A combination of host genetic markers for active pulmonary tuberculosis, screening method and nucleic acid detection reagent and kit

Abstract

The application discloses a kind of active pulmonary tuberculosis host gene marker combination, screening method and nucleic acid detection reagent and kit, belong to molecular diagnosis, bioinformatics, machine learning assisted biomarker screening and in vitro detection reagent technical field.The host gene marker combination disclosed in the application is selected from at least three kinds in ARID4B, CXorf38, HLA-DMB, HLA-DQA1 and SLC25A37.The above-mentioned marker is determined by the present application through public peripheral blood mononuclear cell transcriptome data integration, differential expression analysis, weighted gene co-expression network analysis, random forest, XGBoost, LASSO feature screening and independent sample qRT-PCR verification.The present application also provides corresponding primer group, kit and data interpretation model.The product of the present application is only used for ex vivo sample detection and auxiliary judgment of active pulmonary tuberculosis, does not replace etiology, imaging and clinical comprehensive diagnosis.

Description

Technical Field

[0001] This invention relates to the fields of molecular diagnostics, bioinformatics, machine learning-assisted biomarker screening, and in vitro diagnostic reagents, and more specifically to a combination of host gene markers for active pulmonary tuberculosis, a method for screening the markers, nucleic acid detection reagents, and in vitro diagnostic kits. Background Technology

[0002] The diagnosis of active pulmonary tuberculosis usually relies on a comprehensive assessment of etiological, imaging, and clinical manifestations. Some patients experience delays in diagnosis due to low bacterial loads, difficulty in obtaining samples, or negative etiological results. Peripheral blood transcriptional responses can reflect the immune activation state induced by Mycobacterium tuberculosis infection, and are characterized by convenient sampling, quantifiability, and suitability for multi-indicator combined analysis.

[0003] Single-host gene assays are susceptible to individual differences, sample processing, co-infections, and technological platforms. Combinations of multiple biomarkers can leverage complementary information between different biological pathways to improve model stability; however, an excessive number of biomarkers can increase detection costs and the risk of model overfitting. Therefore, it is necessary to compare the detection efficacy of single, dual, triple, quadruple, and pentabiomarkers to determine the minimum effective combination that balances accuracy, stability, and detection efficiency.

[0004] Therefore, providing a combination of host gene markers for active pulmonary tuberculosis, a screening method, and nucleic acid detection reagents and kits is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a combination of host gene biomarkers for active pulmonary tuberculosis, a screening method, and nucleic acid detection reagents and kits. It provides a set of gene biomarkers that can stably reflect the transcriptional characteristics of hosts with active pulmonary tuberculosis. Through multi-dataset integration, network analysis, machine learning, and independent sample validation, a technical solution for the joint detection of at least three biomarkers is determined. While maintaining controllable detection throughput and cost, it improves AUC, sensitivity, specificity, accuracy, and repeatability.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A host gene biomarker combination for in vitro auxiliary detection of active pulmonary tuberculosis, wherein the host gene biomarker combination includes at least three of ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37. Preferably, a three-biomarker combination of ARID4B, HLA-DMB, and SLC25A37 is used; CXorf38 is added to form a four-biomarker combination; or HLA-DQA1 is further added to form a five-biomarker combination.

[0008] Furthermore, the aforementioned host gene marker combination is used in the preparation of products for in vitro auxiliary detection of active pulmonary tuberculosis, risk stratification, efficacy monitoring studies, or evaluation of host transcriptional characteristics.

[0009] Furthermore, qRT-PCR, digital PCR, microarray, or sequencing reagents for detecting the mRNA expression levels of the aforementioned biomarkers, as well as data interpretation kits containing said reagents.

[0010] Furthermore, a nucleic acid detection reagent for in vitro auxiliary detection of active pulmonary tuberculosis, the nucleic acid detection reagent comprising primers, probes, or combinations thereof for detecting the expression level of the host gene marker combination; the probe being a chip probe, a digital PCR probe, or a sequencing capture probe.

