Children refractory mycoplasma pneumonia pneumonia early diagnosis marker and RT-qPCR (real-time quantitative polymerase chain reaction) detection method

Abstract

The invention discloses an early diagnosis marker for children refractory mycoplasma pneumonia (RMPP) and an RT-qPCR detection method, and relates to the technical field of disease detection. According to the invention, PBMC maps of a healthy control group in the early morbidity stage of RMPP sick children, in the early morbidity stage of mycoplasma pneumoniae pneumonia sick children and in the early morbidity stage of RMPP sick children are drawn for the first time by virtue of an scRNA-seq technology, and IGHM, NEAT1, IL32 and ACTG1 four-gene marker combination for early diagnosis of RMPP in children is innovatively discovered; a peripheral blood sampling scheme is combined to thoroughly solve throat swab sampling obstacles of young children; and a high-precision three-classification diagnosis model is constructed. According to the method disclosed by the invention, the diagnosis time is shortened from 7 days to 24 hours, the sensitivity is improved to 93-96%, the detection cost is reduced by 95%, a first molecular solution with high precision, low cost and clinical accessibility is provided for early RMPP intervention of children, and the severe case rate and medical burden are expected to be obviously reduced.

Description

Technical Field

[0001] The present invention belongs to the fields of biomedicine and disease detection technology, and in particular relates to an early diagnostic marker and RT-qPCR detection method for refractory Mycoplasma pneumonia (RMPP) in children. Background Art

[0002] 15%-40% of Mycoplasma pneumonia (MPP) cases progress to RMPP. Early identification is difficult, progression is rapid, and conventional macrolide therapy is ineffective. Without timely intervention, severe complications such as bronchiectasis and necrotizing pneumonia can occur, even leading to death, significantly increasing the medical burden and psychological stress on families. Therefore, early diagnosis of RMPP is crucial for preventing progression, reducing the risk of disability and mortality, and improving prognosis.

[0003] Currently, there are no clear, unified diagnostic criteria for RMPP in children, both domestically and internationally. The consensus view, proposed by Japanese researchers in 2008, is that high fever persists despite treatment with macrolide antibiotics for at least 7 days, and imaging findings gradually worsen. In 2013, the Chinese Medical Association's Pediatrics Branch concluded that RMPP should be diagnosed in children with MPP who, despite regular macrolide antibiotic treatment for ≥1 week, do not experience clinical symptom relief, persistent fever, severe cough, and worsening lung imaging findings.

[0004] Clinically, there is currently a lack of early diagnostic methods and technologies for RMPP, and only the detection technology for Mycoplasma pneumoniae (MP). MP detection is an important basis for diagnosing MPP and is also one of the prerequisites for diagnosing RMPP. Currently, MPP-related tests are divided into non-specific and specific methods. (I) MPP non-specific detection methods: (1) Routine examinations to assess the severity of infection or inflammation: such as white blood cell count, CRP, etc. (2) Imaging examinations: It is difficult to use for early diagnosis of MPP or RMPP. (3) Fiberoptic bronchoscopy: RMPP can be seen under the microscope with congestion and swelling of the bronchial mucosa, mucus plugs, etc., which makes it difficult to use for early diagnosis. (II) MPP specific detection methods: (1) MP culture: It is the gold standard, but the operation is complicated and time-consuming (7-10 days to 3-4 weeks), and its clinical application is limited. (2) Serological testing: It is the main means of clinical diagnosis of MPP. The ELISA method is sensitive, specific, rapid and economical. MP-IgM is produced 7-10 days after infection and reaches its peak 3-4 weeks later; IgG appears later. Diagnosis often requires a quadruple increase in the titer of two serum samples, which takes a long time and affects the clinical value. (3) Polymerase chain reaction (PCR): It can quickly (2-3 hours) amplify and detect MP-DNA. RT-qPCR has high sensitivity and can be detected at any stage of infection. The positive rate for infants and young children is higher than serology. However, throat swab sampling in infants and young children is often difficult due to poor tolerance, low cooperation, narrow oral space, and sensitive pharyngeal reflex, which affects the sample quality and the reliability of the results. As mentioned in the review, the existing MPP diagnostic methods can only detect MP infection and cannot accurately identify RMPP, let alone achieve its early identification.

[0005] The reasons why early diagnosis of RMPP is difficult are: (1) Symptom and sign mismatch: severe symptoms, mild lung signs, and significant imaging changes. The initial symptoms are often non-respiratory manifestations, which can easily lead to misdiagnosis or missed diagnosis. (2) Poor sampling compliance in young children: The main patient population (infants and preschoolers) has low tolerance and compliance, which often leads to sampling difficulties and sample failure, increasing the risk of false negatives and delayed treatment. (3) Poor antibody detection time: specific IgM requires 1 week to produce, and double serum is required for diagnosis, which is time-consuming and can easily delay treatment and progress to RMPP. (4) Complex drug resistance mechanism: drug resistance mutations are not limited to common sites (such as A2063G / A2064G), and existing test kits may miss detection; and the relationship between drug resistance mutations and RMPP is unclear (the presence of drug resistance mutations does not necessarily mean RMPP, and RMPP may not have drug resistance mutations).

[0006] Early diagnosis of RMPP faces two bottlenecks: Current research on early RMPP diagnosis is largely limited to retrospective analysis of clinical indicators, lacking innovation and unvalidated experimentally. The current mainstream rapid MP detection technology, RT-qPCR, relies primarily on throat swabs. However, the primary population affected by MPP (infants and preschoolers) suffers from low tolerance and poor compliance with swab sampling, which can easily lead to substandard specimen quality, increased false-negative rates in RT-qPCR, and delayed treatment, leading to RMPP progression in some patients. In summary, current clinical testing technologies for distinguishing RMPP, GMPP, and non-MPP infections primarily rely on clinical symptoms, imaging, and MP testing. These technologies suffer from insufficient sensitivity and specificity, and lack a comprehensive diagnostic model that can simultaneously distinguish between the three categories of "normal, common, and refractory" outcomes, leading to missed or misdiagnosed cases. Given the limited efficacy of existing clinical testing methods and the continued focus on homogenous case analysis (focusing on retrospective case analysis to identify potentially suggestive clinical features), new technologies are urgently needed to explore molecular markers for the early and accurate diagnosis of RMPP and MPP.

