Probes, marker combinations and predictive methods thereof for neuroblastoma non-invasive risk stratification

Abstract

The present application belongs to the technical field of molecular medicine, and relates to a probe, a marker combination and a prediction method for non-invasive risk grading of neuroblastoma. The probe UUU-DZ-tFNA can detect UDG in vitro, cells, living bodies and plasma exosomes, and it is proved that UDG in plasma exosomes is closely related to the risk grading of NB. The plasma exosomes of NB patients are collected MYCN Whether to amplify, whether the tumor is metastatic, neuron-specific enolase content, lactate dehydrogenase content and plasma exosome UDG content are integrated, and a NB non-invasive risk grading prediction model is established through a machine learning algorithm. The optimal risk grading model is a neural network model. The area under the receiver operating characteristic curve of the combined model is improved by 5.2% compared with the optimal prediction result of the combined prediction model of the single clinical index, and the sensitivity and specificity are significantly improved. The present application has key clinical values for accurate grading of NB children, individualized treatment plan formulation and efficacy monitoring.

Description

Technical Field

[0001] This invention belongs to the field of molecular medicine technology and relates to non-invasive risk grading of neuroblastoma, specifically a predictive method for non-invasive risk grading of neuroblastoma. Background Technology

[0002] Neuroblastoma (NB) is the most common extracranial malignant solid tumor in childhood, accounting for 6%-10% of childhood malignancies but 15% of deaths. The challenge in diagnosing and treating this disease lies in its extreme heterogeneity. Some low- and intermediate-risk cases may spontaneously regress without treatment, but high-risk cases often have metastases at diagnosis. Even with multimodal treatment including chemotherapy, surgery, radiotherapy, and hematopoietic stem cell transplantation, the recurrence rate is approximately 50%-60%, and the 5-year survival rate for typical high-risk cases is only 48%, a stark contrast to the over 96% 5-year survival rate for low- and intermediate-risk cases.

[0003] In clinical practice, accurate risk stratification is a prerequisite for developing individualized treatment plans for neuroblastoma (NB). Currently, the mainstream international stratification system is based on the Childhood Oncology Group (COG) criteria, combined with the International Neuroblastoma Staging System (INSS / INRG staging) and tumor tissue analysis. MYCN Multiple indicators, including gene amplification status, DNA ploidy, and the International Classification of Neuroblastoma Pathology (INPC), are used to classify children into three levels: low-risk, intermediate-risk, and high-risk. Treatment strategies differ significantly depending on the risk level: low-risk children often require only surgery or observation; intermediate-risk children require surgery combined with moderate-dose chemotherapy; and high-risk children undergo a complex treatment process including induction chemotherapy, consolidation-phase hematopoietic stem cell transplantation, and maintenance immunotherapy (such as anti-GD2 monoclonal antibodies). The key goal of this stratified treatment model is to avoid overtreatment of low-risk children and undertreatment of intermediate- and high-risk children, but this requires early and accurate risk assessment.

[0004] However, current clinical grading methods face serious technical bottlenecks—core grading indicators (such as...) MYCN Obtaining copy number and pathological classification is highly dependent on tissue biopsies of the primary or metastatic tumor lesions. For most high-risk children who present at an advanced stage, the tumors are often enormous and frequently encircle important structures such as major abdominal blood vessels, making direct biopsy surgery extremely risky. This can lead to serious complications such as anesthetic accidents, massive intraoperative and postoperative bleeding, tumor rupture, and abdominal infection, even resulting in irreversible medical accidents. Even if biopsy samples are successfully obtained from some children, traditional testing procedures are costly and complex, making them unsuitable for the needs of primary healthcare institutions or emergency treatment scenarios. Furthermore, clinically… MYCN Copy number is another important indicator for NB risk stratification, but it is necessary to test the child's tumor tissue to obtain the MYCN copy number.

[0005] To overcome the limitations of tissue biopsy, non-invasive risk grading technology has become a research hotspot in the diagnosis and treatment of neonatal nephropathy (NB). Among these, liquid biopsy, with its advantages of convenient sample acquisition and dynamic monitoring, has become one of the most promising directions for application. For example, patent 202311019710.4 discloses a method for NB risk grading based on plasma miRNA, and screens out three biomarkers—miR-488-3p, miR-514a-3p, and miR-887-3p—that are highly correlated with NB risk, establishing a plasma risk grading model. This technology detects plasma miRNA expression levels through RT-PCR, without the need to obtain tumor tissue, effectively avoiding the risks of surgical biopsy. However, plasma miRNA detection is highly dependent on specialized laboratory equipment and technicians, resulting in high testing costs and limited accessibility. Therefore, there is an urgent need to establish a timely and accurate non-invasive risk grading prediction strategy that does not rely on tumor tissue for NB diagnosis, which has key clinical value for early screening, accurate grading, individualized treatment planning, and efficacy monitoring in NB patients. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a combination of probes and markers for non-invasive risk grading of neuroblastoma and a prediction method thereof.

[0007] The technical solution of this invention is implemented as follows:

[0008] This invention investigated the correlation between uracil DNA glycosidase (UDG), an enzyme associated with a representative base excision repair pathway, and NB risk stratification, demonstrating a close correlation between UDG in plasma exosomes and NB risk stratification. To further improve the sensitivity and specificity of NB risk prediction, this application combined NB plasma exosome UDG expression with multidimensional clinical indicators and constructed a non-invasive NB risk stratification prediction model using machine learning algorithms. This model exhibits high specificity and sensitivity, and has good clinical application value.

