Construction and application of a model for early risk warning of latent tuberculosis infection

Abstract

The application discloses a kind of early risk warning model of latent tuberculosis infection and its application.The specific disclosure of the expression level of the substance for detecting biomarker is applied in any one of the following:A1) in the preparation for the product for early risk warning of latent tuberculosis infection;A2) in the construction of early risk warning model of latent tuberculosis infection;The biomarker is ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene and ANKRD22 gene.The biomarker of the application and the model constructed using the biomarker are accurate and reliable by ROC curve verification, can provide decision support for doctors, guide the selection and adjustment of treatment plan, provide personalized prevention and treatment suggestions for patients, or early intervention treatment for high-risk patients, etc., realize the best disease management, and have wide clinical application value.

Description

Technical Field

[0001] This invention relates to the fields of bioinformatics and biomedicine, specifically to the construction and application of an early risk warning model for latent tuberculosis infection. Background Technology

[0002] Tuberculosis (TB) is primarily caused by infection with Mycobacterium tuberculosis (MTB). Early identification and treatment of latent tuberculosis infection (LTBI) are particularly important.

[0003] Latent tuberculosis infection (LTBI) refers to a state in which the body produces a sustained immune response to MTB antigen stimulation, but without obvious clinical symptoms or imaging evidence. Individuals with latent tuberculosis infection are not infectious and do not exhibit active tuberculosis symptoms. Without effective intervention, these infected individuals may develop active tuberculosis in the future due to factors such as weakened immunity, becoming new sources of infection. Prophylactic treatment for individuals with latent tuberculosis infection is an important means of reducing the risk of disease and blocking transmission, and is also one of the key preventive measures in the global strategy to end the tuberculosis epidemic. Currently, there is no gold standard for the diagnosis of LTBI, relying mainly on the tuberculin skin test (TST) and the interferon-gamma release assay (IGRA). However, these methods all have certain limitations. Although various tuberculosis diagnostic kits have been introduced globally in recent years, these methods still have limitations; they cannot completely distinguish between ATB and LTBI, nor can they predict whether LTBI patients will progress to ATB. Therefore, finding biomarkers that can effectively distinguish between LTBI and ATB is of great significance for the early detection and preventive treatment of tuberculosis.

[0004] Therefore, early warning and diagnosis of latent tuberculosis infection, along with early intervention and treatment, are crucial strategies for reducing the incidence and mortality of active tuberculosis. Currently, there is a lack of rapid and efficient methods for early risk warning of latent tuberculosis infection. The rapid development of artificial intelligence and multi-omics technologies offers new means to address this challenge. Summary of the Invention

[0005] The purpose of this invention is to provide a sensitive, specific, and accurate early risk warning model for latent tuberculosis infection and its application.

[0006] To achieve the above objectives, the present invention first provides the use of a substance for detecting the expression level of a biomarker in any of the following: A1) Application in the preparation of products for early risk warning of latent tuberculosis infection; A2) Application in constructing an early risk warning model for latent tuberculosis infection; The biomarkers may be the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

[0007] In the above applications, the substance may include reagents for detecting the expression levels of the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

[0008] Furthermore, the substance may include reagents for detecting the expression levels of the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene by reverse transcription-polymerase chain reaction (RT-PCR), reverse transcription-qPCR, and transcriptome sequencing technology (RNA-seq).

[0009] In the above applications, the reagents may include primers for specifically amplifying the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene and ANKRD22 gene and / or probes for specifically recognizing the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene and ANKRD22 gene.

[0010] Furthermore, the primers include primer pair 1 for detecting or specifically amplifying the ID3 gene, primer pair 2 for detecting or specifically amplifying the PSTPIP2 gene, primer pair 3 for detecting or specifically amplifying the VAMP5 gene, primer pair 4 for detecting or specifically amplifying the GBP5 gene, primer pair 5 for detecting or specifically amplifying the BATF2 gene, and primer pair 6 for detecting or specifically amplifying the ANKRD22 gene.

[0011] Furthermore, the primer pair 1 may consist of a forward primer with a nucleotide sequence as shown in SEQ ID NO:1 and a reverse primer with a nucleotide sequence as shown in SEQ ID NO:2; The primer pair 2 may consist of a forward primer with a nucleotide sequence as shown in SEQ ID NO:3 and a reverse primer with a nucleotide sequence as shown in SEQ ID NO:4; The primer pair 3 may consist of a forward primer with a nucleotide sequence as shown in SEQ ID NO:5 and a reverse primer with a nucleotide sequence as shown in SEQ ID NO:6; The primer pair 4 may consist of a forward primer with a nucleotide sequence as shown in SEQ ID NO:7 and a reverse primer with a nucleotide sequence as shown in SEQ ID NO:8; The primer pair 5 may consist of a forward primer with a nucleotide sequence as shown in SEQ ID NO:9 and a reverse primer with a nucleotide sequence as shown in SEQ ID NO:10; The primer pair 6 may consist of a forward primer with a nucleotide sequence as shown in SEQ ID NO:11 and a reverse primer with a nucleotide sequence as shown in SEQ ID NO:12; Furthermore, the application may include using data on the expression levels of the ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD22 genes in samples from known latent tuberculosis infected individuals and active tuberculosis patients as training samples, employing statistical analysis methods (such as binary logistic regression analysis) to establish an early risk warning model for latent tuberculosis infection, and using this model to provide early risk warnings for latent tuberculosis infection to the subjects.

[0012] Furthermore, the sample may be a blood sample (such as whole blood, plasma, and serum).

[0013] Furthermore, the sample may be whole blood.

[0014] In the above applications, the products may include reagents, kits, detection chips, test strips, test cards, immunosensors, or devices.

[0015] Furthermore, the product may include reagents for detecting the expression levels of the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

[0016] Furthermore, the reagent may be primers (e.g., primer pairs 1-6 described herein) for specifically amplifying the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene, as described herein, and / or probes for specifically recognizing the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

[0017] Furthermore, the product may be a kit, which may also include reagents required for PCR detection, such as DNA polymerase (e.g., Taq DNA polymerase, Tth DNA polymerase, Vent DNA polymerase, or Pfu DNA polymerase), dNTPs, Mg... 2+ Solutions (such as MgCl2 or MgSO4 solutions) and PCR buffers (such as Tris-HCl).

[0018] Furthermore, the kit may also include RNA extraction reagents.

[0019] Furthermore, the kit may be a detection kit based on reverse transcription real-time quantitative polymerase chain reaction (RT-qPCR).

[0020] The present invention also provides a method for constructing an early risk warning model for latent tuberculosis infection. The method includes using data on the expression levels of ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD22 genes in samples from known individuals with latent tuberculosis infection and patients with active tuberculosis as training samples, and employing statistical analysis methods to establish an early risk warning model for latent tuberculosis infection.

[0021] The statistical analysis methods described above may include binary logistic regression analysis, LASSO regression analysis, random forest, decision tree analysis, support vector machine analysis, and neural network analysis.

[0022] In the above method, the statistical analysis method can be the binary logistic regression analysis method.