[0007] Single-cell transcriptome sequencing (scRNA-seq) was first reported in 2009. Unlike traditional high-throughput sequencing that captures the average value of a cell population, it sequences messenger RNA (mRNA) from individual cells. In 2017 and 2018, the technology was hailed as the "genetic microscope" by Nature and ranked as the top scientific breakthrough of the year by Science, respectively. scRNA-seq has provided groundbreaking insights into disease research and facilitated the discovery of novel diagnostic and prognostic biomarkers. Its advantages in clinical diagnostic biomarker discovery include: (1) Cell heterogeneity analysis: revealing gene expression differences between different cell subpopulations in the disease microenvironment and discovering rare cell markers missed by traditional methods; (2) Novel biomarker mining: identifying cell-specifically expressed genes (such as endothelial cell inflammatory factors) to provide highly specific targets for early diagnosis; (3) Unbiased exploration: comprehensive screening of disease-related pathways without relying on prior knowledge.

[0008] In recent years, a large number of scRNA-seq studies have focused on viral pneumonia, primarily coronavirus disease 2019 (COVID-19). Analysis of peripheral blood from patients infected with the novel coronavirus revealed a loss of non-classical CD14LowCD16High monocytes, an increase in HLA-DRLow classical monocytes, and a significant release of calreticulin (S100A8 / S100A9) in severe cases. Therefore, measuring plasma calreticulin levels and detecting reduced abundance of non-classical monocytes by flow cytometry can distinguish patients with severe disease. This demonstrates that scRNA-seq-analyzed cell subtypes and gene expression changes can be used for disease diagnosis. In contrast, only one scRNA-seq study has been conducted on bacterial pneumonia (secondary sepsis), which profiled changes in immune cells, cytokines, and pathways in peripheral blood mononuclear cells (PBMCs) of patients, revealing persistent immune dysfunction. There is also only one scRNA-seq study on fungal pneumonia (Pneumocystis jirovecii). Analysis of mouse lung tissue identified seven macrophage subsets, finding that Mmp12+ macrophages are enriched after infection and highly express pro-inflammatory factors in the bronchoalveolar lavage fluid of infected mice. Extensive literature research has revealed no studies using scRNA-seq to reveal changes in MPP or RMPP cell types / subpopulations or gene expression, either in children or adults. There are also no reports using this technology to explore early diagnostic markers for RMPP. Reflect the innovation of this invention ).

[0009] RT-qPCR technology principle: (1) Fluorescent labeling: Add fluorescent dyes (such as SYBR Green) (this method is used in the present invention) or labeled probes to the reaction; (2) Real-time monitoring: During each round of amplification, the enzyme-digested probe or dye binds to DNA and releases fluorescence, which is detected in real time by the instrument; (3) Quantitative analysis: The threshold cycle number (Ct value) is used to determine the amount of starting template: the smaller the Ct value, the more template there is, and the higher the gene expression. Advantages in clinical marker gene expression detection: (1) Ultra-high sensitivity: It can detect changes in gene expression at very low copy numbers, suitable for early screening; (2) Accurate quantification: Absolute or relative quantification (with internal reference correction) directly reflects expression levels; (3) High specificity: (a) The probe method has strong targeting, and (b) the dye method uses melting curves to verify purity; (4) Rapid and efficient: The test is completed within 2 hours, suitable for rapid clinical diagnosis.

[0010] Disadvantages of scRNA-seq technology in the discovery of clinical disease diagnostic markers: (1) Technical complexity: Samples must be protected from RNA degradation, and cell separation can easily introduce bias; (2) High cost and low throughput: Single samples cost thousands to tens of thousands of yuan, making it difficult to verify large samples; (3) Difficult data analysis: Experts are required to process massive amounts of data, and batch effects and annotation standards are not unified; (4) Limited sensitivity: The detection rate of low-expression genes is low, and key factors may be missed; (5) Difficult clinical translation: The discovered markers need to be verified by traditional technologies.

[0011] Disadvantages of RT-qPCR in detecting gene expression of clinical markers: (1) Dependence on primers: Improper primer / probe design can lead to false positives / negatives and inability to detect unknown variants; (2) High sample requirements: RNA is easily degraded and requires strict treatment, and inhibitors may affect amplification; (3) High cost: Probe synthesis and instrument maintenance costs are high; (4) Quantitative limitations: A stable internal reference gene is required, but it may fluctuate during disease.

[0012] In summary, scRNA-seq can discover new biomarkers at the cellular level and is particularly suitable for disease typing. However, high costs and technical barriers limit its immediate clinical application. RT-qPCR has become the gold standard for clinical gene expression testing due to its high sensitivity, accurate quantification, and rapid speed. However, design dependency, sample requirements, and cost remain bottlenecks. Summary of the Invention

[0013] To address the shortcomings of the aforementioned existing technologies, the present invention, using scRNA-seq technology, for the first time mapped the PBMC profiles of children with early-stage RMPP, children with common Mycoplasma pneumoniae pneumonia (GMPP), and healthy controls. This innovative discovery of a four-gene marker combination (IGHM, NEAT1, IL32, and ACTG1) for the early diagnosis of RMPP in children is crucial. Combined with a peripheral blood sampling protocol, this approach completely overcomes the hurdle of throat swab sampling in young children. Furthermore, a high-precision three-category (RMPP / GMPP / Normal) diagnostic model was constructed. Compared to existing technologies, the present method shortens diagnosis time from 7 days to 24 hours, increases sensitivity to 93-96% (a >30% improvement over clinical indicators), and reduces testing costs by 95% (compared to the tens of thousands of yuan cost of scRNA-seq). This provides the first molecular solution for early intervention of RMPP in children that combines high accuracy, low cost, and clinical accessibility, potentially significantly reducing the incidence of severe disease and the medical burden.

[0014] In order to achieve the above object, the present invention adopts the following technical solutions:

[0015] One of the objects of the present invention is to provide a gene marker combination, which includes IGHM, NEAT1, IL32 and ACTG1.

[0016] A second object of the present invention is to provide the use of the above-mentioned gene marker combination in the preparation of RMPP diagnostic products, early screening products, disease monitoring products and / or drug efficacy monitoring products.

[0017] A third object of the present invention is to provide a screening method for the above-mentioned gene marker combination, which includes the use of scRNA-seq technology and RT-qPCR technology.

[0018] Furthermore, the sample source in the screening method includes peripheral blood.

[0019] A fourth object of the present invention is to provide a product for the diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of RMPP, wherein the product comprises a reagent for detecting the above-mentioned gene marker combination.

[0020] Furthermore, the product also contains one or more pharmaceutically acceptable carriers or excipients.

[0021] Furthermore, the product form includes a kit.

[0022] A fifth object of the present invention is to provide a model for diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of RMPP, wherein the method for constructing the model comprises using the above-mentioned gene marker combination.

[0023] Furthermore, the model is based on scRNA-seq technology and RT-qPCR technology.