[0011] Preferably, the reagent includes the target gene primers shown in SEQ ID NO.1 to SEQ ID NO.10, and the GAPDH internal reference primers shown in SEQ ID NO.11 to SEQ ID NO.12.

[0012] Furthermore, a kit for in vitro auxiliary detection of active pulmonary tuberculosis includes the aforementioned nucleic acid detection reagent, as well as one or more of the following: RNA extraction reagent, reverse transcription reagent, nucleic acid amplification reagent, positive control, negative control, calibrator, and data interpretation instructions.

[0013] Furthermore, the positive control is a nucleic acid reference material containing at least three target genes from ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37, as well as the corresponding target sequence of the internal reference gene GAPDH. The nucleic acid reference material includes synthetic RNA, in vitro transcribed RNA, cDNA, recombinant plasmid, or a mixture thereof. The negative control is RNA or cDNA from healthy controls and / or a template-free control whose comprehensive target gene score is verified to be lower than a preset threshold. The calibrator is an RNA or cDNA reference material containing the target gene and the internal reference gene and having a pre-assigned expression value, used to calibrate the Ct value, internal reference corrected expression value, or comprehensive score of different detection batches.

[0014] Preferably, the data interpretation instructions state that: the expression values ​​of each target gene are corrected and standardized by internal references, and a weighted comprehensive score of at least three target genes is calculated; when the comprehensive score reaches a preset threshold, an in vitro auxiliary detection result of positive host transcriptional characteristics related to active pulmonary tuberculosis is output.

[0015] The kit is used to detect peripheral blood mononuclear cells, stable whole blood samples, or total RNA extracted from them, and the total time for a single test does not exceed 3 hours.

[0016] Furthermore, a computer-implemented biomarker screening method is proposed, which uses the intersection of multiple algorithms such as differential expression analysis, WGCNA, LASSO, and random forest to screen candidate genes, and compares the model performance of different biomarker numbers in independent in vitro samples.

[0017] Furthermore, a comprehensive scoring model is proposed. For each target gene, the internal reference corrected expression value is calculated and standardized to a Z-value, according to S=Σw i Z i Calculate the overall score, where w i The non-negative weights are determined for the training set; when S is not lower than the threshold T, the positive result of the host transcription feature is output.

[0018] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a combination of host gene markers for active pulmonary tuberculosis, a screening method, and nucleic acid detection reagents and kits, which have the following beneficial effects: (1) By integrating three sets of public transcriptome data and verifying them independently by qRT-PCR, the bias of a single cohort was reduced.

[0019] (2) By comparing single markers to five markers step by step, we determine that three markers are the minimum optimal number for efficient detection, thus avoiding the instability caused by relying solely on a single gene.

[0020] (3) Biomarkers (ARID4B, CXorf38, HLA-DMB, HLA-DQA1 and SLC25A37) were identified by integrating public peripheral blood mononuclear cell transcriptome data, differential expression analysis, weighted gene co-expression network analysis, random forest, XGBoost, LASSO feature screening and independent sample qRT-PCR validation. According to the validation data, the optimal AUC for a single biomarker was 0.913. The AUCs for dual biomarkers, triple biomarker combinations (ARID4B, HLA-DMB, and SLC25A37), quadruple biomarker combinations (ARID4B, HLA-DMB, SLC25A37, and CXorf38), and pentabiomarker combinations (ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37) were 0.931, 0.946, 0.956, and 0.963, respectively. The sensitivity, specificity, and accuracy of the triple biomarker combination reached 90.0%, 87.5%, and 88.8%, respectively, indicating that the combined detection of at least three biomarkers can significantly improve the stability of identifying host transcriptional features associated with active pulmonary tuberculosis.

[0021] (4) Host gene marker combinations are suitable for conversion into qRT-PCR, digital PCR, microarray or targeted sequencing products, requiring less sample volume and the total detection time can be controlled within 3 hours. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 This is a transcriptome differential expression analysis; where A: differential gene volcano plot; B: differential gene heatmap.

[0024] Figure 2 Construct co-expression network analysis for WGCNA; where A: scale-free topology analysis and average connectivity analysis for soft threshold power selection; B: gene clustering tree diagram; C: module feature gene relationship diagram.