[0023] In the above method, the formula for the early risk warning model can be: logit(P) = 16.257 + 1.6481 * ID3 - 1.3932 * GBP5 + 0.1276 * BATF2 - 0.4185 * ANKRD22 - 0.9597 * VAMP5 - 0.0115 * PSTPIP2; where logit(P) represents the probability of latent tuberculosis infection in the Logit model, and the gene name in the model represents the expression level of that gene.

[0024] Furthermore, the application method of the early risk warning model for latent tuberculosis infection constructed by any of the above methods may include the following steps: (1) Data on the expression levels of ID3, PSTPIP2, VAMP5, GBP5, BATF2 and ANKRD22 genes in the samples of the subjects to be tested; (2) Input the data into the early risk warning model for latent tuberculosis infection constructed by any of the methods described above; (3) The model is used to obtain early risk warning results of latent tuberculosis infection in the subjects to be tested.

[0025] Furthermore, based on the early risk warning results of latent tuberculosis infection, personalized or further prevention or treatment suggestions can be provided to the subjects to be tested, or early intervention treatment can be carried out for high-risk patients.

[0026] The present invention also provides a device for early risk warning of latent tuberculosis infection, the device comprising: i) A data receiving module, used to receive data on the expression levels of the biomarkers described herein in a sample of a subject to be tested from at least one terminal; ii) A data processing module, used to input the data into a model constructed by any of the methods described herein to obtain the probability value of latent tuberculosis infection; iii) A data output module for outputting the probability value to at least one client.

[0027] This invention also provides a method for early risk warning of latent tuberculosis infection, the method comprising the following steps: S1. Data reception: Receive data on the expression levels of the biomarkers described herein in the sample of the subject to be tested from at least one terminal; S2. Data processing: Input the data into the model constructed by any of the methods described in this paper to obtain the probability value of latent tuberculosis infection; S3. Data output: Output the probability value to at least one client.

[0028] In this document, the sequences of the ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD22 genes are well-known in the art and are publicly available from the NCBI website. Exemplary sequences of these genes are shown below: The coding sequence (CDS) of the ID3 gene described in this article is shown as positions 76-435 of GenBank Accession No. NM_002167.5, and its encoded amino acid sequence is shown as SEQ ID NO:13.

[0029] The coding sequence (CDS) of the PSTPIP2 gene described in this article is shown as positions 73-1077 of GenBank Accession No. NM_024430.4, and its encoded amino acid sequence is shown in SEQ ID NO:14.

[0030] The coding sequence (CDS) of the VAMP5 gene described in this article is shown as positions 61-411 of GenBank Accession No. NM_006634.3, and its encoded amino acid sequence is shown in SEQ ID NO:15.

[0031] The coding sequence (CDS) of the GBP5 gene described in this article is shown as positions 429-2189 of GenBank Accession No. NM_001134486.4, and its encoded amino acid sequence is shown in SEQ ID NO:16.

[0032] The coding sequence (CDS) of the BATF2 gene described in this article is shown as positions 136-888 of GenBank Accession No. NM_001300807.2, and its encoded amino acid sequence is shown in SEQ ID NO:17.

[0033] The coding sequence (CDS) of the ANKRD22 gene described in this article is shown as positions 238-813 of GenBank Accession No. NM_144590.3, and its encoded amino acid sequence is shown in SEQ ID NO:18.

[0034] The expression levels of the biomarkers (ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD2 genes) described herein are applicable to routine reverse transcription real-time quantitative polymerase chain reaction (RT-qPCR) detection experiments. While the embodiments provided in this invention exemplarily employ the RT-qPCR method to detect the expression levels of biomarkers in whole blood samples, this invention is not limited to this specific detection method. Those skilled in the art can use any other suitable and well-known detection methods (such as RNA-seq, Northern blot, gene chip detection, Nanopore sequencing, and PacBio sequencing, etc.) to detect the expression levels of biomarkers in whole blood samples. These alternative methods do not depart from the scope of this invention, and this invention should include these alternative methods.

[0035] This invention is the first to propose using a combination of ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD22 genes as biomarkers for early risk warning of latent tuberculosis infection, and to construct an early risk warning model for latent tuberculosis infection. This invention involves in-depth mining of transcriptome data from the GEO database, identifying 306, 742, and 206 differentially expressed genes associated with LTBI in three datasets, respectively. The intersection of these differentially expressed genes yielded 26 intersection genes. Through GO and KEGG enrichment analyses, combined with two machine learning algorithms and artificial intelligence technologies such as PPI networks, and integrating bioinformatics methods, ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD22 genes were ultimately selected as relevant biomarkers for patients with latent tuberculosis infection. Furthermore, an early risk warning model was constructed based on these six genes, and the model was validated through internal and external datasets, sequencing cohorts, prospective cohort studies, and RT-qPCR. The results showed that the AUCs were 0.911 (CI: 0.842–0.980) and 0.954 (CI: 0.924–0.984) in the two internal datasets, respectively; 0.977 (CI: 0.961–0.993) in the external dataset; and 0.770 (CI: 0.551–0.989) in the sequencing cohort. The expression levels of these six genes were validated through population cohort studies and RT-qPCR, showing trends consistent with the training set, and the overall AUC of the model was 0.899 (CI: 0.805–0.993). At a Youden index of 0.722, the sensitivity was 0.833, the specificity was 0.889, and the accuracy was 0.867.

[0036] Furthermore, considering the crucial role of differentially expressed genes related to LTBI in its occurrence and development, we not only identified disease-related biomarkers but also discovered important immune signaling pathways through KEGG enrichment analysis. These pathways, such as the B cell receptor signaling pathway, NOD-like receptor signaling pathway, JAK-STAT signaling pathway, and cytokine-cytokine receptor interactions, play a key role in the pathogenesis of latent tuberculosis infection. These new findings not only provide potential biological targets for the early detection and screening of LTBI but also offer new ideas and directions for the treatment and prevention of related diseases. This invention developed six reliable LTBI biomarkers, and the constructed risk assessment model demonstrated significant predictive efficiency, providing an early screening strategy for LTBI.

[0037] The biomarkers of this invention and the models constructed using these biomarkers have been verified as accurate and reliable by ROC curve analysis, making them suitable for widespread application. By providing early risk warnings of latent tuberculosis infection (LTBI) in test subjects, this invention can provide decision support for physicians, guide the selection and adjustment of treatment plans, offer personalized prevention and treatment suggestions to patients, or provide early intervention treatment for high-risk patients, thereby achieving optimal disease management and possessing broad clinical application value. Attached Figure Description

[0038] Figure 1 Box plots before and after standardization of GSE37250, GSE28623 and GSE19491.

[0039] Figure 2 Volcano plots of DEGs in GSE37250(A), GSE28623(B), and GSE19491(C). The plots contain two vertical dashed lines representing the log2 FC at -1 and 1. The downregulated genes to the left of the vertical dashed lines are represented by blue, and the downregulated genes to the right are represented by pink. The horizontal dashed line indicates that the adjusted p-value is 0.05.

[0040] Figure 3 Heatmaps of DEGs in GSE37250(A), GSE28623(B), and GSE19491(C).

[0041] Figure 4 Venn diagram of the intersecting genes in GSE37250, GSE28623 and GSE19491.