[0024] Furthermore, the model formula sets X1-X4 to represent the 2 of IGHM, NEAT1, IL32, and ACTG1, respectively. –ΔΔCt value, then:

[0025] P_GMPP=exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4)

[0026] / [1+exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4)

[0027] +exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)];

[0028] P_RMPP=exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)

[0029] / [1+exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4)

[0030] +exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)].

[0031] Compared with the prior art, the present invention has the following technical effects:

[0032] (1) Breaking through existing technical bottlenecks and achieving early molecular typing diagnosis of RMPP

[0033] 1. Solving the problem of biomarker discovery

[0034] Current limitations: RMPP lacks specific molecular markers, and clinical diagnosis relies on retrospective symptom / imaging analysis, which has low sensitivity (missed diagnosis rate >30%) and poor timeliness (delayed treatment ≥7 days).

[0035] Innovation and improvement:

[0036] (1) Single-cell sequencing technology to discover new markers: Shenzhen has more than 2.65 million children aged 0-14 years old (statistics in 2023), and the annual incidence of MPP is about 12%-16% (national data), which means that there are more than 30,000 new cases of MPP each year in the local area. Based on a 20% proportion of RMPP (conservative estimate), Shenzhen has more than 6,000 new high-risk children with RMPP each year. Shenzhen has a high proportion of newborns (the birth rate in 2023 was 1.75%, ranking first among megacities), and children under 3 years old account for more than 65% of MPP cases. The failure rate of throat swab sampling in this group exceeds 40% (data from Shenzhen Children's Hospital). Therefore, peripheral blood specimens were collected from children suspected of MPP in the respiratory department of the only tertiary-level A-class children's hospital in South China (Shenzhen Children's Hospital), which are rich in diversity and representative, and well compensate for the shortcomings of single-center cases. Peripheral blood samples were collected within 24 hours of admission and PBMCs were extracted and stored (maximum response to the early stage of the disease). MP-DNA and other tests and subsequent clinical treatment were used to confirm whether the child was a MP patient, and then differentiated into GMPP or RMPP. The stored samples were reviewed and regrouped ( Ensure that the diagnostic markers mined from sequencing samples are suitable for early diagnosis of the disease Break ). For the first time, scRNA-seq was performed on RMPP / GMPP / Normal PBMCs ( Figure 1-2 ), breaking through the blind spot of cell heterogeneity in traditional whole-body sequencing. Through high-resolution clustering ( Figure 2-3 ) Changes in cell type ratios ( Figure 4 ) and differential gene analysis ( Figure 5-7 ), found that the RMPP group has specific high expression gene groups (IGLC2, IGHM, HSP90B1, IL32, ACTG1, NEAT1, etc.) ( Figure 6 ) and the proportion of neutrophil subsets (Neu) were significantly increased (Neu proportion: RMPP group vs GMPP group, P < 0.01, Figure 7 B), where the Neu subpopulation marker gene MMP8 is specifically highly expressed in RMPP ( Figure 7 C), although it did not enter the final model due to primer specificity issues, it suggests that neutrophil-mediated immune microenvironment disorder is one of the core pathological characteristics of RMPP.

[0037] (2) Mechanism association verification: GO enrichment showed that differentially expressed genes were concentrated in the immune-inflammatory pathway ( Figure 2 E), such as IL32 (a proinflammatory cytokine that activates the NF-κB pathway), HSP90B1 (endoplasmic reticulum stress response, involved in protein folding), ACTG1 (γ-actin, driving cytoskeletal remodeling), and NEAT1 (a long noncoding RNA that regulates inflammatory gene transcription). These genes align with the immunopathological features of RMPP clinical symptoms, providing biological plausibility for the biomarkers.

[0038] 2. Overcoming sampling barriers for young children

[0039] Current defects: The failure rate of throat swab sampling in infants and young children is >40% (due to the narrow oral space and sensitive pharyngeal reflex), which leads to an increased false negative rate of RT-qPCR (described in the background technology).

[0040] Innovation and improvement:

[0041] (1) Peripheral blood instead of pharyngeal swabs: PBMCs were isolated from venous blood for testing, which circumvented the problems of narrow oral anatomical space and pharyngeal reflex sensitivity in young children. The sampling success rate was 100% (samples were successfully obtained from all 349 cases in the prospective cohort).

[0042] (2) Sample quality meets the requirements of single-cell sequencing ( Figure 1 ).

[0043] 2. Building a high-precision, practical diagnostic model

[0044] 1. Multi-gene combination improves classification efficiency

[0045] Current defects: A single clinical indicator (such as CRP or MP-DNA load) cannot distinguish the three-category outcomes of RMPP / GMPP / Normal (the sensitivity described in the background technology is <60%).

[0046] Innovation and improvement: After sample size estimation, a training cohort (n=295) was prospectively established, including 99 normal cases, 104 GMPP cases, and 92 RMPP cases. RT-qPCR was used to detect the expression levels of 6 candidate marker genes (obtained by scRNA-seq analysis and screening). Then, a multinomial logistic regression model was constructed to screen out the core combination of IGHM+NEAT1+IL32+ACTG1 ( Figure 10 Three-category diagnostic performance (One-vs-Rest strategy): Normal vs Rest 0.978; GMPP vs Rest 0.974; RMPP vs Rest 0.952. Overall three-category accuracy 88.5% ( Figure 11), significantly better than any single gene ( Figure 10 ). Subsequent re-prospective collection of validation set specimens (Normal group 18 cases, GMPP group 19 cases, RMPP group 17 cases) were performed for external validation, and the AUC of the validation set ( Figure 12 ) showed that the model still maintained excellent discrimination ability in time external validation, and no significant performance degradation was observed. Finally, the three-class modeling of the whole sample was performed using the specimens collected in the first and second prospective collections (training set + validation set), and the internal validation was performed using 1000 times of stratified bootstrap resampling, to obtain the robust performance indicators (95% CI, see Table 4), and the "box- average curve" graph was constructed using the discrete ROC dataset generated by bootstrap ( Figure 13 ). The difference in coefficients between the two models (training set model, whole sample model) was very subtle (see Table 5), further proving the excellent stability of the four-gene combined model.

[0047] 2. RT-qPCR detection adapts to clinical scenarios

[0048] Innovative improvements: low-cost RT-qPCR transformation scheme.

[0049] Design specific primers (Table 1), verify amplification efficiency by melting curve ( Figure 8 Eliminate double-peak genes MMP8 / SUB1).

[0050] Only 1 ml of peripheral blood is required, RNA extraction to result output ≤ 3 hours, detection cost < ￥200 / sample (reagent and consumable cost).