[0009] On the one hand, the present invention provides a probe for non-invasive risk grading of neuroblastoma, the preparation method of which is as follows:

[0010] (1) Four single-stranded DNAs S1, S2, S3 and S4 were uniformly mixed in TEM buffer, denatured and cooled, and assembled into DNA tetrahedral tFNA;

[0011] (2) DNA single strand S5 and S6 are mixed in TEM buffer, denatured and cooled to form S5-S6 complex;

[0012] (3) Mix single-stranded DNA S7 with TEM buffer, and then denature and cool to prepare HP2;

[0013] (4) After mixing tFNA, S5-S6 complex and HP2, incubate to obtain probe UUU-DZ-tFNA.

[0014] Preferably, in step (1) above, the sequence of DNA single strand S1 is as shown in SEQ ID No.1, the sequence of S2 is as shown in SEQ ID No.2, the sequence of S3 is as shown in SEQ ID No.3, and the sequence of S4 is as shown in SEQ ID No.4; the concentrations of S1, S2, S3 and S4 are all 1-100 μM and the molar ratio is 1:1:1:1.

[0015] Preferably, in step (2) above, the sequence of DNA single strand S5 is as shown in SEQ ID No. 5; the sequence of S6 is as shown in SEQ ID No. 6, and the nucleotides at positions 7, 10, 11, 14 and 18 from the 5' end are deoxyuridine dU; the concentrations of S5 and S6 are both 1-100 μM and the molar ratio is 1:1.

[0016] Preferably, in step (3) above, the DNA single strand S7 has the sequence shown in SEQ ID No.7, with a Cy5 fluorescent group linked between the 24th and 25th nucleotides from the 5' end, the 39th nucleotide being riboadecanoside rA, and the 3' end modified with a BHQ3 fluorescent quencher group; the concentration of S7 is 3-300 μM.

[0017] Preferably, the TEM buffer contains 20 mM Tris-HCl, 2 mM EDTA, and 12.5 mM MgCl2·6H2O; denaturation is performed at 95°C for 10 min, cooling is performed at a rate of -1°C / min to 25°C, and incubation is performed at 37°C for 30 min.

[0018] Secondly, a reagent or kit containing the aforementioned probe UUU-DZ-tFNA.

[0019] The probe UUU-DZ-tFNA can detect UDG at in vitro, cellular, in vivo and plasma exosome levels. Its response signal is significantly increased in high-risk NBPDX tumor tissues, proving that UDG can be used as an indicator for NB risk grading.

[0020] Thirdly, the present invention also provides a combination of markers for non-invasive risk grading of neuroblastoma, the combination of markers comprising plasma. MYCN Whether amplification occurs, whether the tumor has metastasized, the content of neuron-specific enolase, lactate dehydrogenase, and plasma exosomal uracil DNA glycosylation enzyme; the plasma exosomal uracil DNA glycosylation enzyme content is obtained by detection using the above-mentioned reagents or kits.

[0021] Cell-free DNA was extracted from plasma and detected in the plasma. MYCN Copy number confirmed the presence of [the virus] in the plasma and tissues of NB patients. MYCN The amplification status is consistent. Therefore, in order to achieve non-invasive risk stratification of NB, this invention... MYCN Free plasma used in the amplification state MYCN Gene amplification status.

[0022] The reference values ​​for the marker combination are as follows: neuron-specific enolase content is 0-3000 ng / mL, lactate dehydrogenase content is 0-3000 U / L, plasma exosomal uracil DNA glycosylase content is <0.5 U / mL, whether plasma MYCN gene is amplified (amplified 1 / non-amplified 0), and whether the tumor has metastasized (metastatic 1 / non-metastatic 0).

[0023] Fourthly, a method for predicting the non-invasive risk grading of neuroblastoma using the above-mentioned combination of markers, comprising the following steps:

[0024] 1) Collect feature data of the combination of markers in the sample to be tested to form a joint non-invasive feature set;

[0025] 2) Input the joint non-invasive feature set into the Neural Network (NN) model to obtain the continuous risk prediction value of the test sample. Based on the model prediction results, construct the ROC curve, calculate the Youden index under different thresholds, and select the prediction value corresponding to the maximum value of the Youden index as the optimal cut-off value to determine the risk level of the test sample.

[0026] A neural network model with optimized hyperparameters through 10-fold cross-validation on the training set and independent validation on the validation set was adopted. The joint feature set was used as the input feature, and the clinical risk level (high risk / low intermediate risk) of NB children was used as the target variable to construct a "clinical + biomarker" joint prediction model CO-CPM. The prediction results were calculated based on the NN model, and the corresponding ROC curves were plotted. The core performance indicators (AUC value, sensitivity, specificity, Youden index, accuracy) of each model were extracted and compared comprehensively.

[0027] Preferably, the Youden index, by simultaneously considering sensitivity and specificity, reflects the overall discriminative power of the model at different thresholds. Within the ROC curve framework, the maximum value of the Youden index can determine the threshold at which the sum of sensitivity and specificity reaches its maximum value, and is therefore used to determine the optimal cut-off value for continuous risk predictions, where the Youden index = sensitivity + specificity - 1.

