A biomarker, kit and application thereof for HIV / Mtb co-infection immune state evaluation and clinical diagnosis

Abstract

The application relates to the field of medical diagnosis, and particularly discloses a biomarker combination, a kit and application thereof for HIV / Mtb co-infection immune state evaluation and clinical diagnosis. Peripheral blood samples of HIV infectors, Mtb infectors and HIV / Mtb co-infectors are subjected to multi-factor detection, and Lasso regression is adopted to screen out cytokines such as FGF-2, IL-7, I-309, TNF-alpha, MMP-1 and MMP-7 closely related to HIV / Mtb co-infection, so as to construct a multi-factor diagnosis model. The model shows high sensitivity, specificity and AUC value in the training set and the test set, can effectively distinguish HIV single infection from HIV / Mtb co-infection individuals, and is particularly suitable for the diagnosis of patients with low CD4 + T cell level. The biomarker combination and the kit prepared by the biomarker combination provided by the application can be used as a supplement to immunological detection methods, and provide a new technical means for the clinical diagnosis and immune state evaluation of HIV / Mtb co-infection.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostics, specifically to a biomarker, kit, and its application for assessing the immune status and clinical diagnosis of HIV / Mtb co-infection. Background Technology

[0002] Human Immunodeficiency Virus (HIV) combined with Mycobacterium tuberculosis ( Mycobacterium tuberculosis, Mtb infection is a priority for tuberculosis outbreak prevention and control. HIV-infected individuals are approximately 14 times more likely to develop tuberculosis than uninfected individuals. According to the 2024 Global Tuberculosis Report, there were 10.8 million new tuberculosis cases globally in 2023, of whom 662,000 were co-infected with HIV. Tuberculosis causes 1.25 million deaths, including 161,000 HIV-positive individuals, accounting for 12.9% of all tuberculosis-related deaths. Therefore, timely, accurate, and feasible screening and diagnostic strategies can ensure timely intervention in individuals co-infected with HIV / Mtb and effectively reduce hospitalization and mortality rates.

[0003] The currently accepted gold standard for diagnosing tuberculosis remains a positive Mtb culture or a histopathological biopsy consistent with Mtb infection pathology. However, due to the complexity of the condition in HIV / Mtb co-infected individuals and the difficulty in obtaining biopsy specimens, traditional detection methods may lead to missed or misdiagnosed cases, delaying diagnosis and treatment. Immunological detection techniques (including the tuberculin skin test (TST), T-SPOT.TB, QFT-GIT, etc.) have developed rapidly in recent years and have been widely used in the clinical diagnosis of Mtb infection. However, in HIV-infected individuals, the positive rate of TST for tuberculosis detection is only 19.1%, and the sensitivity is only 35.96%, and with the decline of CD4+... + A decrease in T lymphocyte count gradually reduces the positivity rate. Similarly, QFT-GIT has low sensitivity in diagnosing HIV / Mtb co-infection, and the test results may be affected by CD4 count. +T cell count is affected, therefore, it is necessary to seek new strategies to improve the diagnostic performance of the experiment. The sensitivity of T-SPOT.TB in HIV / Mtb co-infection detection is higher than that of TST, but T-SPOT.TB cannot effectively distinguish LTBI and ATB, and the effect is poor in the prediction of ATB. Emerging rapid and instant molecular detection (including Xpert MTB / RIF detection, linear probe detection, etc.) can efficiently realize MTB gene identification and drug resistance mutation gene detection, however, the cost is high, which limits its application in resource-limited areas, and reducing the cost may be the primary problem to be solved for future large-scale application. SUMMARY

[0004] The purpose of the present application is to overcome the defects of the existing HIV / Mtb co-infection diagnosis technology, such as insufficient sensitivity, low specificity, and poor diagnostic accuracy in patients with low CD4 + T cell level, and provide a new scheme for immune status evaluation and clinical diagnosis based on a combination of multiple factor biomarkers.