[0051] Supporting automated algorithm: input 2 –ΔΔCt values, real-time output of three-class probability and diagnosis conclusion ( Figure 12 ), compatible with conventional RT-qPCR instruments in primary hospitals.

[0052] (Three) Clinical value and technical innovation points

[0053] 1. Early intervention window moves forward

[0054] Specimens are collected and detected within 24 hours of admission, which is earlier than the time of serum antibody production (7-10 days), allowing RMPP to be identified in the treatment golden window period. It is expected to reduce the risk of RMPP progressing to severe conditions such as lung necrosis and bronchiectasis by >50% (mortality rate of complications described in background technology is 15-30%).

[0055] 2. Promote precision medicine practice

[0056] (1) First establish the association model of "PBMC atlas-RMPP gene tag" ( Figure 5-7 ). Reveal the expansion of Neu neutrophil subpopulations ( Figure 7B) and abnormally high expression of IGHM / IGLC2 / IL32 genes ( Figure 6 ), providing new targets for targeted therapy (such as anti-IL32 antibodies and MMP8 inhibitors).

[0057] (2) Multidimensional signal integration: B cell activation (IGHM: immunoglobulin heavy chain), inflammation regulation (IL32, NEAT1), and cell function homeostasis (ACTG1), overcoming the limitations of single pathway markers.

[0058] 3. Technical scalability and robustness

[0059] (1) The model established by the training set is externally validated by the validation set, and the validation set AUC ( Figure 12 ) This shows that the model still maintains excellent discrimination ability in external validation and no obvious performance degradation is observed.

[0060] (2) Internal validation reliability of the entire sample: After 1000 bootstrap resamplings, the model AUC stability was excellent (Table 4, Figure 13 ).

[0061] (3) The coefficient differences between the two models (training set model and full sample model) are very subtle (see Table 5), which further proves the excellent stability of the four-gene joint model. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 These are the data quality inspection results after single-cell sequencing of 8 samples in Example 1 of the present invention.

[0063] Figure 2 Figure 1 shows the scRNA profiles and GO enrichment analysis results for the control, GMPP, and RMPP groups in Example 1. (A) Clustering and annotation of all cells in eight samples, with UMAP displaying single-cell profiles; (B-D) Single-cell profiles for each of the three sample groups; (E) GO enrichment analysis of highly expressed genes in four cell types.

[0064] Figure 3 It is the marker gene of the cell type in Example 1 of the present invention.

[0065] Figure 4 is the ratio of cell types in Example 1 of the present invention.

[0066] Figure 5 It is a gene specifically expressed in the GMPP group in Example 1 of the present invention.

[0067] Figure 6 These are genes specifically expressed in the RMPP group in Example 1 of the present invention.

[0068] Figure 7The Myeloid cells of the RMPP group in Example 1 of the present invention are subdivided into 7 subclasses, as well as the marker genes and subclass cell ratios.

[0069] Figure 8 These are the amplification curves and melting curves of the MMP8 and SUB1 genes detected by RT-qPCR in Example 1 of the present invention.

[0070] Figure 9 These are the results of detecting the expression levels of six candidate marker genes using RT-qPCR on prospective cohort samples in Example 1 of the present invention.

[0071] Figure 10 These are the ROC curves for single-gene diagnosis and the ROC curves for the four-gene combined diagnosis model in Example 1 of the present invention.

[0072] Figure 11 The cut-off value of the ROC curve of the four-gene combined diagnosis model in Example 1 of the present invention and the corresponding sensitivity and specificity.

[0073] Figure 12 This is the result of external validation of the four-gene combined diagnostic model by prospectively collecting validation set samples in Example 1 of the present invention.

[0074] Figure 13 These are the bootstrap 1000 results of the four-gene combined diagnosis model for all samples (training set + validation set) in Example 1 of the present invention. DETAILED DESCRIPTION

[0075] The following examples are intended to illustrate the present invention but are not intended to limit the scope of the invention. Without departing from the spirit and essence of the present invention, modifications or substitutions made to the inventive method, steps or conditions are intended to fall within the scope of the present invention. The reagents, products and instruments used in the following examples are all commercially available, and the methods used in the examples are consistent with conventional methods unless otherwise specified.

[0076] The technical solution of the present invention is further elaborated in detail below in conjunction with embodiments.

[0077] Example 1

[0078] (I) Enrollment of children in the RMPP, GMPP, and Normal groups, and collection and processing of peripheral blood samples

[0079] (1) Clinical data of children with suspected MPP (including RMPP, GMPP, and non-MPP) admitted to the only tertiary-level A children's hospital in South China (Shenzhen Children's Hospital) were prospectively collected, including age, sex, white blood cell count, CRP, mycoplasma IgM antibody test or MP DNA load in throat swabs within 24 hours of admission, fever duration, cough duration, hospitalization time, severity of lung imaging, i.e., extent of lung infiltration (multiple lobes involved or ≥2 / 3 of the lungs) and extrapulmonary complications (refer to [Chinese Medical Association Pediatrics Branch Respiratory Group. Guidelines for the Management of Community-Acquired Pneumonia in Children (2013 Revision) (Part 2). Chinese Journal of Pediatrics, 2013, 51(11): 856-862]. This study passed medical ethics approval (ethics approval number 202404202), and informed consent was obtained from the children and their guardians.

[0080] Inclusion criteria for the MPP group: ① Age < 14 years; ② MPP ​​diagnosis: in accordance with the diagnostic criteria of the expert consensus on the diagnosis and treatment of Mycoplasma pneumonia in children (2015 edition) (reference [Respiratory Group of the Pediatric Branch of the Chinese Medical Association, Editorial Committee of the Chinese Journal of Practical Pediatrics. Expert consensus on the diagnosis and treatment of Mycoplasma pneumonia in children. Chinese Journal of Practical Pediatrics, 2015, 30(17): 1304-1308]), that is, clinical manifestations of pneumonia (fever, cough, etc.) and / or imaging changes (sheet-like infiltration shadows, lobar parenchymal infiltration shadows, atelectasis, pleural effusion, etc.), and combined with MP pathogen examination (serum MP-IgM antibody detection and MP-DNA load detection) for diagnosis; ③ Family members agreed to the examination and signed the consent form, informed the family members of the research topic, and obtained consent.

[0081] Exclusion criteria for the MPP group: ① incomplete clinical data; ② the presence of underlying diseases such as chronic lung disease (bronchopulmonary dysplasia, congenital lung malformations), bronchial foreign bodies, congenital heart disease, and autoimmune deficiency; ③ concurrent infection with other pathogens; ④ family refusal.

[0082] RMPP group: patients who have not improved their clinical signs, have persistent fever, or have worsening lung imaging changes after standardized macrolide treatment for 7 days or more.