[0042] Figure 5 GO enrichment analysis for intersecting genes. Green indicates GO enrichment results for biological processes; blue indicates GO enrichment results for cellular components; orange indicates GO enrichment results for molecular functions. The x-axis represents the number of enriched genes. The y-axis represents the entries for GO-BP, GO-CC, and GO-MF.

[0043] Figure 6 This is a KEGG enrichment analysis of the intersecting genes. The x-axis represents the enrichment factor, and the y-axis represents the KEGG enrichment pathway name.

[0044] Figure 7 The results of feature selection and model optimization for LASSO regression on the GSE37250 dataset.

[0045] Figure 8 Important genes selected by RF in dataset GSE37250.

[0046] Figure 9The results of feature selection and model optimization for LASSO regression on dataset GSE28623.

[0047] Figure 10 Important genes selected by RF in dataset GSE28623.

[0048] Figure 11 The results of feature selection and model optimization for LASSO regression on the GSE19491 dataset.

[0049] Figure 12 Important genes selected by RF in the dataset GSE19491.

[0050] Figure 13 PPI network of intersecting genes and screening of core genes. A: Original STRING protein interaction network diagram; B: Visualization diagram using Cytoscape software.

[0051] Figure 14 To validate the expression levels of the core gene in three training datasets (GSE37250, GSE28623, GSE19491).

[0052] Figure 15 Diagnostic potential of single-gene and multi-gene combined models in dataset GSE37250.

[0053] Figure 16 To assess the diagnostic potential of single-gene and multi-gene combined models in the internal validation set GSE28623.

[0054] Figure 17 To assess the diagnostic potential of single-gene and multi-gene combined models in the internal validation set GSE19491.

[0055] Figure 18 To assess the diagnostic potential of single-gene and multi-gene combined models in the external validation set GSE107995.

[0056] Figure 19 To assess the diagnostic potential of combined single-gene and multi-gene models in sequencing cohorts.

[0057] Figure 20 A nomogram of core genes.

[0058] Figure 21 The calibration curve for the core gene.

[0059] Figure 22 The decision curve for the core gene.

[0060] Figure 23 This is for the expression analysis of core genes in the sequencing cohort.

[0061] Figure 24 This study analyzed the expression of core genes in an RT-qPCR cohort.

[0062] Figure 25 To assess the diagnostic potential of single-gene and multi-gene combined models in the RT-qPCR validation cohort.

[0063] Figure 26 For xCell immune infiltration analysis (stacked plot).

[0064] Figure 27 For xCell immune infiltration analysis (box plot).

[0065] Figure 28 Spearman correlation heatmap of immune cell subsets.

[0066] Figure 29 This study analyzed the correlation network between core genes and immune infiltrating cells. Detailed Implementation

[0067] The present invention will now be described in further detail with reference to specific embodiments. The given embodiments are merely illustrative of the invention and not intended to limit its scope. The embodiments provided below can serve as a guide for further improvements by those skilled in the art and do not constitute a limitation on the invention in any way.

[0068] Unless otherwise specified, the experimental methods used in the following examples are conventional methods, performed according to the techniques or conditions described in the literature in this field or according to the product instructions. Unless otherwise specified, the materials and reagents used in the following examples are commercially available.

[0069] Example 1: Inclusion of four datasets and sample features A search was conducted in the GEO database (https: / / www.ncbi.nlm.nih.gov / geo / ) using the keywords "latent tuberculosis infection" and "tuberculosis". Based on this, careful screening and identification were performed according to the following inclusion criteria.

[0070] Inclusion criteria: (1) Samples were derived from ATB patients or LTBI individuals; (2) Sample size ≥ 6 per group; (3) Each sample had complete gene expression data; (4) Subjects were > 15 years old; (5) Subjects had not received tuberculosis-related treatment; (6) Samples were blood samples not affected by HIV infection, autoimmune diseases, or malignancies; (7) Samples used for gene expression profiling studies must contain RNA expressed at the whole human genome level. Four datasets that met the criteria were finally identified: GSE37250, GSE28623, GSE19491, and GSE107995. The first three will be used as the training set, and GSE107995 will be used as the validation set. The inclusion of GEO data is shown in Table 1.

[0071]

[0072] Abbreviations: "ATB" stands for Active Tuberculosis; "LTBI" stands for Latent Tuberculosis Infection; "GEO" stands for Gene Expression Omnibus.

[0073] Example 2: Raw Data Processing Based on the four dataset formats, the `read.ilmn` function from the `limma` package, the `getGEO` function from the `GEOquery` package, the `fread` function from the `data.table` package, and the `read_tsv` function from the `readr` package were used to read the data. After mapping probe IDs (e.g., `EnsemblID`) to gene symbols, data cleaning was performed: ① Filtering probes without symbol annotations; ② Removing duplicate mappings to ensure that multiple probes for the same gene were merged and deduplicated; ③ Filtering low-expression genes (defined as genes whose expression level is below the 25th percentile of the dataset in ≥75% of samples); ④ Retaining samples from the ATB and LTBI groups and removing the healthy controls (HCs) group. Targeted preprocessing workflows were adopted for different platform characteristics: GSE37250 was standardized in one step using the `neqc` function of `limma`; GSE28623 underwent background correction, log2 transformation, and quantile normalization sequentially; GSE19491 was preprocessed with lumiExpresso and then normalized between rows. The standardized box plots show that the medians of the samples in each dataset are aligned and the dispersion is consistent, the batch effect is significantly eliminated, and the technical bias is effectively controlled, meeting the requirements for subsequent difference analysis. Figure 1 ).

[0074] Example 3: Identification of DEGs in the Training Set Based on the preprocessed expression matrices, the R language's limma package was used to conduct inter-group differential expression analysis on the training datasets (GSE37250, GSE28623, and GSE19491). The differential comparison was set as LTBI group (experimental group) versus ATB group (control group). The complete analysis process included: defining grouping factors, constructing a design matrix, fitting a linear model, Bayesian robust variance correction, extracting differential analysis results, and finally screening for significantly differentially expressed genes (DEGs). The screening threshold was a corrected P-value (adj.P.Val) < 0.05 and |log2FC| > 1. The analysis results showed that 306 (80 upregulated / 226 downregulated), 742 (316 upregulated / 426 downregulated), and 206 (168 upregulated / 38 downregulated) DEGs were obtained from the GSE37250, GSE28623, and GSE19491 training sets, respectively. To more intuitively display the distribution and expression patterns of differentially expressed genes, we plotted a volcano plot (…). Figure 2 ) and heatmap ( Figure 3 The volcano plot shows the relationship between the significance of DEGs (-log10 p-value) and the fold change in expression (log2FC), while the heatmap visually presents the expression patterns of DEGs in the LTBI and ATB samples.