[0005] To achieve the above-mentioned purpose, the present application provides a biomarker combination, a kit and its application for HIV / Mtb co-infection immune status evaluation and clinical diagnosis. The biomarker combination comprises at least one or more of the following: fibroblast growth factor-2 (FGF-2), interleukin-7 (IL-7), chemokine I-309, tumor necrosis factor-α (TNF-α), matrix metalloproteinase-1 (MMP-1) and matrix metalloproteinase-7 (MMP-7). The present application realizes effective differentiation of HIV infection alone, Mtb infection alone and HIV / Mtb co-infection by performing Luminex multi-factor detection on peripheral blood samples, combining Lasso regression model for feature screening, and using machine learning methods such as XGboost to construct a diagnosis model. Specifically:

[0006] In a first aspect of the present application, a biomarker combination for HIV / Mtb co-infection immune status evaluation and clinical diagnosis is provided, and the biomarker comprises at least one or more of the following: fibroblast growth factor-2 (FGF-2), interleukin-7 (IL-7), chemokine I-309, tumor necrosis factor-α (TNF-α), matrix metalloproteinase-1 (MMP-1) and matrix metalloproteinase-7 (MMP-7).

[0007] In an embodiment, the combination contains FGF-2, IL-7, I-309 and TNF-α.

[0008] In an embodiment, further comprising MMP-1 and / or MMP-7.

[0009] In a second aspect of the present application, an in vitro detection kit for detecting HIV / Mtb co-infection is provided, wherein the kit comprises detection reagents for detecting the biomarkers.

[0010] In an embodiment, the detection reagents are selected from ELISA reagents, immunochromatography reagents, flow cytometry reagents or Luminex multi-factor detection reagents.

[0011] In a third aspect of the present application, the use of the above biomarker combination in the preparation of a product for the evaluation of the immune status of HIV / Mtb co-infection is provided.

[0012] In a fourth aspect of the present application, the use of the above biomarker combination in the preparation of a product for the clinical diagnosis of HIV / Mtb co-infection is provided.

[0013] In a fifth aspect of the present application, the use of the above biomarker combination in the preparation of a product for the diagnosis of HIV / Mtb co-infection in a population of HIV-infected patients with low CD4 + T cell count (≤200 / µL) is provided.

[0014] In an embodiment, the product is a kit.

[0015] Compared with the prior art, the present application has the following beneficial effects:

[0016] 1. The multi-factor combined detection significantly improves the diagnostic sensitivity and specificity for HIV / Mtb co-infection, and the diagnostic performance is better than that of the traditional TST, QFT-GIT and T-SPOT.TB methods;

[0017] 2. The XGboost multi-factor diagnostic model established in the present application has an AUC greater than 0.80 in both the training set and the test set, and has stable and reliable diagnostic ability;

[0018] 3. In particular, in the patient population with CD4 + T cell count ≤200 / µL, the diagnostic performance of the multi-factor model of the present application is significantly better than that of the IGRA detection method, solving the problem of difficult diagnosis of patients with low immune status;

[0019] 4. The biomarker combination of the present application can be widely applied to the development of in vitro detection kits, and has strong generalizability and clinical application prospect.

[0020] Therefore, the present application not only provides a new technical approach for the clinical diagnosis of HIV / Mtb co-infection, but also provides a solid foundation for the improvement and popularization of immunological detection methods, and has important public health value and industrial application potential. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of this specification, illustrate embodiments of the application, and are meant to explain the present application, but are not intended to limit the application. In the drawings:

[0022] Figure 1 Variable selection plot based on Lasso regression model, where Figure 1 A: Association between log(λ) of variables included in Lasso analysis and regression coefficients; Figure 1 B: Process of screening the best λ value in Lasso regression model by 10-fold cross-validation method;

[0023] Figure 2 ROC plot for cytokines for HIV / Mtb co-infection diagnosis; where Figure 2 A: ROC plot for FGF-2 for HIV / Mtb co-infection diagnosis; Figure 2 B: ROC plot for IL-7 for HIV / Mtb co-infection diagnosis; Figure 2 C: ROC plot for I-309 for HIV / Mtb co-infection diagnosis; Figure 2 D: ROC plot for TNF-α for HIV / Mtb co-infection diagnosis; Figure 2 E: ROC plot for MMP-1 for HIV / Mtb co-infection diagnosis; Figure 2 F: ROC plot for MMP-7 for HIV / Mtb co-infection diagnosis;