[0083] GMPP group: Clinical symptoms were significantly alleviated within 7 days of standardized macrolide treatment, and there was no progressive worsening of lung imaging.

[0084] (2) Clinical data of children who underwent circumcision in the Department of Urology and children with tic disorders in the Department of Pediatrics (Normal group) were collected, including age, gender, and medical history; peripheral blood was collected.

[0085] Inclusion criteria for the Normal group: ① No MPP. ② Age <14 years. ③ No other respiratory diseases. Family members agreed to the examination and signed a consent form. ④ Family members were informed of the study topic and consent was obtained.

[0086] Exclusion criteria for the Normal group: ①Meet the inclusion criteria but have immune system diseases and other diseases, such as congenital heart disease, rhinitis and sinusitis, gastroesophageal reflux disease, etc. ②The child or guardian refuses to sign the informed consent form.

[0087] (3) Collect peripheral blood from the enrolled cases and the Normal group, isolate PBMCs, and preserve them.

[0088] Peripheral blood samples were collected from children suspected of MPP within 24 hours of admission to the respiratory department of Shenzhen Children's Hospital, the only tertiary-level A-grade pediatric hospital in South China. PBMCs were extracted and stored at -80°C (maximum reflection of the early onset of the disease). MP-DNA assays and subsequent clinical treatment progress were used to confirm MP and differentiate between GMPP and RMPP. Preserved samples were retrospectively regrouped and selected for scRNA-seq sequencing to ensure that the diagnostic markers discovered by sequencing reflected the early onset of GMPP and RMPP. Approximately 2 ml of peripheral blood was collected within 24 hours of admission, and PBMCs were prepared by density gradient centrifugation. Briefly, whole blood cells were diluted with 1× DPBS (1:1), layered on a Ficoll-Paque Plus instrument, and centrifuged at 500 g for 20 minutes at room temperature. The white layer containing PBMCs was isolated and washed twice with DPBS, followed by removal of erythrocytes using ice-cold erythrocyte lysate. After centrifugation, the cell pellet was resuspended in RPMI / 10% FBS medium with 10% DMSO. 1ml aliquots of the cell suspension were stored at -80°C overnight and transferred to liquid nitrogen the next day for storage. After the MPP child received clinical treatment and the deputy chief physician or chief physician of the respiratory department confirmed GMPP or RMPP, peripheral blood specimens collected within 24 hours of admission were grouped and labeled. The specimens were divided into GMPP, RMPP, and an uninfected control group. Three samples were randomly selected from each group and sent for sequencing according to the sequencing company's specimen storage and transportation requirements.

[0089] (II) scRNA-seq and data analysis

[0090] Three samples / group were randomly selected and sent to the company for scRNA-seq. After sample processing, loading, library construction, and sequencing, the sequencing data of two uninfected control group samples (numbered: SBS-P1, CKJ-P1) and one sample that failed the data quality inspection (not included in subsequent data analysis), three GMPP group samples (numbered: G1-pbmc, AX-P1, WJY-P1), and three RMPP group samples (numbered: WSJ-P1, WYT-P1, WY-P1) were determined to meet the quality standards ( Figure 1 Single-cell bioinformatics analysis was performed on data that met the quality requirements. First, cluster analysis was performed ( Figure 2 AD), annotate cells according to their classic markers ( Figure 3 ), and obtained four cell types (Myeloid cells, T cells, B cells, NK cells) ( Figure 2 AD). GO analysis was performed on the highly expressed genes in each type of cells ( Figure 2 E), the enriched terms are consistent with the cell types. Then, the proportions of the four cell types in each of the three groups of samples were calculated and analyzed. The proportions of each cell type varied greatly between the groups ( Figure 4 Then, the differentially expressed genes among the three groups of samples were analyzed. Figure 5 and Figure 6 Represent the genes specifically expressed in the GMPP group and the RMPP group, respectively. Focusing on the RMPP group, compared with the GMPP group and the Normal group, the genes specifically expressed by cells in this group include IGLC2, IGHM, HSP90B1, IL32, ACTG1, NEAT1, SUB1, PPIB, and TUBA1B. Finally, for Myeloid cells, T cells, B cells, and NK cells, we further subdivided each type of cell into cell subclasses, and then calculated the changes in the proportion of cells in each subclass between the three groups of cells. Among them, we only found that Myeloid cells were further subdivided into 7 subclasses: CD14Mono1, CD14 Mono2, CD14 Mono3, Mac, Neu, DC, MK ( Figure 7 A), the proportion of Neu subtype cells increased significantly in the RMPP group ( Figure 7 B). When the other three cell types were further divided into subtypes, no significant changes in cell ratios were found. We further analyzed the marker genes of the seven subtypes and showed that Neu subtype cells specifically expressed the MMP8 gene ( Figure 7 C). By Figure 6 and Figure 7The RMPP group showed specific expression of the genes IGLC2, IGHM, HSP90B1, IL32, ACTG1, NEAT1, SUB1, PPIB, TUBA1B, and MMP8. We selected eight of these genes, including IGLC2, IGHM, HSP90B1, IL32, ACTG1, NEAT1, SUB1, and MMP8, as candidate marker genes for distinguishing between the GMPP and RMPP groups for subsequent RT-qPCR validation.

[0091] The specific experimental techniques and methods for each step of scRNA-seq sample processing and sequencing analysis are as follows:

[0092] (1) PBMC isolation steps

[0093] After receiving the sample, the sequencing company will thaw the cells, discard the supernatant, and resuspend the PBMC in phosphate-buffered saline (PBS, prepared by HyClone) to obtain a single-cell suspension. The suspension is then stained with trypan blue and assessed for cell viability under a microscope. Only after cell viability and cell number meet the required standards can subsequent library construction and sequencing be performed.

[0094] (2) Reverse transcription & amplification & library construction

[0095] Single cell suspension (2×10 5 cells / mL) were loaded onto a microwell chip using the Singleron Matrix® Single-Cell Processing System. Barcoding beads were then collected from the microwell chip, and the captured mRNA was reverse transcribed to generate cDNA, which was then amplified by PCR. The amplified cDNA was fragmented and ligated with sequencing adapters. scRNA-seq libraries were constructed according to the GEXSCOPE® Single-Cell RNA Library Kit (Singleron) instructions. Individual libraries were diluted to 4 nM, pooled, and sequenced using paired-end 150 bp sequencing on an Illumina Novaseq 6000 sequencer.