[0025] Figure 3 Modular gene analysis was constructed for WGCNA; where A: adjacency heatmap of module gene features; B: module-trait relationship heatmap. This demonstrates the correlation between each module gene feature and clinical traits (tuberculosis status).

[0026] Figure 4 Screening target genes for machine learning; where A: scatter plot of module membership and gene significance; B: confusion matrix of random forest; C: confusion matrix of XGBoost classifier; D: receiver operating characteristic (ROC) curve of classifier.

[0027] Figure 5 For modular gene analysis; where A: LASSO coefficient plot of modular genes; B: LASSO coefficient curve of modular genes; C: 10 gene features screened by LASSO regression; D: the top 15 important genes ranked by average Gini reduction index in the random forest model.

[0028] Figure 6 For independent sample qRT-PCR validation; where A: ARID4B; B: CXorf38; C: HLA-DMB; D: HLA-DQA1; E: SLC25A37. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1: Transcriptome Data Integration and Differential Expression Analysis Peripheral blood mononuclear cell transcriptome data from patients with active pulmonary tuberculosis and healthy controls in GSE19439, GSE28623, and GSE73408 were downloaded from the Gene Expression Overall (GEO) database. Background correction, quantile normalization, and log2 transformation were used to process the raw data. ComBat was used to correct for batch effects between datasets; when multiple probes corresponded to the same gene, the median expression value was used. A linear model was used to screen for differentially expressed genes, with the optimal threshold being |log2FC|≥2 and p<0.01. Example analysis revealed 6919 upregulated genes and 1341 downregulated genes (…). Figure 1 (A and B).

[0031] Example 2: WGCNA and Machine Learning Screening A co-expression network was constructed using WGCNA, with a soft threshold power β=10 and a minimum module gene count of 30. Figure 2 A). Using hierarchical clustering based on topological overlap and dynamic tree cutting, genes are grouped and assigned to different co-expression modules ( Figure 2 B). The relationship between module characteristic genes ( Figure 2 C) and characteristic gene adjacency analysis ( Figure 3 A) demonstrated high connectivity within modules and good independence between modules. Associating module-specific genes with clinical characteristics (TB=1, HC=0) identified 12 co-expression modules. Figure 3 B). The correlation coefficient between the turquoise module and the active pulmonary tuberculosis phenotype was r=0.75, p=2×10⁻⁶. -27 First, the overall diagnostic potential of this module was assessed. Gene importance analysis showed a high correlation between "module membership" and "gene importance" for the TB trait. Figure 4 A) confirmed the module's enrichment in functionally relevant genes. Subsequently, random forests and XGBoost classifiers constructed using the gene expression profiles of this module both exhibited perfect classification performance, with no error in the confusion matrix. Figure 4 (B, C), the area under the curve (AUC) is 1.00 ( Figure 4 D), validating the module's strong diagnostic potential. Then, two independent feature selection methods were applied to the 683 genes in this module. LASSO regression (10-fold cross-validation, optimal λ = 12.76) retained 10 genes with non-zero coefficients ( Figure 5 A, B, C). An independent random forest model ranked genes based on the importance of average Gini reduction and extracted the top 15 from them. Figure 5D). LASSO regression and random forest importance ranking were performed on the genes of this module, and the intersection of the two algorithms was taken to obtain AMD1, ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37. Combining expression significance and co-expression network connectivity, the relatively isolated AMD1 was excluded, resulting in five candidate host biomarkers.