[0075] To obtain a stable and reproducible set of target genes, the intersection of the three datasets (DEGs) was taken as the intersection genes. The intersection results were visualized using an online Venn diagram tool (Venny 2.1, https: / / bioinfogp.cnb.csic.es / tools / venny / ). Venn diagram analysis showed ( Figure 4 The three datasets contained 26 overlapping genes, of which 10 were upregulated relative to ATB in the LTBI group (ID3, CCR6, CD79B, EBF1, BLK, FCRLA, CD19, GZMK, AIM2, GNLY), and 16 were downregulated (GBP5, C1QB, ANKRD22, PSTPIP2, VAMP5, GBP1, FCGR1B, LMNB1, OSM, SOCS3, ATF3, TNFRSF25, IFI44, IFI44L, KLRG1, BATF2), showing a highly consistent expression trend.

[0076] Example 4: GO and KEGG enrichment analysis of differentially expressed genes The ClusterProfiler R package was used to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses on the genes intersecting ATB and LTBI. The significance criterion for the enrichment results was adj. p-value ≤ 0.05.

[0077] GO results indicate that ( Figure 5 In terms of biological processes, the interleaved genes were primarily enriched in terms related to immune responses and inflammation. Significantly enriched terms included: antigen receptor-mediated signaling pathway, positive regulation of inflammatory responses, B cell receptor signaling pathway, cellular defense response, and pyroptotic inflammatory response. Regarding cellular components, the interleaved genes were significantly enriched in components such as the outer plasma membrane, parts of other organisms, and the surface of symbiotic cells. In terms of molecular function, the interleaved genes were significantly enriched in tumor suppressor protein-M receptor binding and leprosycin B binding functions.

[0078] KEGG analysis revealed that the overlapping genes were mainly enriched in pathways such as the supplemental B cell receptor signaling pathway, the NOD-like receptor signaling pathway, the JAK-STAT signaling pathway, and cytokine-cytokine receptor interaction. Figure 6 ).

[0079] Example 5: ML algorithm for screening core genes This application employs a hierarchical strategy of "primarily using machine learning (ML) algorithms and secondarily using protein-protein interaction (PPI) networks" to screen core genes. First, a Least Absolute Shrinkage and Selection Operator (LASSO) regression model and a Random Forest (RF) model are constructed. Based on gene importance ranking, the top two feature genes are initially screened. Then, combined with the PPI interaction network, using GBP5 as the core anchor, genes with close interaction and high connectivity with GBP5 are selected to form a candidate gene set.

[0080] First, to improve the accuracy of the screening results and prevent overfitting, two machine learning algorithms—Least Absolute Shrinkage and Selection Operator (LASSO) regression and Random Forest (RF)—were used to further screen the intersection genes. The glmnet package in R was used for modeling, and 10-fold cross-validation was used to screen for the optimal regularization coefficient (lambda.min). Genes with non-zero regression coefficients were retained as candidate features for LASSO screening. The randomForest package in R was used to build a classification model, with disease group (LTBI / ATB) as the dependent variable and the differentially expressed genes in the intersection as the independent variable; parameter settings: number of decision trees ntree=500, other parameters remained at default. The feature genes screened by the LASSO regression and RF algorithms were ranked according to gene importance, and the top two feature genes were initially selected for each algorithm.

[0081] Based on LASSO regression and RF algorithms, the genes GBP5 and CD79B were identified in the dataset GSE37250. Figure 7 ) and ID3, GBP5 ( Figure 8 ) were identified as characteristic genes, and CCR6 and ID3 were determined in the dataset GSE28623 respectively. Figure 9 ) and PSTPIP2, ID3 ( Figure 10 ) were identified as characteristic genes. In GSE19491, VAMP5 and CCR6 were respectively identified as characteristic genes. Figure 11 ) and VAMP5, ID3 ( Figure 12 Based on the results of the three datasets mentioned above, we identified ID3, PSTPIP2, VAMP5, GBP5, CCR6, and CD79B as core genes for subsequent analysis. These genes may serve as potential biomarkers for LTBI, providing important evidence for subsequent functional validation and clinical applications.

[0082] Example 6: PPI Network Construction To analyze the interaction relationships of core encoded proteins, a PPI was constructed based on the STRING online database. The combined interaction score was set to >0.4, while isolated nodes with no interactions were removed. The string_interactions.tsv interaction file exported from STRING was imported into Cytoscape 3.10.4 software for visualization and topology analysis. We used the CytoHubba plugin to calculate the centrality index connectivity (Degree) and screened out core genes with high connectivity and significant regulatory effects in the network. GBP5 had the highest node degree and was the core hub of the network. In addition, genes such as BATF2, ANKRD22, and PSTPIP2 acted as bridge nodes, connecting the innate immune module centered on GBP5. Figure 13 ).

[0083] Since CD79B has weaker connectivity than ANKRD22 in the PPI network, we replace CD79B in the core genes initially screened by the ML algorithm with ANKRD22, which has broader connectivity. This gives us 6 core genes: ID3, PSTPIP2, VAMP5, GBP5, CCR6, and ANKRD22.

[0084] Example 7: Validation of Core Gene Expression Levels To verify whether the core genes screened showed intergroup expression differences, the expression levels of the core genes ID3, PSTPIP2, VAMP5, GBP5, CCR6, and ANKRD22 were analyzed in three training sets (GSE37250, GSE28623, and GSE19491). Figure 14 In data preprocessing, we need to extract the standardized expression matrix of the core genes. Statistical tests were then performed using GraphadPrism software (version 10.1.2), with disease status (LTBI vs. ATB) as the grouping variable. Different statistical descriptive methods were selected based on the data distribution characteristics: normally distributed continuous variables were expressed as mean ± standard deviation (Mean ± SD); non-normally distributed continuous variables were expressed as quartiles [50% (25%-75%)]; and count data were presented directly as numerical values. In comparing two groups of data, if both groups conformed to a normal distribution, an unpaired t-test was used; otherwise, the Mann-Whitney U test was used. In this study, P < 0.05 was used as the statistical significance threshold.

[0085] ID3, PSTPIP2, VAMP5, GBP5, and ANKRD22 showed significant inter-group expression differences in all datasets (P<0.0001), while CCR6 only showed differences in datasets GSE37250 and GSE28623 (P<0.0001), and was slightly significant in dataset GSE19491 (P=0.0172). Based on the validation results of the three datasets, ID3, PSTPIP2, VAMP5, GBP5, and ANKRD22 showed stable and significant expression differences in both LTBI and ATB groups, demonstrating good potential for differential diagnosis. CCR6 only showed highly significant differences in two datasets, and was only slightly significant in the third dataset, suggesting insufficient robustness as a differential diagnostic biomarker; therefore, it was not included in the final core gene set. Based on the PPI network topology analysis results, CCR6 was replaced with BATF2, which is tightly connected to GBP5 in the network and ranks among the top 3 in RF feature importance. Validation showed that BATF2 exhibited significant inter-group expression differences across the three datasets (P < 0.0001). Therefore, the final optimized core gene set is ID3, PSTPIP2, VAMP5, GBP5, ANKRD22, and BATF2.