[0096] (3) Primary analysis of raw sequencing data

[0097] Using CeleScope ( https: / / github.com / singleron-RD / CeleScope) v1.9.0 analysis pipeline was used to process the raw reads of single-cell RNA sequencing and generate a gene expression matrix. The brief process is as follows: (1) Data preprocessing: Cutadapt v1.17 was used to remove low-quality reads and trim the poly-A tail and adapter sequences; (2) Information extraction: Cell barcodes and unique molecular identifiers (UMIs) were parsed; (3) Sequence alignment: STAR v2.6.1a was used to align the reads to the reference genome GRCh38 (Ensembl version 92 annotation); (4) Quantitative analysis: FeatureCounts v2.0.1 was used to count the UMI counts and gene expression levels of each cell, and finally an expression matrix file was generated for subsequent analysis.

[0098] (4) Contamination Removal

[0099] Visualizing cell marker gene expression levels using the FeaturePlot function in the Seurat package v3.1.2 revealed significant environmental RNA contamination. To address this, environmental RNA decontamination was subsequently performed using the decontX algorithm in the celda package v1.3.8. This algorithm assumes the existence of K cell populations and uses Bayesian variational inference to resolve environmental RNA contamination as a weighted combination of cell population distributions. The algorithm takes as input the original UMI count matrix and several cell populations calculated using the FindClusters function in the Seurat package v3.1.2. It then derives the decontamination count matrix from the raw data. A default random seed was used throughout the analysis to ensure reproducible results.

[0100] (5) Quality control, dimensionality reduction, and clustering (Seurat)

[0101] Cell quality control first eliminated cells with fewer than 200 genes or ranked in the top 2% for gene or UMI counts. Cells with a mitochondrial gene count exceeding 20% ​​were also excluded. Retained cells were used for downstream analysis to determine the average number of genes and UMIs detected per cell. Seurat v3.1.2 software was used for dimensionality reduction and cluster analysis: gene expression was normalized and standardized using the NormalizeData and ScaleData functions. Principal component analysis (PCA) was performed using FindVariableFeatures to select the top 2000 highly variable genes. Cell clustering was achieved using the FindClusters function based on the top 20 principal components. The Harmony algorithm was used to eliminate batch effects between samples. Finally, the UMAP algorithm was used to visualize cell distribution in two dimensions.

[0102] (6) Statistics and reproducibility

[0103] The Shapiro-Wilk method was used to test the normality of gene expression (R4.4.1). (2) The Kruskal-Wallis rank sum test was used to test the differences between groups. All statistical analyses and results were performed using the R language. The statistical test methods used in the figures are described in the figure legends, and statistical significance was set at p < 0.05.

[0104] (7) Analysis of differentially expressed genes (DEGs)

[0105] To identify differentially expressed genes (DEGs), we used the FindMarkers function in the Seurat software package, based on the Wilcox likelihood ratio test (default parameters). The screening criteria were: genes expressed in more than 10% of cells in a given cluster and with an average log(fold change) value greater than 0.25 were considered DEGs. For cell type annotation of each cluster, we combined the expression of classic marker genes found in DEGs with literature knowledge. Seurat's DoHeatmap / DotPlot / VlnPlot functions were used to generate heatmaps / dot plots / violin plots to display marker expression for each cell type. Doublet cells were defined as cells expressing markers of different cell types and were manually excluded.

[0106] (8) Cell type annotation

[0107] The cell type identity of each cell cluster was determined by combining the expression of classic marker genes found in the DEGs with reference to the SynEcoSys database. The expression of marker genes used to identify each cell type was generated as heatmaps / dot plots / violin plots using the DoHeatmap / DotPlot / VlnPlot functions in Seurat v3.1.2.

[0108] (III) Prospective collection of clinical specimens, detection of candidate marker gene expression levels in the RMPP group, and verification of differences in marker gene expression levels between groups

[0109] We selected 8 genes, including IGLC2, IGHM, HSP90B1, IL32, ACTG1, NEAT1, SUB1, and MMP8, as candidate diagnostic markers for distinguishing the GMPP group ( Figure 6 and Figure 7Five samples were randomly selected from the Normal, GMPP, and RMPP groups. RT-qPCR (primer sequences are shown in Table 1) was used to perform preliminary experiments on the mRNA levels of the above eight genes. The primer amplification efficiency and melting curve were evaluated to determine whether the primer sequence met the RT-qPCR requirements. We found that the primers for SUB1 and MMP8 produced double peaks after amplification, and the melting curves were less consistent ( Figure 8 ), the amplification curves and melting curves of the other 6 genes were good, and further verification and statistical analysis will be carried out by expanding clinical samples.

[0110] Table 1 RT-qPCR primer sequence information

[0111]

[0112] Sample size calculation:

[0113] The joint gene model is guaranteed to have an AUC ≥ 0.85, a 95% confidence interval width ≤ 0.10, a two-sided α = 0.05, and a test power 1-β = 0.90 in the three-classification scenario (Normal vs GMPP vs RMPP).

[0114] Design parameters: Class allocation: Normal:GMPP:RMPP = 1:1:1. Based on pre-experimental results (AUC and confidence interval widths), the expected minimum AUCs were: Normal 0.85, GMPP 0.85, and RMPP 0.85. The tolerance for error (δ) (half-width): 0.06. Using the multiclass ROC formula, the required sample size for each class was calculated using the PASS15 "ROC–Confidence Intervals for Area Under ROC Curve" module. The total sample size was 97 cases for each class, 291 cases for the three groups, and 320 cases for a 10% dropout rate.

[0115] We prospectively collected peripheral blood samples from the training set (Normal group 99 cases, GMPP group 104 cases, RMPP group 92 cases) and used RT-qPCR to detect the mRNA levels of six candidate genes (IGLC2, IGHM, HSP90B1, IL32, ACTG1 and NEAT1). -ΔΔCt All experiments were performed under blinded conditions, and data reliability was ensured by the use of internal reference genes and repeated testing.

[0116] Data analysis was performed using R4.4.1: (1) Normality test: The Shapiro-Wilk method was used to test the normality of the expression of each gene (R4.4.1). The results showed that the six genes did not obey the normal distribution in the three groups (P < 0.05). (2) Inter-group difference test: The Kruskal-Wallis rank sum test was used to confirm that the differences in the four genes (HSP90B1, IL32, ACTG1 and NEAT1) between the Normal, GMPP and RMPP groups were statistically significant (P < 0.05) ( Figure 9 ).

[0117] RT-qPCR experimental steps:

[0118] 1. Total RNA Extraction

[0119] Peripheral blood specimens were prospectively collected from 117 patients in the Normal group (99 in the initial prospective training set and 18 in the subsequent prospective validation set), 123 patients in the GMPP group (104 in the initial prospective training set and 19 in the subsequent prospective validation set), and 109 patients in the RMPP group (92 in the initial prospective training set and 17 in the subsequent prospective validation set). One ml of fresh peripheral blood was collected within 24 hours of admission and centrifuged at 3500 rpm for 5 minutes at 24°C. The supernatant was discarded to obtain the cell pellet.