[0024] Figure 3 Machine learning model performance evaluation in training and test set data; where Figure 3 A: Accuracy, sensitivity, specificity, precision and F1 score of 6 machine learning models in training set data; Figure 3 B: Accuracy, sensitivity, specificity, precision and F1 score of 6 machine learning models in test set data;

[0025] Figure 4 ROC curve of diagnostic model in training and validation set data;

[0026] Figure 5 Variable importance ranking of XGboost model;

[0027] Figure 6 ROC curve of multi-factor model and IGRA in patients with CD4 + T cell ≤ 200 / µL; where Figure 6 A: ROC curve and AUC of multi-factor model in patients with CD4 + T cell ≤ 200 / µL; where Figure 6B: IGRA in CD4 + ROC curve and AUC in patients with T cells <200 cells / µL. DETAILED DESCRIPTION

[0028] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, it should be understood that the preferred embodiments described here are only used to illustrate and explain the present application, and are not used to limit the present application.

[0029] Example 1 HIV / Mtb co-infection immune status evaluation and clinical diagnosis marker screening

[0030] 1. Study subjects

[0031] This study was approved by the Ethics Committee of Beijing You'an Hospital, Capital Medical University (Ethics Approval No: JYKL

[2023] 020), HIV-infected and HIV / Mtb co-infected patients treated in the Department of Infection and Immunology of Beijing You'an Hospital from January 2023 to June 2024, Mtb-infected patients treated in the Department of Respiratory and Critical Care Medicine, and healthy people of the same age and gender who underwent physical examination at the same time were recruited. Subjects who met the inclusion and exclusion criteria were included in this study. A total of 90 HIV-infected patients and 81 HIV / Mtb co-infected patients were finally included. The subjects were randomly divided into training set and test set in the ratio of 7:3. The training set was used to construct the Lasso regression model and the subsequent diagnostic model, and the test set was used to test the accuracy of the constructed model.

[0032] After fully informing the research process and the risks and benefits of the subjects, the researchers obtained the informed consent of the subjects and collected relevant clinical data (including age, gender, HIV diagnosis time, HIV viral load, HBV, HCV, and syphilis co-infection, blood routine, liver function, ART start time and medication, anti-tuberculosis treatment, and other medical history data) and peripheral blood specimens. The specific inclusion and exclusion criteria are as follows:

[0033] Inclusion criteria:

[0034] (1) Age 18 years old ≤ age ≤ 65 years old, men and women are not limited;

[0035] (2) The diagnosis of HIV infection meets any of the following conditions: ① HIV-1 / 2 antibody screening test positive and HIV supplementary test (including antibody confirmation test and nucleic acid test) result positive; ② Epidemiological history, both nucleic acid test results are positive; ③ HIV isolation test result is positive.

[0036] (3) HIV infected subjects enrolled in the study also had similar respiratory or systemic symptoms (including symptoms such as fever, cough, sputum, etc.) as Mtb infection, and were excluded from Mtb infection (IGRA negative);

[0037] (4) Diagnosis of AIDS stage: meet the diagnosis of HIV infection, and CD4 + T cell count is less than 200 / µL. Or meet the diagnosis of HIV infection plus any one of the following: ① Unexplained persistent irregular fever above 38℃ for more than 1 month; ② Diarrhea for more than 1 month; ③ Body weight loss of more than 10% within 6 months; ④ Recurrent oral fungal infection, herpes simplex virus infection or herpes zoster virus infection; ⑤ Pneumocystis pneumonia, ATB or non-tuberculous mycobacteriosis and other opportunistic infections; ⑥ Recurrent bacterial pneumonia; ⑦ Deep fungal infection; ⑧ Central nervous system space-occupying lesions; ⑨ Dementia in young and middle-aged people; ⑩ Active cytomegalovirus infection; Toxoplasma encephalopathy; Marnifi basket fungus disease; Recurrent sepsis; Kaposi's sarcoma, lymphoma.

[0038] (5) The diagnosis of latent tuberculosis mainly depends on clinical manifestations and IGRA results. Patients without tuberculosis-related clinical manifestations (fever, cough, hemoptysis, night sweats, weight loss, chest pain, fatigue, dyspnea) and IGRA positive results are diagnosed as latent tuberculosis.