[0086] Example 8: Single-gene diagnostic efficacy analysis and construction of a multi-gene combined diagnostic model To evaluate the diagnostic efficacy of each core gene in distinguishing between LTBI and ATB individually, receiver operating characteristic (ROC) curves were constructed for each core gene based on the training set GSE37250, and the area under the curve (AUC) was calculated. To integrate multi-gene features and improve identification accuracy, a multi-gene joint diagnostic model was constructed. The six core genes were included in a binary logistic regression analysis, and the multi-gene joint diagnostic model was constructed as follows: logit(P) = 16.257 + 1.6481*ID3 - 1.3932*GBP5 + 0.1276*BATF2 - 0.4185*ANKRD22 - 0.9597*VAMP5 - 0.0115*PSTPIP2, named the Hexa-TB Model. Here, logit(P) represents the probability of latent tuberculosis infection in the Logit model (i.e., the probability of latent tuberculosis infection calculated by the model), and the gene names in the model represent the expression levels of those genes. Finally, external validation of the single-gene and multi-gene combined models was completed sequentially in the internal validation datasets (GSE28623, GSE19491), the external independent dataset GSE107995, and the prospective sequencing cohort (see Example 10 for details).

[0087] In the training set GSE37250, all six core genes showed the ability to distinguish between LTBI and ATB. Figure 15 Among the various diagnostic methods, ANKRD22 showed the best diagnostic efficacy, with an AUC of 0.943 (95% CI: 0.910–0.974); followed by BATF2 (AUC = 0.933, 95% CI: 0.896–0.970); ID3 and GBP5 showed similar diagnostic efficacy, with AUCs of 0.929 (95% CI: 0.886–0.971) and 0.928 (95% CI: 0.886–0.970), respectively, followed by PSTPIP2 (AUC = 0.913, 95% CI: 0.870–0.954) and VAMP5 (AUC = 0.910, 95% CI: 0.867–0.953). The combined model demonstrated significantly better diagnostic efficacy than any single gene (see [link to original text]). Figure 15 Its AUC value reached 0.968 (95% CI: 0.947–0.990).

[0088] Within the GSE28623 intimatum set, ID3 showed the best diagnostic efficacy with an AUC of 0.936 (95% CI: 0.878–0.994); PSTPIP2 had an AUC of 0.916 (95% CI: 0.848–0.983); BATF2 had an AUC of 0.887 (95% CI: 0.812–0.962); ANKRD22 had an AUC of 0.887 (95% CI: 0.812–0.962); GBP5 had an AUC of 0.865 (95% CI: 0.778–0.952); and VAMP5 had an AUC of 0.810 (95% CI: 0.709–0.912). Figure 16 The AUC of the multi-gene combined model on this dataset was 0.911 (95% CI: 0.842–0.98). Figure 16 ).

[0089] In the internal validation set GSE19491, the AUCs in order of diagnostic efficacy were as follows: VAMP5 AUC 0.956 (95% CI: 0.927-0.985), ID3 AUC 0.909 (95% CI: 0.864-0.954), GBP5 AUC 0.885 (95% CI: 0.833-0.936), ANKRD22 AUC 0.819 (95% CI: 0.747-0.892), BATF2 AUC 0.812 (95% CI: 0.739-0.884), and PSTPIP2 AUC 0.807 (95% CI: 0.733-0.881). Figure 17The AUC value of the multi-gene combined model was 0.954 (95% CI: 0.924–0.984). Figure 17 ).

[0090] The results of the external validation set GSE107995 showed that the diagnostic efficacy of the single genes, from highest to lowest, was as follows: ANKRD22: AUC 0.978 (95% CI: 0.963-0.994), BATF2: 0.968 (95% CI: 0.949-0.988), GBP5: 0.947 (95% CI: 0.92-0.974), PSTPIP2: 0.947 (95% CI: 0.92-0.973), VAMP5: 0.919 (95% CI: 0.881-0.958), ID3: 0.870 (95% CI: 0.818-0.922). Figure 18 The multi-gene joint model had an AUC of 0.977 (95% CI: 0.961–0.993) on external datasets. Figure 18 This confirms that the model has good generalization ability across datasets.

[0091] Sequencing data (ATB=10; LTBI=10) validation showed that the diagnostic efficacy of single genes, from highest to lowest, was as follows: BATF2: 0.95 (95% CI: 0.865-1), ANKRD22: 0.92 (95% CI: 0.797-1), GBP5: 0.82 (95% CI: 0.604-1), VAMP5: 0.8 (95% CI: 0.579-1), PSTPIP2: 0.67 (95% CI: 0.417-0.923), ID3: 0.51 (95% CI: 0.235-0.785). Figure 19 The multi-gene joint model had an AUC of 0.77 (95% CI: 0.551–0.989) on external datasets. Figure 19 ).

[0092] All statistical analyses and ROC curve plotting were performed in R software (version: 4.4.3, R Foundation for Statistical Computing); the "pROC" package was used to plot ROC curves, calculate AUC, and estimate 95% CI (DeLong method) to quantify the diagnostic efficacy of each biomarker and the joint model; the "ggplot2" package was used to complete all statistical graphical visualizations.

[0093] Example 9: Construction and Evaluation of Nodal Charts To evaluate the model's predictive diagnostic capability and clinical value, a nomogram based on the aforementioned characteristic genes was constructed using the R package "rms" in R software (version 4.3.2). Nomograms are widely used in the medical field, especially in oncology and other areas requiring precise prediction. A nomogram can include multiple different influencing factors that may simultaneously affect prognosis, predicting survival or occurrence in the study cohort. To validate the predictive accuracy of the nomogram, a calibration curve was plotted to assess the consistency between expected probabilities and actual outcomes. Higher closeness of the calibration curves indicates stronger predictive ability of the model. Furthermore, decision curve analysis (DCA) was used to evaluate the clinical usefulness and advantages of the nomogram. DCA assesses the model's practical application value in clinical decision-making by quantifying the model's net benefit at different threshold probabilities.

[0094] To visualize the Hexa-TB Model and improve its ease of use in clinical practice, we created a nomogram of the Hexa-TB Model. Figure 20 In a nomogram, the contribution value of each gene can be visually displayed using its position on a scale. We can calculate the total score by locating the score of each gene on the scale and projecting these scores onto the total score scale at the bottom. The total score can then be mapped to the probability scale at the bottom to assess the likelihood of a patient having LTBI.

[0095] Through calibration curve analysis, we found a high degree of consistency between the predicted probabilities of the Hexa-TB Model and the actual observed values, indicating that the Hexa-TB Model has a good fit. Figure 21 Furthermore, DCA showed that, over a wider range of risk thresholds, the net benefit of the 6-gene model was higher than that of the "all-treatment" and "no-treatment" strategies, indicating that the model has good clinical application value. Figure 22 ).

[0096] In conclusion, the constructed 6-gene diagnostic model, Hexa-TB Model, demonstrates certain clinical application potential in distinguishing between LTBI and ATB. This model not only provides a new tool for the precise diagnosis of tuberculosis but also lays an important foundation for subsequent clinical translational research.

[0097] Example 10: Real-world prospective cohort validation of significant differences in the expression of six core genes. 1. Cohort study validation of the LTBI early risk warning model To further validate the expression of the six core genes mentioned above in a real-world population, we recruited 10 patients with ATB, 10 patients with LTBI, and 10 patients with HCs for a prospective cohort study. This study was approved by the Medical Ethics Committee of the Eighth Medical Center of the PLA General Hospital (Ethics Approval No.: 3092023122013297234), and all participants signed informed consent forms. We collected 5 ml of peripheral blood from each participant to isolate peripheral blood mononuclear cells (PBMCs), and then extracted total RNA. After ensuring that the RNA sample quality met the standards, a small RNA library was constructed and sequenced using the Illumina sequencing platform.