[0032] Example 3 Independent Sample qRT-PCR Validation Peripheral blood samples were collected from 40 patients with active pulmonary tuberculosis and 40 age- and sex-matched healthy controls. Peripheral blood mononuclear cells were isolated, and total RNA was extracted and reverse transcribed into cDNA. Using cDNA as a template, qRT-PCR amplification was performed using the primers shown in Table 1. Each reaction had a total volume of 20 μL, including 10.0 μL of 2×SYBR Green qPCR Mix, 0.4 μL of 10 μmol / L forward primer, 0.4 μL of 10 μmol / L reverse primer, 2.0 μL of cDNA template, and 7.2 μL of nuclease-free water. The qRT-PCR reaction conditions were: 95℃ pre-denaturation for 30 s; followed by 40 cycles, each cycle consisting of 5 s at 95℃ and 30 s at 60℃. GAPDH was used as an internal control, and the relative expression levels of each target gene were calculated using the 2^-ΔΔCt method. ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37 were all significantly elevated in the active pulmonary tuberculosis group. Figure 6 ).

[0033] Table 1 qRT-PCR primer sequences

[0034] Example 4: Modeling Scheme with Increasing Number of Markers To avoid repeated parameter tuning on independent validation queues, model weights and thresholds should be fixed in the training queue. Candidate genes are first sorted by single-marker AUC, cross-dataset directional consistency, and biological complementarity, and then two, three, four, and five-marker models are constructed using a forward addition approach. The preferred addition order is ARID4B, HLA-DMB, SLC25A37, CXorf38, and HLA-DQA1. For each target gene, the internal reference corrected expression value is calculated and normalized to a Z-value, according to S=Σw. i Z i Calculate the overall score, where w iNon-negative weights are determined for the training set; when S is not lower than the threshold T, a positive result for host transcriptional features is output. An exemplary scoring formula for the joint model is shown in Table 2. Table 2 is used to convert the standardized expression values ​​of multiple target genes into a single comprehensive score. Specifically, the mean and standard deviation of the corrected expression values ​​of each target gene within the training cohort are used to Z-score the test samples, and the comprehensive score S is calculated according to the weights shown in Table 2. In the training cohort, active pulmonary tuberculosis samples are considered positive, and healthy control samples are considered negative. ROC curves are plotted based on the comprehensive scores, and the comprehensive score corresponding to the largest Youden index is used as the decision threshold T. This threshold is fixed during the independent validation phase and not adjusted further. For the test sample, when the comprehensive score S ≥ the threshold T of the corresponding model, it is determined to be positive for active pulmonary tuberculosis-related host transcriptional features; when S < T, it is determined to be negative. For example, the decision threshold T for the three-marker model is 0.43; when S ≥ 0.43, it is determined to be positive, and when S < 0.43, it is determined to be negative. Based on the corresponding judgment threshold, the test sample is interpreted as positive or negative, thereby realizing the quantitative and standardized output of the joint detection results of multiple markers.

[0035] Table 2. Exemplary scoring formulas for the joint model

[0036] Example 5: Comparison of detection efficacy between single markers and combined markers In 80 independent validation samples, 40 were active tuberculosis patients and 40 were healthy controls. Blind interpretation was performed using weights and thresholds fixed during the training phase. The confusion matrix and detection power of each model were calculated using the following formulas: Sensitivity = TP / (TP+FN), Specificity = TN / (TN+FP), Accuracy = (TP+TN) / Total Number of Samples, Positive Predictive Value = TP / (TP+FP), Negative Predictive Value = TN / (TN+FN). AUC was obtained by plotting ROC curves using continuous expression values ​​or comprehensive scores for each sample, with the false positive rate (1-specificity) on the x-axis and sensitivity on the y-axis, and the AUC obtained by calculating the area under the ROC curve. The final classification result of each model was determined using a threshold T predetermined during the training phase.

[0037] Table 3 Detection efficacy for different marker quantities

[0038] Note: M2 = ARID4B + HLA-DMB; M3 = M2 + SLC25A37; M4 = M3 + CXorf38; M5 = M4 + HLA-DQA1. TP: True positive, active pulmonary tuberculosis is judged as positive; FN: False negative, active pulmonary tuberculosis is misjudged as negative; TN: True negative, healthy controls are judged as negative; FP: False positive, healthy controls are misjudged as positive; PPV: Positive predictive value; NPV: Negative predictive value.