[0039] (6) ATB is mainly determined by comprehensive judgment of clinical manifestations, imaging examination results and etiological detection results, and at least one of the following conditions is met for diagnosis: ① Etiological evidence: positive results of sputum acid-fast staining; positive culture of Mycobacterium tuberculosis in sputum, bronchoalveolar lavage fluid and pleural effusion; positive detection of Mtb-DNA in sputum, bronchoalveolar lavage fluid and pleural effusion; positive detection of Mycobacterium tuberculosis and rifampicin resistance detection (Xpert MTB / RIF); ② Pathological evidence: biopsy tissue shows tuberculous granuloma or caseous necrosis, and positive acid-fast staining;

[0040] (7) The diagnosis of HIV / Mtb co-infection mainly includes the following two cases: ① Diagnosis of AIDS patients with ATB: meet the diagnosis criteria of AIDS stage and ATB; ② HIV infection combined with latent tuberculosis infection: meet the diagnosis criteria of HIV infection, without tuberculosis-related clinical manifestations, and IGRA results are positive.

[0041] (8) Healthy controls: ① HIV-1 / 2 antibody screening test negative; ② No clinical manifestations of Mtb infection, IGRA results negative;

[0042] (9) No anti-tuberculosis treatment or less than 1 month of anti-tuberculosis treatment;

[0043] (10) Fully understand the purpose and requirements of this trial, voluntarily sign a written informed consent form.

[0044] Exclusion criteria:

[0045] (1) The investigator judged that there were any unstable or serious cardiovascular, renal, liver, blood, tumor, endocrine metabolism, mental or rheumatic diseases, so that they were not suitable for participating in this study.

[0046] 2. Peripheral blood collection and plasma separation

[0047] According to the standard method, 20 mL of peripheral venous blood of the enrolled subjects was collected in EDTA anticoagulant vacuum blood collection tube without separation gel. After blood collection, the blood collection tube was stored in the 4 ℃ refrigerator for a short time, and the peripheral blood was separated and stored within 4 hours.

[0048] 3. Luminex multi-factor detection

[0049] (1) All reagents were equilibrated to room temperature (20 ~ 25℃) before detection;

[0050] (2) Plasma sample preparation: Take the frozen plasma sample from the -80 ℃ freezer, vortex mix after the sample is completely thawed on the vortex shaker, then place the sample in a micro high-speed centrifuge, centrifuge at 13000 × g / min for 10 min, carefully aspirate 50 µL of the upper clear liquid into a new sterile EP tube, avoiding aspirating particles or the lower lipid layer;

[0051] (3) Preparation of pre-coupled antibody magnetic beads: vortex mix the pre-coupled antibody magnetic bead reagent for 1 min;

[0052] (4) Preparation of washing buffer: take 10 × Wash Buffer to room temperature, vortex mix, then take 60 mL 10 × Wash Buffer to 540 mL pure water, mix well for use (unused part can be stored in the 4 ℃ refrigerator for 1 month);

[0053] (5) Standard reconstitution: Before detection, reconstitute the PLEX pedition standard according to the kit instructions. ① Take out the 7 PLEX pedition standards in the kit, gently invert several times to mix the samples in the bottle, then place the standards in the centrifuge for a short time. After centrifugation, the standards are left at room temperature for 10 min. Then add 25 μL of pure water to the standards and vortex to mix well; ② Take out 20 μL of each reconstituted standard and transfer it to a sterile EP tube, and add Assay buffer to make the final volume 200 μL (Standard 7);

[0054] (6) Standard preparation: Prepare 6 sterile EP tubes and label them as Standard 1 to Standard 6. Add 150 μL of Assay buffer to each of the 6 EP tubes, and add 50 μL of Standard 7 solution to the EP tube labeled Standard 6. Mix well and then pipette 50 μL into the EP tube labeled Standard 5. Mix well and then pipette 50 μL into the EP tube labeled Standard 4. Dilute the subsequent standards by 4-fold gradient, a total of 7 standards. The specific dilution process is shown in the table below:

[0055] Table 1 Standard preparation process

[0056]

[0057] (7) Take out the 96-well plate in the kit and arrange the samples according to the kit instructions. Then add 200 μL of washing buffer to each well of the 96-well plate. Seal the 96-well plate and incubate it at room temperature for 10 min on a plate shaker. Then discard the washing buffer and invert the 96-well plate on absorbent paper to absorb the residual liquid in the wells;

[0058] (8) Add 25 μL of Assay buffer to the 96-well plate. According to the pre-designed sample arrangement, add 25 μL of Assay buffer, standards and samples to the corresponding blank wells, standard wells and sample wells, respectively. Then add 25 μL of pre-coupled antibody magnetic beads to each well of the 96-well plate;

[0059] (9) Seal the 96-well plate with plate sealing film, then place it in a plate shaker at 2-8 °C for 16-18 h;

[0060] (10) Place the 96-well plate in a handheld magnetic stand for 60 s to allow the magnetic beads to settle completely at the bottom of the 96-well plate. Then gently discard the contents of the 96-well plate (the 96-well plate remains attached to the handheld magnetic stand during this process) and invert the 96-well plate on absorbent paper to absorb the residual liquid in the wells;

[0061] (11) Take off the 96-well plate from the handheld magnetic stand, add 200 μL of the washing buffer to each well, and shake on the plate shaker for 30 s to rinse the magnetic beads;

[0062] (12) Repeat steps (10) and (11) twice;

[0063] (13) Add 25 μL of the detection antibody to each well of the 96-well plate, seal the plate with the tin foil, and incubate on the plate shaker at room temperature (20-25 °C) for 1 h;

[0064] (14) Repeat the washing of the 96-well plate according to steps (10) and (11) for 3 times;

[0065] (15) Add 25 μL of the phycoerythrin-labeled streptavidin to each well of the 96-well plate, seal the plate with the tin foil, and incubate on the plate shaker at room temperature (20-25 °C) for 30 min;

[0066] (16) Repeat the washing of the 96-well plate according to steps (10) and (11) for 3 times;

[0067] (17) Add 150 μL of the sheath solution to each well of the 96-well plate, then shake on the plate shaker for 5 min to resuspend the magnetic beads, and then detect the cytokine levels in the plasma sample on the instrument;

[0068] (18) The instrument reads the fluorescence value of the standard sample well, obtains the fitting curve by using the recommended 5-parameters logistic method, reads the fluorescence value of the sample to be detected, and substitutes it into the 5-parameters logistic fitting curve to calculate the concentration of each marker;

[0069] (19) Data quality control standards: ① R2 of the standard curve > 0.99; ② Coefficient of variation within the batch < 15%, and inter-batch variation effect < 20%.

[0070] 4. Evaluation of the immune status of cytokines and construction and verification of the clinical diagnosis model

[0071] Based on the cytokine detection data of HIV / Mtb co-infected persons and HIV infected persons with corresponding respiratory and systemic symptoms, we explored the value of cytokines in the immune status evaluation and clinical diagnosis of HIV / Mtb co-infection.

[0072] 4.1 Data segmentation

[0073] Based on the "sample.split" function in the "Caret" package, we divided the subjects into training and test sets in a ratio of 7:3. The data in the training set were used to build the machine learning model, and the data in the test set were used to validate the model.

[0074] 4.2 Construction of the least absolute shrinkage and selection operator (Lasso) regression model for feature variable screening

[0075] (1) Standardize the training set data using the "scale" function;

[0076] (2) Use the "glmnet" function to build the Lasso model, and select "binomal" for the family parameter;

[0077] (3) Use the "cv.glmnet" function for 10-fold cross-validation to select the most appropriate lambda value. Select lambda.1se in the output results as the optimal lambda for the final model construction;

[0078] (4) The variables with non-zero coefficients in the final model output are the most relevant feature variables for HIV / Mtb co-infection diagnosis in this study.