[0098] Inclusion criteria for HCs: A. No history of contact with tuberculosis patients; B. Negative enzyme-linked immunospot test; C. No clinical manifestations of tuberculosis, normal chest X-ray, and exclusion of active tuberculosis; D. HIV negative.

[0099] Exclusion criteria for HCs: A. History of travel to or residence in high-risk areas for tuberculosis; B. Staff of tuberculosis specialist hospitals or laboratories; C. Children under 12 years of age; D. Those with a history of tuberculosis or old lesions on lung imaging; E. Those who cannot undergo CE antigen testing or are allergic to it; F. HIV positive; G. Those who cannot undergo CE antigen testing or are allergic to it.

[0100] LTBI inclusion criteria: A. History of close contact with patients with pulmonary tuberculosis; B. Staff of specialized tuberculosis hospitals or laboratories; C. No clinical symptoms of pulmonary tuberculosis; D. Normal chest X-ray; E. Positive IGRA; F. HIV negative; G. 12 years of age and older.

[0101] LBTI exclusion criteria: A. Patients with pulmonary tuberculosis; B. Pregnant or lactating women; C. HIV positive; D. Patients undergoing anti-tuberculosis treatment for one month or more; F. Children under 12 years of age.

[0102] ATB Inclusion Criteria: This study defines ATB patients as those diagnosed with tuberculosis of the lungs, trachea, bronchi, and pleura according to the Diagnostic Criteria for Pulmonary Tuberculosis (WS288-2017). The diagnosis of tuberculosis is based on etiological and pathological results, combined with a comprehensive analysis of epidemiological history, clinical manifestations, chest imaging, relevant auxiliary examinations, and differential diagnosis.

[0103] ATB exclusion criteria: A. Those using hormones; B. Those with diseases affecting immune function, such as HIV infection, post-transplant, and autoimmune diseases; C. Those who are malnourished; D. Children under 12 years of age.

[0104] Significant differences were found in the expression of genes VAMP5, ANKRD22, and BATF2 between the ATB and LTBI groups (P<0.05). Figure 23Specifically, compared to the ATB group, genes VAMP5, ANKRD22, and BATF2 were significantly downregulated in the LTBI group (P<0.05). While the remaining three genes showed some trend of difference between the ATB and LTBI groups, these differences did not reach statistical significance. This result may be related to the small sample size or population heterogeneity, suggesting that further validation of these gene expression differences and their clinical significance in a larger population cohort is necessary.

[0105] 2. Core gene expression analysis of the RT-qPCR validation cohort 2.1 Sub-queue design and sample collection and processing We validated the reliability of the bioinformatics-based predictions using RT-qPCR. A total of 102 participants were recruited from the Eighth Medical Center of the PLA General Hospital, including individuals with HCs (n=32) and individuals with LTBI (n=32). n =30) and ATB patients ( n =40).

[0106] The inclusion and exclusion criteria for each group were as described above. The diagnosis of ATB patients was based on the "Diagnostic Criteria for Pulmonary Tuberculosis WS288-2017," while the diagnosis of LTBI patients was based on the results of the Interferon-Gamma Release Assay (IGRA). Immunodeficiency, use of glucocorticoids, pregnancy, and children under 12 years of age were excluded.

[0107] Screening for HCs and LTBI involved collecting 6 mL of peripheral blood from each subject who had fasted for 8 hours (4 mL of heparin anticoagulant tube for ELISPOT and 2 mL of EDTA anticoagulant tube for RNA extraction and PCR verification). For ATB, blood was collected according to the diagnostic criteria for tuberculosis.

[0108] 2.2 ELISPOT screening of LTBI and HCs populations 1) Prepare several 10 mL EP tubes (depending on the number of blood vessels). First, add 3 mL of PBMC cell separation solution to the EP tube. Then, invert the blood collection tube to mix it well. Finally, add 4 mL of blood to the EP tube to form a clear interface.

[0109] 2) Centrifuge at 2500 rpm for 20 min at room temperature (centrifuge speed 9 and deceleration speed 1). Gently aspirate the PBMC cell cloud layer into a 10 mL sterile centrifuge tube (pre-filled with 5 mL of 1640 cell culture medium). Gently invert to resuspend and centrifuge at 1500 rpm for 10 min at room temperature (centrifuge speed 9 and deceleration speed 9).

[0110] 3) Discard the supernatant, add 5 mL of 1640 culture medium to resuspend the precipitated cells, and centrifuge at 1500 rpm for 10 min at room temperature (centrifuge speed 9 for both ascending and descending).

[0111] 4) Discard the supernatant and resuspend the cells in 400 µL of serum-free AIM medium. Add 20 µL of the PBMC cell suspension to a cell counting chamber and count the number of PBMC cells per mL of suspension using a computer. Adjust the PBMC cell concentration to 2.5 × 10⁻⁶ cells / mL using AIM medium. 6 Quantity / mL, for later use.

[0112] 5) Subsequently, the ELISPOT plates were pretreated: washed 4 times with 200 µL PBS per well, 1 min each time, and then patted dry. 200 µL of PBS containing 2% PSA and 1% trehalose was added to each well to block the ELISPOT plate, and the plate was incubated at room temperature for 2 h.

[0113] 6) Dilute the antigens to the following concentrations using AIM (blood-free medium): CE protein, 90 µg / mL; PHA, 90 μg / mL. Then, remove the pretreated ELISPOT plate, discard the blocking solution, and rinse once with 50 µL of 1640 culture medium before use.

[0114] 7) Add AIM (negative control), PHA (positive control), and CE antigen to each well of the ELISPOT plate at a volume of 50 µL / well. Subsequently, add 100 µL of a working concentration of 2.5 × 10⁻⁶ to each well of the ELISPOT plate. 6 A cell suspension of 1 PBMC / mL was prepared. ELISPOT plates were incubated at 37°C in a 5% CO2 incubator for 18-20 h.

[0115] 8) Remove the EISPOT plate from the incubator and discard the supernatant. Wash the plate 5 times with 200 µL 1×PBS, 1 min each time, and pat dry after the last wash. Add 100 µL of IFN-γ alkaline phosphatase antibody (ALP) to each well, incubate at room temperature for 2 h, discard the polyclonal antibody solution, wash the plate 5 times with 200 µL 1×PBS, 1 min each time, and pat dry after the last wash.

[0116] 9) Add 100 µL of BCIP / NBT substrate solution to each well and incubate at room temperature in the dark for 5-15 min until the spots grow to the appropriate size. Once blue-purple spots appear on the bottom of the plate, wash the plate three times with distilled water or water for injection to stop the reaction. Place the plate in a 37°C oven to dry or air dry at room temperature.