[0028] Preferably, if the risk score of the sample to be tested output by the neural network model is higher than the cut-off value corresponding to the maximum value of the Youden index, it is identified as a high-risk sample; if the risk score of the sample to be tested output by the neural network model is lower than the cut-off value corresponding to the maximum value of the Youden index, it is identified as a medium- or low-risk sample.

[0029] The present invention has the following beneficial effects:

[0030] 1. This invention constructs and characterizes an in-situ UDG detection probe, UUU-DZ-tFNA, confirming successful synthesis and good sensitivity and specificity. The UUU-DZ-tFNA probe can detect UDG in vitro, in cells, in vivo, and in plasma exosomes. Its reaction signal is significantly increased in high-risk NB PDX tumor tissues, demonstrating that UDG can serve as an indicator for NB risk stratification. The in-situ detection method of UUU-DZ-tFNA is highly consistent with ELISA results, confirming the accuracy and reliability of UUU-DZ-tFNA for detecting UDG in exosomes. Transmission electron microscopy reveals that the exosomes exhibit a typical cup-shaped vesicle morphology with visible surface depressions and clear membrane edges, confirming successful extraction of plasma exosomes. Thermographic results show that 60.0%-86.8% of UDG in plasma is present in exosomes. Exosomal UDG has the greatest value for NB risk stratification, demonstrating a close correlation between UDG in plasma exosomes and NB risk stratification.

[0031] 2. This application integrates plasma exosome UDG expression with multiple clinical indicators (plasma). MYCN This invention integrates amplification status, tumor metastasis, neuron-specific enolase (NSE), and lactate dehydrogenase (LDH) to establish a non-invasive risk stratification prediction model for neonatal neuropathy (NB) using machine learning algorithms. The optimal NB risk stratification model constructed in this invention is the NN model, with an AUC of 0.904 and a corresponding 95% confidence interval of 0.76–1. The ROC curve closely follows the upper left corner of the coordinate system, demonstrating strong discriminative ability between the two patient groups. Based on confusion matrix analysis, the training set model has a sensitivity of 91%, a specificity of 89%, and an overall accuracy of 90%, with positive predictive value (PPV) and negative predictive value (NPV) of 83.3% and 94.4%, respectively. The validation set model has a sensitivity of 100%, a specificity of 86%, and an overall accuracy of 93%, with positive predictive value (PPV) and negative predictive value (NPV) of 87.5% and 100%, respectively, fully meeting the accuracy requirements for clinical stratification. In terms of comparative advantages, the AUC of this combined model is 5.2% higher than the best prediction result of the combined prediction model of individual clinical indicators. Both sensitivity and specificity are significantly improved, fully demonstrating the synergistic effect of multiple indicators combined. It has key clinical value for early screening, accurate classification, individualized treatment plan formulation and efficacy monitoring of NB children. Attached Figure Description

[0032] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is the standard curve for UDG quantification based on the ELISA method.

[0034] Figure 2 The expression of UDG in NB cytoplasm and plasma is shown in Figures AB and C. Figures AB and C show the quantitative determination of UDG levels in NB cytoplasm using an UDG ELISA kit (n = 3), CD shows the quantitative determination of UDG levels in tissue cell cytoplasm using an UDG ELISA kit (n = 3), E shows the quantitative determination of UDG levels in plasma using an UDG ELISA kit (n = 13), and F shows the quantitative determination of UDG levels in plasma using an UDG ELISA kit (n = 10). *P < 0.05, **P < 0.01, ***P < 0.001.

[0035] Figure 3 The synthesis and characterization of UUU-DZ-tFNA are shown in Figure A, which is a schematic diagram of the self-assembly synthesis process of UUU-DZ-tFNA. Figure B shows the successful synthesis of UUU-DZ-tFNA confirmed by AGE (M: DNA marker, L1: S1, L2: S1+S2, L3: S1+S2+S3, L4: tFNA, L5: S5-S6, L6: HP2, L7: UUU-DZ-tFNA). Figure C shows the morphology and size of UUU-DZ-tFNA characterized by AFM, with a scale bar of 50 nm.

[0036] Figure 4 To demonstrate the feasibility of UUU-DZ-tFNA responding to UDG in vitro; Figure A shows the flowchart of UUU-DZ-tFNA detection of UDG, B shows the 1.5% AGE detection results showing the reactivity of UUU-DZ-tFNA to UDG (M: label, L1: UUU-DZ-tFNA, L2: UUU-DZ-tFNA+UDG), C shows the fluorescence spectral response of UUU-DZ-tFNA to different concentrations of UDG, and D shows the specificity of UUU-DZ-tFNA in detecting UDG.

[0037] Figure 5Figure A shows the kinetic characteristics of the probe's response to UDG; Figure B shows the capillary electrophoresis results showing the response of UUU-DZ-tFNA to UDG; and Figure C shows the gel images corresponding to the response of UUU-DZ-tFNA to UDG (L1: UUU-DZ-tFNA, L2: UUU-DZ-tFNA + UDG).

[0038] Figure 6 The results of UUU-DZ-tFNA detection of UDG in cytoplasm are shown in Figure A; Figure B shows the results of UUU-DZ-tFNA detection of UDG in cytoplasm, and Figure C shows the flow cytometry results. Figure C is a bar chart of the fluorescence signal of UUU-DZ-tFNA in Figure A. "***" P<0.001.