[0079] The results are shown in Figure 1 , as shown in Figure 1 A shows the regression coefficients of the cytokine variables included in the Lasso analysis with respect to the Lambda value. As the parameter Lambda increases, the number of variables with regression coefficients of 0 increases. 10-fold cross-validation is used to select the most appropriate Lambda value, and the results are shown in Figure 1 B. Select Lambda = 1se as the best Lambda value to obtain the best Lasso regression model. The most relevant feature variables for HIV / Mtb co-infection, as output by the model, include FGF-2, IL-7, I-309, TNF-alpha, MMP-1, and MMP-7. Mtb

[0080] Example 2: Immune status evaluation and clinical diagnosis model construction and evaluation

[0081] ​Based on the training set data, we constructed six models, including support vector machines (SVM), logistic regression (LR), eXtreme gradient boosting (XGboost), random forest (RF), Naive Bayes (NB), and neural network (NNET), to explore the value of the selected variable combinations in the evaluation and clinical diagnosis of HIV / Mtb co-infection. Subsequently, we used the test set data to evaluate the models:

[0082] (1) The "svm" function in the "e1071" package was used to construct the SVM model, the "lrm" function in the "rms" package was used to construct the LR model, the "xgboost" package was used to construct the XGboost model, the "randomForest" package was used to construct the RF model, the "NaïveBayes" function in the "e1071" package was used to construct the NB model, and the "nnet" package was used to construct the NNET model.

[0083] (2) After the initial model construction, we used 10-fold cross-validation and grid search to optimize the model parameters, and then constructed the final model based on the optimized parameters.

[0084] (3) We evaluated the models using accuracy, sensitivity, specificity, precision, F1 score, and Area Under Curve (AUC), and then selected the optimal diagnostic model based on the evaluation results.

[0085] The results, as shown in Figure 2 FGF-2, IL-7, I-309, and TNF-α are potential markers for the evaluation and clinical diagnosis of HIV / Mtb co-infection, and can be used to distinguish HIV-infected individuals from HIV / Mtb co-infected individuals. The AUC of FGF-2, IL-7, I-309, and TNF-α are 0.748, 0.731, 0.797, and 0.726, respectively, while the AUC of MMP-1 and MMP-7 are less than 0.70 Figure 2 when used alone for the evaluation and clinical diagnosis of HIV / Mtb co-infection (A-F). The effect of these cytokines may not be satisfactory when used alone for the evaluation and clinical diagnosis of HIV / Mtb co-infection.

[0086] Based on the above results, we consider that the combination of multiple cytokines may improve the HIV / Mtb To evaluate the sensitivity and specificity of the immune status and clinical diagnosis, we used the feature variables screened by the Lasso regression model to construct various machine learning models (including SVM model, LR model, XGboost model, RF model, NB model, and NNET model), and compared the accuracy, sensitivity, specificity, precision, F1 score, and AUC of the constructed models in the training set and validation set data to screen the best immune status evaluation and clinical diagnosis model.

[0087] XGBoost is an ensemble learning model based on gradient boosting decision trees (GBDT). Each decision tree ft is not directly outputting a class label, but a score (or "log odds"). For a sample, the model adds up the output scores of all trees to get the total score of the sample. The core of this model is to realize prediction through weighted ensemble of multiple decision trees. For binary or multi-classification tasks (such as distinguishing healthy controls, HIV-only infection, and HIV / Mtb co-infection), the final prediction formula can be expressed as:

[0088]

[0089] In this formula, S represents the total score of each decision tree, f t represents the function used for different applications, and Σ is the mathematical summation symbol, m represents the index category (0 represents "healthy control" class, 1 represents "HIV-only infection" class, and 2 represents "HIV / Mtb co-infection" class), f t m represents the tth decision tree under the mth category, f(x) is the feature vector, i.e., the test value group of the patient.

[0090] According to this formula, assuming there are K decision trees, for a sample, the feature vector of the sample is input f ( x ), and the final total score is calculated by the corresponding function f t (Sigmoid function for binary classification, Softmax function for multi-classification). In actual operation, the model will convert the obtained total score SConvert to the probability of belonging to each class through the Softmax function, and finally select the class with the maximum probability as the final prediction label and output the prediction result.

[0091] In short, XGBoost integrates multiple weak learners through additive models (adding the prediction values of multiple trees), and then converts the aggregated scores into probabilities through the Sigmoid (binary classification) or Softmax (multi-classification) function, and finally makes a classification prediction according to the probability.