[0120] ② Lyse red blood cells: Add 4 ml of red blood cell lysis buffer to the cell pellet, mix well, let it stand for 15 minutes, centrifuge at 3500 rpm at 4°C for 5 minutes, discard the red supernatant, and obtain the white blood cell pellet.

[0121] ③ Wash: Add 200 μl of 1× PBS, pipette repeatedly to mix, centrifuge at 3500 rpm for 5 min at 4°C, discard the supernatant, and repeat once.

[0122] ④RNAiso Plus lysis: Add 1 ml of RNAiso Plus to the cells, pipette up and down repeatedly, and let it stand for 5 minutes.

[0123] ⑤Chloroform extraction and separation: Add 200 μl RNA Extraction Assisted Reagent (REX), mix thoroughly by inverting for 15 seconds, let it stand for 3 minutes, and centrifuge at 4°C, 12800 rpm, for 15 minutes.

[0124] ⑥ Isopropanol precipitation: Transfer the supernatant (about 500 μl) to a new 1.5 ml EP tube, add an equal volume (500 μl) of pre-cooled isopropanol, gently mix by inverting, let stand for 10 minutes, and centrifuge at 4°C, 11,800 rpm, for 10 minutes.

[0125] ⑦ Rinse with anhydrous ethanol: discard the supernatant, add 70% anhydrous ethanol (freshly prepared with depcol water) with an equal volume to Trizol (1 ml), turn the tube upside down to float the precipitate, and centrifuge at 4°C, 11800 rpm, for 5 min.

[0126] ⑧Dissolve: Discard the supernatant, dry at room temperature, and wait until the white precipitate at the bottom of the tube becomes transparent. Then add approximately 10 μl of RNase-free double-distilled water to dissolve the RNA. Store at -80°C.

[0127] 2. First-Strand cDNA Synthesis (TaKaRa-PrimeScript™ RT Master Mix (Perfect RealTime)-RR036A)

[0128] (1) RNA quantification and purity detection (NanoDrop One / Onec detection to obtain the absorption wavelength of the highest absorption peak of each component): Open the software, select "Nucleic Acid", and after optical path calibration, select "RNA"; aspirate 1 μl of nuclease-free water, load the sample, and click "blank"; aspirate 1 μl of total RNA, load the sample, and click "measure"; obtain the total RNA concentration, OD values ​​of A230nm, A260nm, and A280nm.

[0129] (2) PCR (perform on ice throughout): ① Prepare the RT reaction solution (perform on ice) according to the following composition: 5× PrimeScript RT Master Mix (Perfect Real Time), 2 μl; Total RNA, 500 ng; RNase-free ddH2O, make up to 10 μl. The reaction system can be scaled up as needed (refer to the instructions). ② After mixing, perform reverse transcription at 37°C for 15 min; 85°C for 5 sec; and 4°C.

[0130] 3. RT-qPCR (TaKaRa-TB Green® Premix Ex Taq™ II (Tli RNaseH Plus) -RR820A)

[0131] (1) Prepare the PCR reaction solution according to the following components (the reaction solution should be prepared on ice): TB Green Premix ExTaq II (Tli RNaseH Plus), 10.0 μl; PCR Forward Primer (10 μM), 0.8 μl; PCR Reverse Primer (10 μM), 0.8 μl; DNA template (<100 ng), 2.0 μl; sterile water, 6.4 μl; a total of 20.0 μl.

[0132] (2) Perform Real-Time PCR reaction<LightCycler480 System> Denaturation: 95°C, 30 sec (Ramp Rate 4.4°C / sec), 1 cycle; PCR (Analysis Mode: Quantitative Analysis, 40 cycles): 95°C, 5 sec (Ramp Rate 4.4°C / sec); 60°C, 30 sec (Ramp Rate 2.2°C / sec, Acquisition Mode: Single). Melting (Analysis Mode: Melting Curve, 1 cycle): 95°C, 5 sec (Ramp Rate 4.4°C / sec); 60°C, 1 min (Ramp Rate 2.2°C / sec); 95°C (Ramp Rate 0.11°C / sec, Acquisition Mode: Continuous, Acquisitions: 5 per°C). Cooling (1 cycle): 50°C, 30 sec (Ramp Rate 2.2°C / sec).

[0133] 4. Data processing and analysis: After the RT-qPCR reaction is completed, the Ct value of the results is calculated by the LightCycler 480 amplification instrument. -ΔΔCt The relative quantitative analysis method was used, ΔCt 实验组 =Ct 实验组目标基因 -Ct 实验组内参基因 ;ΔCt 对照组 = Ct 对照组目标基因 -Ct 对照组内参基因 ; ΔΔCt=ΔCt 实验组 -ΔCt 对照组 . Use 2 -ΔΔCt The formula is used for the final data statistics to express the relative multiple relationship between the experimental group and the control group.

[0134] (IV) Construction of a three-classification ROC diagnostic model based on the expression levels of candidate marker genes in the RMPP group

[0135] 1. Based on the expression levels of six candidate diagnostic marker genes from prospectively collected training set specimens (Normal group 99 cases, GMPP group 104 cases, RMPP group 92 cases), to explore the discriminative ability of each gene for the three-category outcome (Normal / GMPP / RMPP), this study implemented a one-vs-rest strategy for each category, constructed a univariate multinomial logistic regression model, and plotted a smoothed ROC curve. The AUC results are shown in Figure 2. Figure 10 .

[0136] (1) Normal vs Rest: ACTG1 (0.892) > NEAT1 (0.733) > IL32 (0.711) > HSP90B1 (0.572) > IGLC2 (0.551) > IGHM (0.53);

[0137] (2) GMPP vs Rest: ACTG1 (0.701) > IL32 (0.667) > HSP90B1 (0.646) > NEAT1 (0.618) > IGLC2 (0.579) > IGHM (0.507);

[0138] (3) RMPP vs Rest: IL32 (0.877) N > EAT1 (0.844) > HSP90B1 (0.726) > ACTG1 (0.672) > IGLC2 (0.537) > IGHM (0.517).

[0139] 2. Construction of a four-gene combined diagnostic model

[0140] (1) Variable screening: Based on multinomial logistic regression, the backward stepwise method based on the Wald test (α = 0.05) was used. After iteration, HSP90B1 and IGLC2 were eliminated, and finally 4 genes, IGHM, NEAT1, IL32, and ACTG1, were included to form a four-gene joint diagnosis model. Figure 10 .