[0039] Figure 7 To detect UDG expression at the in vivo level using UUU-DZ-tFNA; Figure A shows the tumor fluorescence image within 100 min after intratumoral injection of UUU-DZ-tFNA in low-risk NB PDX mice, B shows the fluorescence intensity corresponding to Figure A, C shows the tumor fluorescence image within 100 min after intratumoral injection of UUU-DZ-tFNA in high-risk NB PDX mice, D shows the fluorescence intensity corresponding to Figure C, EF shows the fluorescence images of ex vivo tumors and organs obtained 100 min after injection of UUU-DZ-tFNA in low-risk and high-risk NB PDX mice, G shows the relative expression of UDG in anatomically dissected tumors quantitatively analyzed by ELISA, H shows the fluorescence image of ex vivo tumor tissue from low-risk NB PDX mice (scale bar: 200 μm), and I shows the fluorescence image of ex vivo tumor tissue from high-risk NB PDX mice (scale bar: 200 μm). The data in the figures are mean ± standard deviation (n=3), and "***" P < 0.001.

[0040] Figure 8 The results of UUU-DZ-tFNA detection of UDG in exosomes are shown in Figure A, where the results of plasma exosome UDG content determination by UDG ELISA kit are shown in Figure B, the results of plasma exosome UDG content determination by UUU-DZ-tFNA are shown in Figure C, and the correlation analysis between ELISA results and UUU-DZ-tFNA results is shown in Figure C.

[0041] Figure 9 Transmission electron microscopy observation of exosomes in NB plasma, scale bar: 100 nm.

[0042] Figure 10 Heatmaps of total UDG, exosomal UDG, and plasma free UDG activity levels in serum samples from low- to intermediate-risk NB children (n=19) and high-risk NB children (n=11), and the percentage of UDG content in exosomals.

[0043] Figure 11 The enzyme activities in serum samples from low- to intermediate-risk NB (LIR) and high-risk NB (HR) patients are shown in Figure A, where total UDG enzyme activity is shown in Figure B, EV UDG enzyme activity is shown in Figure C, and free UDG enzyme activity is shown in Figure C. "*" P < 0.05.

[0044] Figure 12 ROC curves for total plasma UDG, plasma exosome UDG, and plasma free UDG used for NB risk stratification.

[0045] Figure 13 The heatmap shows the relative abundance of the four clinical indicators selected in low-to-intermediate-risk (LIR-NB) and high-risk (HR-NB) NB patients.

[0046] Figure 14 for MYCN ROC curves of amplification status, metastasis, neuron-specific enolase, and lactate dehydrogenase are used to differentiate between low- and intermediate-risk neoplasms (NBs) and high-risk NBs.

[0047] Figure 15 The ROC curves of clinical indicators combined with non-invasive prediction models were compared using 11 different machine learning algorithms. The horizontal axis represents the false positive rate, the vertical axis represents the true positive rate, AUC is the area under the receiver operating characteristic curve, and 95% CI is the 95% confidence interval of the AUC value.

[0048] Figure 16 Four clinical multidimensional indicators were used to predict the risk stratification of NB; Figure A shows the ROC curve of the combination of four clinical multidimensional indicators for the differentiation between low- and intermediate-risk NB and high-risk NB, and Figure B shows the diagnostic efficacy of individual clinical indicators through confusion matrix analysis.

[0049] Figure 17 The ROC curves of the CB-CPM model were compared after analysis by 11 different machine learning algorithms. The horizontal axis represents the false positive rate, the vertical axis represents the true positive rate, AUC is the area under the receiver operating characteristic curve, and 95% CI is the 95% confidence interval of the AUC value.

[0050] Figure 18 The combination of plasma exosome UDG and clinical multidimensional indicators was used to predict the risk stratification of NB. Figure A shows the ROC curve of the combined prediction model of exosome UDG and clinical multidimensional indicators for the differentiation between low- and intermediate-risk NB and high-risk NB. Figure B shows the confusion matrix analysis of the training set, which shows the diagnostic efficacy of the model. Figure C shows the confusion matrix analysis of the validation set, which shows the diagnostic efficacy of the model.

[0051] Figure 19 The present invention is a non-invasive risk grading prediction platform for NB, wherein Figure A is the homepage of the prediction platform, B is the calculation result of Example 2, and C is the calculation result of Example 2. Detailed Implementation

[0052] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0053] Unless otherwise specified, the experimental methods used in the following experimental examples are conventional methods; the materials and reagents used are commercially available unless otherwise specified.

[0054] Example 1: Detection of UDG expression levels in cells, animals, and clinical plasma

[0055] 1. Extraction and determination of UDG in cytoplasm at the cellular level

[0056] Two NB cell lines (SK-N-BE(2) and SK-N-AS), a drug-resistant NB cell line (BE(2)-DDP) and a normal cell line (293T) were cultured in DMEM containing 10% fetal bovine serum (FBS), penicillin (100 μg / mL) and streptomycin (100 μg / mL) at 37°C and 5% CO2. After the cells reached the logarithmic growth phase, the cell pellet was collected after trypsin digestion and cytoplasmic protein extraction was performed using the nuclear / plasmic protein extraction kit (C510001) from Sangon Biotech Co., Ltd. (Shanghai, China). The extracted cytoplasmic proteins were measured using the UDG ELISA kit from Wuhan Enzyme Immunosorbent Assay Biotechnology Co., Ltd. (Wuhan, China).