[0092] The XGBoost model outputs a classification result or probability by inputting the concentrations of key immune factors related to HIV / Mtb infection (x) selected in the pre-screening process, thereby assisting in the evaluation of the immune status and clinical diagnosis of HIV infection and HIV / Mtb co-infection. The specific process is as follows:

[0093] 1. Selection and standardization of input features.

[0094] The input of the model is the key immune factors (FGF2, IL7, I309, TNF-a, MMP-1, MMP-7) related to HIV / Mtb infection selected in the pre-screening process. Before input, the features need to be standardized (such as z-score processing) to ensure that factors of different magnitudes are given fair weight in the model.

[0095] 2. Prediction process of the model and association with immune status.

[0096] The model performs "layer-by-layer analysis" on the input features through multiple decision trees: each decision tree splits the sample according to the feature threshold, and finally falls into a leaf node, outputting the weight of the node. After weighting and summing the outputs of all decision trees, the probability value is converted through the activation function.

[0097] Specifically, after inputting the selected meaningful cytokines (FGF2, IL7, I309, TNF-a, MMP-1, MMP-7), the model outputs a continuous probability value (e.g. 0.23, 0.78). The prediction model converts this probability value into the final classification decision (0 or 1, i.e. "uninfected" or "infected"). The prediction rule is determined according to the preset threshold. In this embodiment, the preset threshold is > 70% for "high risk", 30% ~ 70% for "medium risk", and < 30% for "low risk".

[0098] The results are shown in Table 1. Figure 3 As shown in Table 1, in the training set, the accuracy of the XGboost model is 0.875, the sensitivity is 0.842, the specificity is 0.905, the precision is 0.889, and the F1 score is 0.865. Figure 3A). In the test set, the accuracy of the XGboost model was 0.804, the sensitivity was 0.792, the specificity was 0.815, the precision was 0.792, and the F1 score was 0.792. Figure 3 B). In addition, the AUC of the XGboost model in the training set and the test set was 0.873 and 0.803, respectively. Figure 4 A and Figure 4 B). Compared with other models, the XGboost model performed well in both the training set and the test set, so we chose the XGboost model constructed with the feature variables screened by the Lasso regression model as the best model for the evaluation of the immune status of HIV / Mtb co-infection and clinical diagnosis. Mtb Subsequently, we evaluated the importance of the variables included in the XGboost model according to the SHAP value, and the results showed that I-309, IL-7, and FGF-2 were the top three variables that contributed most to the evaluation of the immune status of HIV / Mtb co-infection and clinical diagnosis. Figure 5 .

[0099] IGRA is a reliable indicator for the diagnosis of HIV / Mtb co-infection, but its sensitivity is low in patients with CD4 + T cell counts below 200 / µL. Therefore, we compared the diagnostic performance of the multi-cytokine model developed in this study with IGRA in patients with CD4 + T cells below 200 / µL. As shown in Figure 6 , the AUC of the multi-cytokine model for the diagnosis of Mtb infection was 0.815 in HIV-infected individuals with CD4 + T cells below 200 / µL, while the AUC of IGRA for the diagnosis of Mtb infection was 0.692. Our diagnostic model can provide a more reliable diagnosis for HIV / Mtb co-infected individuals, especially for patients with low CD4 + T cell counts.

[0100] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. The application of a reagent for detecting combinations of biomarkers in the preparation of clinical diagnostic products for differentiating between HIV infection and HIV / Mtb co-infection, characterized in that, The biomarker combination includes FGF-2, IL-7, I-309, and TNF-α, and the reagents for detecting the biomarker combination are selected from ELISA reagents, immunochromatographic reagents, flow cytometry reagents, or Luminex multifactor detection reagents.

2. The application according to claim 1, characterized in that, The biomarker combination further includes MMP-1 and / or MMP-7.

3. The application according to claim 1 or claim 2, characterized in that, The HIV-infected and HIV / Mtb co-infected populations are those with low CD4 counts. + T-cell count population.

4. The application according to claim 1 or claim 2, characterized in that, The clinical diagnostic product is a reagent kit.

Images