[0039] The results showed that the AUC of the five single biomarkers ranged from 0.872 to 0.913, with an accuracy of 81.2% to 83.8%. The AUC of the two-biomarker combination improved to 0.931, but there were still 5 false negatives and 6 false positives. The AUC of the three-biomarker combination reached 0.946, with a sensitivity of 90.0%, specificity of 87.5%, and accuracy of 88.8%, reducing misclassified samples by 5 compared to the optimal single biomarker, meeting the preset minimum power standard. The four-biomarker combination further improved the accuracy to 91.3%, and the five-biomarker combination improved both specificity and accuracy to 92.5%.

[0040] Considering overall detection efficiency, number of genes, reagent cost, and model robustness, the minimum preferred solution of this invention is the combined detection of at least three biomarkers; when pursuing higher specificity or requiring scientific stratification, a combination of four or five biomarkers is preferred.

[0041] The host gene biomarker combination of the present invention can be made into qRT-PCR kits, digital PCR kits, chips, targeted sequencing panels or data analysis systems, which are suitable for in vitro auxiliary detection of host transcriptional characteristics related to active pulmonary tuberculosis, scientific research stratification and efficacy monitoring studies, and have clear prospects for industrialization and clinical testing translation.

[0042] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A combination of host gene markers for active pulmonary tuberculosis, characterized in that, The host genetic marker combination includes at least three of ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37.

2. The combination of host gene markers for active pulmonary tuberculosis according to claim 1, characterized in that, The host genetic marker combination consists of ARID4B, HLA-DMB, and SLC25A37.

3. The combination of host gene markers for active pulmonary tuberculosis according to claim 1, characterized in that, The host genetic marker combination consists of ARID4B, HLA-DMB, SLC25A37, and CXorf38.

4. The host gene marker combination for active pulmonary tuberculosis according to claim 1, characterized in that, The host genetic marker combination consists of ARID4B, CXorf38, HLA-DMB, HLA-DQA1, and SLC25A37.

5. The use of the host gene marker combination according to any one of claims 1 to 4 in the preparation of products for in vitro auxiliary detection of active pulmonary tuberculosis, risk stratification, efficacy monitoring studies or host transcriptional characterization.

6. A nucleic acid detection reagent for in vitro auxiliary detection of active pulmonary tuberculosis, characterized in that, The nucleic acid detection reagent includes primers, probes, or combinations thereof for detecting the expression level of the host gene marker combination as described in any one of claims 1 to 4; the probe is a chip probe, a digital PCR probe, or a sequencing capture probe.

7. A nucleic acid detection reagent for in vitro auxiliary detection of active pulmonary tuberculosis according to claim 6, characterized in that, The nucleic acid detection reagent includes the target gene primers shown in SEQ ID NO.1 to SEQ ID NO.10, and the GAPDH internal reference primers shown in SEQ ID NO.11 to SEQ ID NO.

12.

8. A kit for the in vitro auxiliary detection of active pulmonary tuberculosis, characterized in that, It includes the nucleic acid detection reagent as described in claim 6 or 7, as well as one or more of the following: RNA extraction reagent, reverse transcription reagent, nucleic acid amplification reagent, positive control, negative control, calibrator, and data interpretation instructions.

9. The reagent kit according to claim 8, characterized in that, The data interpretation instructions state that: the expression values ​​of each target gene are corrected and standardized by internal references, and a weighted comprehensive score of at least three target genes is calculated; when the comprehensive score reaches a preset threshold, an in vitro auxiliary detection result of positive host transcriptional characteristics related to active pulmonary tuberculosis is output.

10. A computer-based method for screening combinations of host gene markers for active pulmonary tuberculosis, characterized in that, include: Transcriptome data from isolated peripheral blood mononuclear cells of patients with active pulmonary tuberculosis and healthy controls were obtained; standardization and batch effect correction were performed; differentially expressed genes were identified; disease-related modules were obtained through weighted gene co-expression network analysis; candidate genes were screened using at least two machine learning feature selection algorithms; candidate genes were validated by nucleic acid testing of independent isolated samples; and the AUC, sensitivity, specificity, and accuracy of single biomarkers and different numbers of combined biomarkers were compared to determine the detection combination containing at least three biomarkers.

Images