[0057] 2. Extraction and determination of UDG in cytoplasm at the tissue level

[0058] SPF-grade female BALB / c nude mice (3-4 weeks old, 14-29 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China) and housed in an SPF-grade animal facility under the following environmental conditions: temperature 22 ± 2℃, humidity 50 ± 5%, 12 h light / dark cycle, and free access to food and water. After 14 days of acclimatization, experiments were conducted. Four-week-old female BALB / c nude mice were subcutaneously inoculated with 6 × 10⁶ mol / L mol / L scalp filaments in the right scapular region. 6 SK-N-BE(2), SK-N-AS, BE(2)-DDP (50 μL PBS / Matrelle mixture, volume ratio 1:1); when the tumor volume grows to approximately 150-300 mm 3Mice were euthanized, and 20 mg of tumor tissue was collected. Tissue homogenate was prepared, and cytoplasmic proteins were extracted from the tissue cells using the Cell Nuclear / Plasma Protein Extraction Kit (C510001) manufactured by Sangon Biotech Co., Ltd. (Shanghai, China). The extracted cytoplasmic proteins were then measured using the UDG Enzyme-Linked Immunosorbent Assay (ELISA) kit manufactured by Wuhan Enzyme Immunosorbent Assay Biotechnology Co., Ltd. (Wuhan, China).

[0059] 3. Measurement of plasma UDG levels

[0060] Peripheral blood was collected from 36 children with neonatal neuropathy (NB), with 1 mL collected from each case. The blood was placed in a centrifuge tube containing EDTA-K2 and gently inverted 3-5 times to mix. The plasma was centrifuged at 3000 r / min for 15 min at 4°C, carefully aspirating the upper pale yellow plasma layer, avoiding contact with the leukocyte layer and erythrocyte sediment. The plasma was aliquoted into sterile EP tubes and stored at -80°C for later use. Before testing, the plasma was slowly thawed at 4°C, then centrifuged at 12000 r / min for 5 min to remove any potential precipitate. The supernatant was collected and analyzed using the UDG ELISA kit from Wuhan Enzyme Immunosorbent Assay Biotechnology Co., Ltd. (Wuhan, China).

[0061] The test results of the NB cells, tissues, and plasma samples obtained through the above steps are as follows:

[0062] (1) Results of standard curve plotting

[0063] Plotting UDG standard concentration (ng / mL) on the x-axis, corresponding to OD 450 The value of nm is used as the ordinate. A standard curve was plotted using Origin software, and the fitting equation was Y = 0.0741 + 0.1544X, with a correlation coefficient R0. 2 =0.95, linear range is 0.156-10 ng / mL ( Figure 1 ).

[0064] (2) Results of UDG content determination in different samples

[0065] To clarify the expression patterns of UDG in the development and progression of tumors and its potential clinical application value, this study systematically detected the expression levels of UDG at three levels: cellular, tissue, and clinical plasma. The results showed that the expression level of UDG in the cytoplasm of NB cells was significantly higher than that in the cytoplasm of normal cells. Figure 2 A). The expression level of UDG in the cytoplasm of drug-resistant NB cells was significantly higher than that in drug-sensitive NB cells. Figure 2 B); further tissue-level validation confirmed that the expression level of UDG in the cytoplasm of tumor tissue was significantly higher than that in normal tissue, and the expression level of UDG in drug-resistant NB tissue was further increased (B). Figure 2CD); Clinical plasma sample analysis showed that there were significant differences in plasma UDG levels between high-risk and low-risk children, and plasma UDG levels in drug-resistant NB children were significantly higher than those in untreated NB children. Figure 2 (EF). In summary, UDG is expected to serve as an indicator for risk grading in children with NB.

[0066] Example 2: Synthesis and Characterization of UDG Detection Probe

[0067] Currently, classic ELISA-based UDG detection methods can only detect UDG at the in vitro level and cannot detect UDG expression in situ in cells, animals, or exosomes. Therefore, this patent constructs an in situ UDG detection probe (UUU-DZ-tFNA). The specific synthesis and characterization results of UUU-DZ-tFNA are as follows:

[0068] (1) First, four single-stranded DNAs (S1-S4) were mixed in equal amounts (1 μL each, 100 μM) in 18 μL of TEM buffer (20 mM Tris-HCl, 2 mM EDTA, 12.5 mM MgCl2·6H2O) and 38 μL of DEPC-treated water (pH 7.4). The system was heated to 95 °C for 10 min to denature, and then rapidly cooled to 25 °C at a rate of -1 °C / min to allow it to self-assemble into DNA tetrahedrons (tFNA).

[0069] (2) Next, S5 (1 μL, 100 μM) and S6 (1.2 μL, 100 μM) were mixed in 3 μL of TEM buffer and 4.8 μL of DEPC-treated water, heated to 95°C for 10 min to denature, and then rapidly cooled to 25°C at a rate of -1°C / min to form the S5-S6 complex.

[0070] (3) Subsequently, S7 (3 μL, 100 μM) was mixed with 9 μL of TEM buffer and 18 μL of DEPC-treated solution to prepare HP2 (heated to 95°C and held for 10 min to achieve denaturation, and then rapidly cooled to 25°C at a rate of -1°C / min).

[0071] (4) Finally, tFNA, S5-S6, and HP2 were incubated together at 37°C for 30 min to hybridize and obtain the UUU-DZ-tFNA complex. The assembly of UUU-DZ-tFNA was verified by agarose gel electrophoresis (AGE) at 120 V for 50 min. The morphology of UUU-DZ-tFNA was further analyzed by atomic force microscopy (AFM).