[0141] (2) Model formula

[0142] Let X1-X4 represent the 2 of IGHM, NEAT1, IL32, and ACTG1 respectively. –ΔΔCt value, then:

[0143] P_GMPP=exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4)

[0144] / [1+exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4)

[0145] +exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)];

[0146] P_RMPP=exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)

[0147] / [1+exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4)

[0148] +exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)]. The coefficients are detailed in Table 2.

[0149] Table 2 Coefficients of the three-category multivariate logistic model for the four-gene combined diagnosis

[0150]

[0151] 3. ROC evaluation of the four-gene joint model

[0152] Under the One-vs-Rest strategy, the AUCs of the four-gene joint model in the three categories were: Normal vs Rest 0.978; GMPP vs Rest 0.974; RMPP vs Rest 0.952 ( Figure 10 ).

[0153] To comprehensively compare the overall discrimination of the single-gene and combined models, the macro-average method was used to calculate the average AUC, which was 0.968 for the combined model. The ranking of the single genes was: IL32 0.752, ACTG1 0.755, NEAT1 0.731, HSP90B1 0.648, IGLC2 0.556, and IGHM 0.5158. Therefore, the four-gene combined model significantly outperformed any of the single-gene models.

[0154] 4. Optimal cutoff value and diagnostic performance of the four-gene joint model

[0155] Based on the principle of maximizing the Youden index, the optimal cut-off value (cut-off) was determined in the original ROC curve of the combined diagnosis model (see Figure 11): (1) Normal vs Rest: cut-off=0.246, Se (sensitivity)=0.98, Sp (specificity)=0.91; (2) GMPP vs Rest: cut-off=0.369, Se=0.94, Sp=0.93; (3) RMPP vs Rest: cut-off=0.246, Se=0.91, Sp=0.86.

[0156] A three-category decision matrix was constructed based on the above thresholds, and the overall diagnostic accuracy was 88.5%.

[0157] Macro-average AUC: Calculate the AUC of each of the three One-vs-Rest ROC curves first, and then average them. This measures the overall discrimination ability of the model and does not directly correspond to any specific threshold.

[0158] Overall diagnostic accuracy: 295 samples were classified into three categories using the three cut-off thresholds obtained using the Youden index. The correct proportion was calculated by comparing the samples with the true labels.

[0159] In summary, we provide a set of peripheral blood four-gene (IGHM, NEAT1, IL32, ACTG1) combined detection kit and its supporting algorithm that can be directly implemented on the conventional RT-qPCR platform, so that grassroots hospitals can use 2 –ΔΔCt The system inputs the value and outputs the Normal / GMPP / RMPP three-class probability and the corresponding diagnosis conclusion in real time.

[0160] 5. External Validation of the Four-Gene Joint Model

[0161] Strategy: After constructing the four-gene joint model, the research team prospectively re-enrolled cases for external validation. 18 cases in the Normal group, 19 cases in the GMPP group, and 17 cases in the RMPP group were collected to form the validation set. The ROC curves of the training set and validation set are shown in detail. Figure 12 The validation set AUCs were 0.985 for Normal vs Rest, 0.997 for GMPP vs Rest, and 0.979 for RMPP vs Rest, respectively. This indicates that the model maintained excellent discriminatory capabilities during external validation, with no significant performance degradation. Overall, these results demonstrate that the four-gene joint model performed robustly in prospective validation and can provide a reliable basis for subsequent clinical implementation.

[0162] 6. Internal Validation of the Four-Gene Joint Model

[0163] Considering that the sample size of each group was relatively small when constructing the model in this study, the stability of the model may be affected by the sample size. Therefore, all samples (training set + validation set) were used in the same variable screening and model construction method to construct a full sample model. The backward stepwise method based on the Wald test was used (α=0.05). After iteration, HSP90B1 and IGLC2 were eliminated, and finally 4 genes, IGHM, NEAT1, IL32, and ACTG1, were included. The model coefficients are shown in Table 3. 1000 stratified bootstrap resamplings were performed for internal validation, and the following robust performance indicators (95% CI) were obtained. See Table 4 for details, as well as the "box-line-mean curve" graph constructed from the discretized ROC data set generated by the bootstrap ( Figure 13 ).

[0164] The coefficient differences between the two models are very subtle, further demonstrating the stability of the four-gene joint model. See Table 5 for details.

[0165] All three categories achieved AUCs >0.95, demonstrating the model's excellent discriminatory ability during internal validation. The 95% confidence interval for the cutoff was <0.20, indicating a stable threshold and that the mean value can be used directly as the clinical decision point. Both mean sensitivity and specificity values ​​were >0.90, further validating the model's high performance and robustness.

[0166] Table 3 Coefficients of the three-category multivariate logistic model for the four-gene combined diagnosis constructed for the entire sample

[0167]

[0168] Table 4 Results of 1000 bootstrap internal validations of the four-gene joint model

[0169]

[0170] Table 5 Comparison of model coefficients between the training set model and the full sample model

[0171]

[0172] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A gene marker combination, characterized in that: The gene marker combination includes IGHM, NEAT1, IL32 and ACTG1.

2. Use of the gene marker combination according to claim 1 in the preparation of diagnostic products, early screening products, disease monitoring products and / or drug efficacy monitoring products for refractory Mycoplasma pneumonia (RMPP).

3. A method for screening the gene marker combination according to claim 1, characterized in that: The screening method includes the use of single-cell transcriptome sequencing (scRNA-seq) technology and RT-qPCR technology.

4. The screening method according to claim 3, wherein The sample source in the screening method includes peripheral blood.

5. A product for diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of RMPP, characterized in that: The product comprises a reagent for detecting the gene marker combination according to claim 1.

6. The product according to claim 5, characterized in that The product further comprises one or more pharmaceutically acceptable carriers or excipients.

7. The product according to claim 6, characterized in that Such product forms include kits.

8. A model for diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of RMPP, characterized in that: The method for constructing the model comprises using the gene marker combination according to claim 1.

9. The model according to claim 8, characterized in that The model is based on scRNA-seq technology and RT-qPCR technology.

10. The model according to claim 9, characterized in that The model formula assumes that X1-X4 represent the 2 of IGHM, NEAT1, IL32, and ACTG1, respectively. –ΔΔCt value, then: P_GMPP=exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4) / [1+exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4) +exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)]; P_RMPP=exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4) / [1+exp(4.800+2.424·X1+0.994·X2+3.286·X3–17.750·X4) +exp(5.706+3.103·X1–2.419·X2–5.327·X3–4.759·X4)].

Images