion and model building module is used to fuse the immune repertoire diversity index and the expression data of the age-related genes and train a machine learning regression model to build the physiological age prediction model as described in claim 1. The prediction module is used to call the physiological age prediction model based on the feature data of the new sample to output the predicted physiological age; The machine learning regression model is a random forest model.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the physiological age prediction method as described in claim 7 or 8.

Images