[0072] Figure 3A is a schematic diagram of the synthesis of UUU-DZ-tFNA. Figure 3 B and 3C were characterized by AGE and AFM, respectively, for the synthesis process, morphology, and particle size of UUU-DZ-tFNA. The results showed that the average particle size was 44.80±1.46 nm, proving that UUU-DZ-tFNA was successfully synthesized.

[0073] Table 1. Sequences used in the synthesis of UUU-DZ-tFNA

[0074]

[0075] Note: Fluorescent group - quencher group: Cy5-BHQ3.

[0076] The sequences used for the synthesis of UUU-DZ-tFNA are shown in Table 1.

[0077] Example 3: Feasibility of in vitro target detection using UUU-DZ-tFNA

[0078] (1) Dilute UUU-DZ-tFNA (100 μM) to 10 µM in buffer (20 mM Tris-HCl, 2 mM EDTA and 12.5 mM MgCl2·6H2O, pH 7.4) to a final volume of 100 μL.

[0079] (2) Subsequently, the diluted solution was reacted with appropriate concentrations of UDG and Zn. 2+ The solutions were mixed and incubated at 37°C for 3 h. The reaction system was characterized by AGE and capillary electrophoresis. The kinetics and fluorescence spectra of the reaction system were determined by fluorescence spectroscopy using a microplate detection system (Spectra Max i3x) at an excitation wavelength of 655 nm.

[0080] (3) In the UDG sensitivity test, UUU-DZ-tFNA was added to UDG solutions with concentrations ranging from 0 U / mL to 2 U / mL (0 U / mL, 0.00005 U / mL, 0.0005 U / mL, 0.001 U / mL, 0.01 U / mL, 0.1 U / mL, 0.5 U / mL, 1 U / mL and 2 U / mL), and Zn was added simultaneously. 2+ (20 μM), incubated at 37°C for 3 h.

[0081] (4) Add the same dose (2 U / mL) of endonuclease (uracil DNA glycosylase (UDG), DNase I, RNase A, APE1), 100 nM UUU-DZ-tFNA, and 20 μM Zn to the reaction system. 2+The cells were incubated for 3 hours, and their fluorescence intensity was measured to investigate the specificity of UUU-DZ-tFNA in detecting UDG.

[0082] Figure 4 and Figure 5 Experimental results show that the UUU-DZ-tFNA detection probe has good sensitivity and specificity, proving the feasibility of using UUU-DZ-tFNA to detect UDG in vitro.

[0083] Example 4: Detection of UDG at the UUU-DZ-tFNA cellular level

[0084] 293T, SK-N-BE(2), and SK-N-AS were seeded in 20 mm glass-bottom confocal culture dishes and cultured at 37°C and 5% CO2 for 12 h to allow them to adhere. After serum starvation treatment, they were first treated with 50 µM Zn 2+ Cells were treated with the solution in serum-free DMEM medium for 30 min, followed by treatment with 200 nM UUU-DZ-tFNA complex in serum-free DMEM medium for 4 h. Unbound complex was washed three times with PBS. The cytoskeleton and nuclei were stained with Tubulin and Hoechst, respectively. Cell images were observed using a 40x objective lens on an Axiovert 5 microscope.

[0085] The above results indicate that UUU-DZ-tFNA can detect UDG in the cytoplasm, and the UDG content in the cytoplasm of NB cells is significantly higher than that in normal control cells. Figure 6 ).

[0086] Example 5: Detection results of UDG at the in vivo level using UUU-DZ-tFNA

[0087] (1) SPF-grade female NOG mice (3-4 weeks old, 12-14 g) were selected from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China) and housed in an SPF-grade animal room. The environmental conditions were 22±2℃, 50±5%, 12 h light / dark cycle, and free access to food and water. After 14 days of acclimatization, the experiment was conducted.

[0088] (2) The surgically removed NB cell tumor tissue was immediately placed in sterile ice bath PBS containing penicillin 100 U / mL + streptomycin 100 μg / mL and transferred to the laboratory within 30 min. In a laminar flow hood, the tissue was rinsed 3 times with sterile PBS to remove surface blood, necrotic tissue and connective tissue, and the treated tumor tissue was cut into 1-2 mm pieces. 3 The tissue block is prepared for later use.

[0089] (3) The cut tumor tissue block was implanted subcutaneously into the mouse, the skin incision was sutured, and the wound was disinfected with iodine. After the operation, the mice were housed in cages of 2-3. The mental state, feeding and wound healing were observed daily. The observation was continued for 7 days to confirm no infection. Starting from the 7th day after transplantation, the tumor growth was monitored every 3 days using calipers, and the volume was calculated using the formula V=(length×width) / 2.

[0090] (4) When the tumor volume reaches 50-100 mm 3 Mice were euthanized by cervical dislocation, and tumor tissue was aseptically removed and cut into 1-2 mm pieces according to the tissue processing method described above. 3 The tissue blocks were transplanted subcutaneously into new NOG mice to complete the model passage, designated as generation F1. Subsequent experiments can use generation F1-F3 xenografts as needed to obtain the NB PDX model.