[0117] 10) Use a CTL (Chemistry, Inc.) enzyme-linked dot analyzer to count and analyze the stained spots, and classify the population based on the number of spots. The judgment criteria are as follows: Positive: 1. When the number of spots in the blank control well is 0-5, and (number of spots in antigen A or antigen B wells) - (number of spots in blank control wells) ≥ 6. 2. When the number of spots in the blank control well is 6-10, and (number of spots in antigen A or antigen B wells) ≥ 2 × (number of spots in blank control wells). Negative: Antigen wells - blank control < 6 and positive control >= 20.

[0118] 2.3 RNA Extraction 1) Add 1 mL of Buffer L9 to a 2 mL centrifuge tube, then add 500 µL of whole blood or bone marrow, vortex for 30 seconds to mix thoroughly. Centrifuge at 13000 rpm for 5 minutes.

[0119] 2) Transfer 700 μL of the supernatant to a 1.5 mL centrifuge tube containing 500 µL of anhydrous ethanol. Do not discard the pipette tip. Mix twice. Then, transfer 600 μL of the mixture to a nucleic acid purification column, cap the tube, and centrifuge at 12,000 rpm for 30 seconds.

[0120] 3) Discard the filtrate, place the nucleic acid purification column back into the 2 mL centrifuge tube, pour all the remaining liquid in the 1.5 mL centrifuge tube into the nucleic acid purification column, cap the tube, and centrifuge at 12000 rpm for 30 seconds.

[0121] 4) Discard the filtrate, place the nucleic acid purification column back into the 2 mL centrifuge tube, add 500 μL Buffer WA to the nucleic acid purification column, cap the tube, and centrifuge at 12000 rpm for 30 seconds.

[0122] 5) Discard the filtrate in the 2 mL centrifuge tube, place the nucleic acid purification column back into the 2 mL centrifuge tube, add 700 μL of Buffer WBR to the nucleic acid purification column, cap the tube, and centrifuge at 12000 rpm for 30 seconds.

[0123] 6) Discard the filtrate, place the nucleic acid purification column back into a 2 mL centrifuge tube, and centrifuge at 14,000 rpm for 1 minute. Then, discard the 2 mL centrifuge tube, place the nucleic acid purification column into a 1.5 mL RNase-free centrifuge tube, add 50 μL of RNase-free water to the purification column, cap the tube, let it stand at room temperature for 1 minute, and centrifuge at 12,000 rpm for 30 seconds.

[0124] 7) Transfer the RNA from the 1.5 mL centrifuge tube to a 0.6 mL centrifuge tube. First, add 1 μL of enzyme-free water to the UV spectrophotometer for calibration (the detection data should be 0). Then, add 1 μL of each experimental group sample to the UV spectrophotometer and measure the RNA concentration.

[0125] 2.4 Reverse transcription and qPCR 1) cDNA synthesis was performed using the FastKing gDNA Dispelling RT SuperMix kit, following the instructions in the manual. Quantitative real-time RT-PCR amplification was then performed using the FastReal qPCR PreMix (SYBR Green) premixed reagent in the QuantStudio™ Design & Analysis Software real-time quantitative PCR instrument, following the manufacturer's instructions.

[0126] 2) Finally, the relative abundance of each gene is calculated by subtracting the CT value of each sample from the corresponding GAPDH CT value (ΔCT). ΔΔCT is obtained by subtracting ΔCT from the reference point. These values ​​are then rounded up to 2-ΔΔct to generate a folded expression relative to the reference point, and the corresponding values ​​are calculated. Representative data are represented as the mean of three samples + SD.

[0127] 3) Primer sequence design is as follows: ID3: forward, 5'-TACAGCGCGTCATCGACTAC-3' (SEQ ID NO: 1); reverse, 5'-TGACAAGTTCCGGAGTGAGC-3' (SEQ ID NO: 2); PSTPIP2: forward, 5'-ATGGCCGCAAGAACTGCAA-3' (SEQ ID NO: 3); reverse, 5'-ACATTGTGCCACATTGTCTACTT-3' (SEQ ID NO: 4); VAMP5: forward, 5'-AGCGTTCAGACCAACTCCTG-3' (SEQ ID NO: 5); reverse, 5'-CGGATGTTCTCCCAGCACTT-3' (SEQ ID NO: 6); GBP5: forward, 5'-AGTTCTGCTTGACACCGAGG-3' (SEQ ID NO: 7); reverse, 5'-GCAGTAGGTCGATAGCACCC-3' (SEQ ID NO: 8); BATF2: forward, 5'-CTGAACCTCCTGTCCAGCTG-3' (SEQ ID NO:9); reverse, 5'-AGGGCACTGAGCTTAGAGGA-3' (SEQ ID NO:10); ANKRD22: forward, 5'-ATGCTACTTGATGCTGGCGT-3' (SEQ ID NO: 11); reverse, 5'-TTATTGTGGGGTCTGCACGG-3' (SEQ ID NO: 12).

[0128] The results showed that the expression of all six core genes differed significantly between the ATB and LTBI groups (P<0.05). Specifically, compared with the ATB group, gene ID3 was significantly upregulated in the LTBI group (P<0.05), while genes PSTPIP2, VAMP5, GBP5, ANKD22, and BATF2 were significantly downregulated in the LTBI group (P<0.05). Figure 24 We then plotted the ROC curve, and the ROC curve analysis results showed ( Figure 25 The overall AUC of the model was 0.899 (CI: 0.805-0.993). At a Youden index of 0.722, the sensitivity was 0.833, specificity was 0.889, and accuracy was 0.867. These results indicate that the six core genes of this invention—ID3, PSTPIP2, VAMP5, GBP5, ANKD22, and BATF2—may play important regulatory roles at different stages of tuberculosis infection, particularly in differentiating between LTBI and ATB. The early risk model for LTBI constructed based on these six genes demonstrates good accuracy, providing a new method for early risk warning of LTBI.

[0129] Example 11: Immune Infiltration Analysis and Visualization Immune response plays a crucial role in the occurrence and development of LTBI and ATB. To gain a deeper understanding of the immune status of patients with LTBI and ATB, this study used the xCell algorithm to perform immune infiltration analysis on the samples.

[0130] The results of the immune cell stacking diagram show that ( Figure 26 Both groups were characterized by neutrophils, monocytes, and CD4+. + Resting memory T cells and CD8 +T cells were the dominant immune cell subset. In the ATB group, the proportion of neutrophils and monocytes was increased, suggesting that the innate immunity of ATB was strongly mobilized, with a large infiltration of myeloid cells to combat pathogens. Simultaneously, the proportion of pro-inflammatory macrophages (M1) and activated dendritic cells (DCs) was slightly increased in the ATB group, reflecting antigen-driven activation of adaptive immunity, presenting a "high inflammation-strong response" immune phenotype. In contrast, the LTBI group showed a more stable balance in the distribution of immune cells: a slightly lower proportion of neutrophils and CD8+. + The relatively stable proportion of T cells and the balanced distribution of regulatory T cells (Tregs) and other suppressive subsets suggest that the immune system can effectively control Mycobacterium tuberculosis replication, maintain the latent infection state, and avoid excessive inflammatory damage. This difference in the immune spectrum suggests that the progression from LTBI to ATB, accompanied by overactivation of innate immunity and imbalance of adaptive immunity, may be a key immunological feature of tuberculosis pathogenesis.