[0091] (5) When the tumor volume grows to approximately 50-100 mm 3 PDX model mice were intratumorally injected with UUU-DZ-tFNA (100 μL, 1 μM). Images were captured at 1 min, 10 min, 20 min, 30 min, 60 min, 80 min, and 100 min after injection using an AniView 600 in vivo imaging system (excitation wavelength: 625 nM, emission wavelength: 680 nM). Immediately after mouse sacrifice, tumors and major organs were harvested for ex vivo imaging. The excised tumors were prepared into 3–4 µM thick sections and subsequently imaged using a fluorescence microscope (Leica, Germany).

[0092] In vivo fluorescence imaging showed that within 100 minutes after injection, the fluorescence signal in both high-risk and low-risk NB tumor tissues gradually increased over time, reaching a peak intensity at 80 minutes. Frozen section analysis showed that compared to low-risk NB PDX tumor tissues, high-risk NB PDX tumor tissues exhibited significantly increased UUU-DZ-tFNA reaction signals, a result consistent with in vivo imaging findings. These results further demonstrate that UDG can serve as an indicator for NB risk grading. Figure 7 ).

[0093] Example 6: Accuracy of UUU-DZ-tFNA in detecting UDG expression in exosomes

[0094] (1) Plasma exosomes were extracted using the plasma exosome extraction kit (R603) from Nanjing Novizan Biotechnology Co., Ltd. (Nanjing, China);

[0095] (2) Resuspend the purified exosome precipitate in RIPA lysis buffer and incubate on ice for 30 min, during which vortexing is performed to promote the rupture of exosome membranes and the release of enzymes. After lysis, centrifuge at high speed at 4°C (e.g., 12,000 r / min, 10 min) and take the supernatant for later use.

[0096] (3) Take the supernatant after lysis and use the UDG ELISA kit from Wuhan Enzyme Immunosorbent Biotechnology Co., Ltd. (Wuhan, China) to measure it. Substitute the OD value of the sample into the regression equation of the standard curve to calculate the concentration of the target enzyme in the sample.

[0097] (4) Exosomes were incubated with UUU-DZ-tFNA for 3 h, and UDG expression in exosomes was obtained by fluorescence signal.

[0098] The experimental results above show that the detection results of the UUU-DZ-tFNA in situ detection method are highly consistent with those of the gold standard ELISA method; correlation analysis results show that the detection data of the two methods are significantly linearly correlated (R0). 2 =0.9097, P<0.0001) further confirms the accuracy and reliability of UUU-DZ-tFNA in detecting UDG in exosomes. Figure 8 ).

[0099] Example 7: Determination of total plasma UDG expression, plasma exosome UDG, and plasma free UDG content

[0100] (1) Determination of total UDG content in plasma

[0101] 90 μL of plasma was added to 10 μL of reaction buffer containing 1 μM UUU-DZ-tFNA. After incubation at 37°C for 3 h, the emission spectrum of the solution was collected using a microplate reader.

[0102] (2) Extraction of plasma exosomes and determination of exosome UDG content

[0103] 500 μL of plasma was collected and plasma exosomes were extracted using the plasma exosome extraction kit (R603) from Nanjing Novizan Biotechnology Co., Ltd. (Nanjing, China), and observed under a transmission electron microscope.

[0104] To assess the UDG content in exosomes of low- and intermediate-risk and high-risk NBs, 90 μL of exosomes resuspended in PBS buffer were added to 10 μL of reaction buffer containing 1 μM UUU-DZ-tFNA. After incubation at 37°C for 3 h, the emission spectra of the solution were collected using a microplate reader.

[0105] (3) Calculation of plasma free UDG content

[0106] Total plasma UDG is composed of plasma exosome UDG and plasma free UDG, therefore plasma free UDG is the difference between the two.

[0107] Experimental results showed that exosomes under transmission electron microscopy exhibited a typical cup-shaped vesicle morphology with visible surface depressions and relatively clear membrane edges, confirming the successful extraction of plasma exosomes. Figure 9 Thermographic results indicate that 60.0%-86.8% of UDG in plasma exists in exosomes. Figure 10 The levels of total UDG, EV-related UDG, and free UDG in plasma samples were significantly higher in the high-risk group than in the intermediate- and low-risk groups. Figure 11 Exosomal UDG has the greatest value in NB risk stratification. Figure 12 The above findings demonstrate that plasma UDG in NB patients shows promise as a risk stratification indicator for NB.

[0108] In summary, plasma exosome UDG shows promise as an indicator for non-invasive risk stratification of neonatal nephropathy (NB), but its sensitivity and specificity are unsatisfactory. To improve the sensitivity and specificity of NB risk stratification, this invention further integrates plasma exosome UDG results with multi-dimensional clinical indicators and establishes a predictive model for NB non-invasive risk stratification using machine learning algorithms.

[0109] Example 8: Plasma MYCN Copy number quantification method

[0110] By screening more than 10 clinical indicators for non-invasive prenatal testing (NB), four indicators with the most significant differences were selected for use in the subsequent construction of the NB non-invasive risk grading model. MYCN Amplification status, whether the tumor has metastasized, neuron-specific enolase (NSE), and lactate dehydrogenase (LDH).

[0111] plasma MYCN The method for quantifying copy number is as follows:

[0112] (1) Extraction of cell-free DNA from plasma: Cell-free DNA was extracted from plasma using the blood / cell / tissue genomic DNA extraction kit (DP304) from Tiangen Biotech (Beijing) Co., Ltd.