[0131] Box plots further quantified the differences in the distribution of various immune cells between the LTBI and ATB groups. Figure 27 The study revealed statistically significant cell types. Box plot analysis showed significantly elevated expression levels of neutrophils and macrophages M0 in the ATB group, suggesting that ATB is accompanied by strong activation of innate immunity and inflammatory response. Meanwhile, the LTBI group showed significantly increased expression levels of naive B cells and CD8+. + T cells (CD8), activated CD4 + Significantly elevated expression levels of memory T cells (CD4 memory activated) and activated NK cells suggest that adaptive and innate-like immunity maintain efficient surveillance functions to control Mycobacterium tuberculosis replication during latent infection. No significant differences were observed in the expression of other immune cell subsets between the two groups. These results indicate that ATB and LTBI exhibit distinctly different immune profiles: ATB is characterized by excessive activation of innate immunity and hyperinflammatory activity, while LTBI is characterized by adaptive immune homeostasis activation and enhanced immune surveillance, providing potential immunological evidence for their differential diagnosis.

[0132] Spearman correlation analysis of immune cells showed that ( Figure 28 A characteristic synergistic regulatory network was formed among various immune cell subsets. A strong positive correlation was observed among innate immune cells such as monocytes, neutrophils, M0 macrophages, γδT cells, activated mast cells, and activated dendritic cells, suggesting that these cells were synergistically activated during tuberculosis infection, collectively constituting the core of the body's inflammatory response. Simultaneously, memory CD4+...+ Activated adaptive T cells, professional antigen-presenting cells (resting dendritic cells, M1 macrophages), humoral immune effector cells (plasma cells), immunomodulatory cells (Tregs), and resting NK innate immune cells each form independent, highly positively correlated cell clusters, reflecting the functional synergy at different stages of the immune response. Furthermore, significant negative correlations exist among various cell subsets, such as CD8... + The presence of effector T cells and monocytes suggests an antagonistic relationship between these two cell types in maintaining immune homeostasis. This complex network of immune cell associations provides a new perspective for elucidating the differences in the immune microenvironment between ATB and LTBI.

[0133] The correlation analysis between the six core genes (ID3, PSTPIP2, VAMP5, GBP5, BATF2, ANKRD22) and immune cells showed that ( Figure 29ID3 showed a strong positive correlation with naive B cells, T cells CD8, and Master cells resting, and a negative correlation with B cell memory, Macrophages M0, and Neutrophils. PSTPIP2 showed a positive correlation with Monocytes, Macrophages M0, Dendritic cells activated, and Neutrophils, and a negative correlation with B cell memory, T cell CD8, T cell CD4 memory resting, NK cell activation, and Master cells resting. VAMP5 showed a positive correlation with Monocytes, Macrophages M0, and T cell CD4 memory activated, and a negative correlation with B cell naive, T cell CD8, and T cell CD4 memory resting. GBP5 showed a positive correlation with Monocytes and Dendritic cells activated, and a negative correlation with Master cells resting, B cell naive, NK cell resting, T cell CD8, and T cell CD4 memory resting. BATF2 was positively correlated with Monocytes, Macrophages M0, Dendritic cells activated, and Neutrophils; and negatively correlated with Mastcells resting, B cells naive, T cells CD8, and T cells CD4 memory resting. ANKRD22 was positively correlated with Monocytes, Macrophages M0, Macrophages M1, and Neutrophils; and negatively correlated with Mastcells resting, NK cells activated, T cells CD4 memory resting, T cells CD8, and B cells naive. These gene-immune cell association patterns support the diagnostic value of the core genes at the immune mechanism level: GBP5, BATF2, and ANKRD22 genes are positively correlated with pro-inflammatory innate immune cells, suggesting their activation role in the ATB inflammatory response; while ID3 and CD8... + The correlation between T cells and adaptive immune cells such as naive B cells is positive, reflecting their role in maintaining LTBI immune surveillance homeostasis.

[0134] The present invention has been described in detail above. Those skilled in the art will recognize that the invention can be practiced in a wide range of ways with equivalent parameters, concentrations, and conditions without departing from its spirit and scope, and without requiring unnecessary experiments. While specific embodiments have been provided, it should be understood that further modifications can be made to the invention. In summary, according to the principles of the invention, this application is intended to include any changes, uses, or improvements to the invention, including changes made using conventional techniques known in the art that depart from the scope disclosed herein.

Claims

1. The application of substances used to detect biomarker expression levels in any of the following: A1) Application in the preparation of products for early risk warning of latent tuberculosis infection; A2) Application in constructing an early risk warning model for latent tuberculosis infection; The biomarkers are the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

2. The application according to claim 1, characterized in that, The substances include reagents for detecting the expression levels of the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

3. The application according to claim 2, characterized in that, The reagents include primers for specifically amplifying the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene, and / or probes for specifically recognizing the ID3 gene, PSTPIP2 gene, VAMP5 gene, GBP5 gene, BATF2 gene, and ANKRD22 gene.

4. The application according to any one of claims 1-3, characterized in that, The products include reagents, kits, detection chips, test strips, test cards, immunosensors, or devices.

5. A method for constructing an early risk warning model for latent tuberculosis infection, characterized in that, The method includes using data on the expression levels of the ID3, PSTPIP2, VAMP5, GBP5, BATF2, and ANKRD22 genes in samples from known individuals with latent tuberculosis infection and patients with active tuberculosis as training samples, and employing statistical analysis methods to establish an early risk warning model for latent tuberculosis infection.

6. The method according to claim 5, characterized in that, The statistical analysis methods include binary logistic regression analysis, LASSO regression analysis, random forest, decision tree analysis, support vector machine analysis, and neural network analysis.

7. The method according to claim 6, characterized in that, The statistical analysis method is binary logistic regression analysis.

8. The method according to any one of claims 5-7, characterized in that, The formula for the early risk warning model is: logit(P) = 16.257 + 1.6481 * ID3 - 1.3932 * GBP5 + 0.1276 * BATF2 - 0.4185 * ANKRD22 - 0.9597 * VAMP5 - 0.0115 * PSTPIP2; where logit(P) represents the probability of latent tuberculosis infection in the Logit model, and the gene name in the model represents the expression level of that gene.

9. A device for early risk warning of latent tuberculosis infection, characterized in that, The device includes: i) A data receiving module, configured to receive data on the expression level of the biomarker as described in claim 1 in a sample of a subject to be tested from at least one terminal; ii) A data processing module, used to input the data into a model constructed by any one of the methods described in claims 5-8 to obtain the probability value of latent tuberculosis infection; iii) A data output module for outputting the probability value to at least one client.

10. A method for early risk warning of latent tuberculosis infection, characterized in that, The method includes the following steps: S1. Data reception: Receiving data on the expression level of the biomarker described in claim 1 from at least one terminal; S2. Data processing: Input the data into the model constructed by any of the methods described in claims 5-8 to obtain the probability value of latent tuberculosis infection; S3. Data output: Output the probability value to at least one client.

Images