[0113] (2) RT-PCR detection of plasma MYCN Copy number. The detection method was based on the paper published by the applicant's team (CancerReports 2023, 6(2): e1688.). The RT-PCR detection results were compared with the clinical FISH results.

[0114] Table 2 NB tissue and plasma MYCN Amplification Status Correspondence Result Table

[0115]

[0116] Table 2 results indicate that plasma and tissue levels in NB patients... MYCN The amplification status is consistent. Therefore, in order to achieve non-invasive risk stratification of NB, this invention... MYCN Free plasma used in the amplification state MYCN Gene amplification status.

[0117] Example 9: Predictive Value of Individual Clinical Multidimensional Indicators for NB Risk Staging

[0118] The model building steps are as follows:

[0119] 1. Screening of clinical indicators

[0120] Based on the international NB hazard classification index, a dual screening condition is set:

[0121] ① It is non-invasive (it can be obtained without invasive procedures, meeting the needs of convenient clinical application);

[0122] ② Statistical analysis revealed a significant difference between the low-to-intermediate risk (LIR) group and the high-risk (HR) group (P<0.05). Four independent predictive indicators were ultimately selected as the basis for subsequent modeling.

[0123] 2. Validation of Independent Predictive Performance of Single Indicators

[0124] For the four non-invasive indicators selected, receiver operating characteristic (ROC) curves were plotted using the pROC package in R language. The area under the curve (AUC) of each indicator in predicting the risk of non-invasive brain injury (NIB) (low to medium risk / high risk) was calculated to evaluate the independent predictive value of individual indicators and provide a reference for subsequent multi-indicator joint modeling.

[0125] 3. Machine learning model construction and training

[0126] Four non-invasive indicators were combined to construct a Clinical-Only Combined Predictive Model (CO-CPM). Eleven commonly used machine learning classification methods were selected (Logistic Regression (LR), Linear Discriminant Analysis (LDA), Elastic Net (Enet), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), Multilayer Perceptron (MLP), Neural Network (NN), Naive Bayes (NB), Flexible Discriminant Analysis (FDA), and Classification and Regression Tree (CART)). The four selected non-invasive feature variables were combined to form a unified input feature set. Based on this feature set, eleven machine learning model classifiers were constructed, forming eleven sets of classification and prediction schemes based on "single non-invasive feature set - different algorithm models". ROC curves were plotted for 11 algorithms, and the core performance indicators (AUC, sensitivity, specificity, accuracy, Kappa value, F1 score, etc.) of each model were extracted for comprehensive comparison. Based on the comprehensive performance of the indicators, the optimal prediction model was selected.

[0127] Confusion matrix construction and model performance evaluation: For the optimal model (NN model) selected through screening, the confusion matrix is ​​calculated to quantify the true positive, true negative, false positive, and false negative results of the model in the low-to-medium risk / high-risk classification, and to further verify the classification reliability and clinical application value of the model.

[0128] Four indicators that are both non-invasive and statistically significant (P<0.05) were successfully selected from the international NB risk classification index. These indicators are plasma. MYCN The four indicators—amplification status, tumor metastasis, neuron-specific enolase (NSE), and lactate dehydrogenase (LDH)—showed clear differences between the low- and intermediate-risk groups and possessed independent risk prediction potential. Figure 13 ).

[0129] Single-indicator prediction results showed that all four non-invasive indicators exhibited certain predictive ability for NB risk, with AUC values ​​all greater than 0.7, validating the effectiveness of each indicator. Among them, the single-factor model corresponding to serum NSE level showed relatively better predictive efficacy, with an area under the ROC curve (AUC) of 0.81. At the optimal cutoff value, the sensitivity and specificity were 81.8% and 89.5%, respectively, which could, to some extent, distinguish between high-risk and low-risk NB infants. Figure 14 However, the predictive performance of a single indicator has limitations and does not meet the requirements for high-precision clinical applications. The results of a multi-algorithm combined model show that the predictive model constructed by combining 11 machine learning algorithms for four non-invasive indicators all exhibited superior predictive performance compared to single indicators. Among them, the neural network (NN) model had the best overall performance, with an area under the ROC curve (AUC) of 0.852, significantly higher than other algorithm models. It also demonstrated the best performance in core indicators such as sensitivity (72.7%), specificity (94.7%), and accuracy (86.7%), effectively distinguishing between low- and high-risk NB patients. Figure 15 ).

[0130] The confusion matrix results of the model showed that the model achieved a true positive rate (sensitivity) of 88.9% for high-risk NB patients and a true negative rate (specificity) of 89.5% for intermediate- and low-risk patients, but the overall accuracy was only 83.3%, indicating that the clinical safety and reliability of the model in non-invasive prediction of NB risk still need to be improved. Figure 16 Further research is needed to explore new methods for predicting the risk of non-invasive NB (notebook) procedures.

[0131] Example 10: The value of combining plasma exosome UDG with multiple clinical indicators for NB risk stratification

[0132] (1) Sample data processing:

[0133] Integrating the above-mentioned plasma exosome UDG with 4 clinical indicators (plasma) MYCN The amplification status, whether the tumor has metastasized, neuron-specific enolase (NSE), and lactate dehydrogenase (LDH) were used to form a combined non-invasive feature set containing five features; all features were obtained non-invasively.

[0134] (2) Joint model construction and optimization:

[0135] Using the 11 machine learning classification models mentioned above, with the joint feature set as the input feature and the clinical risk level (high risk / low intermediate risk) of NB children as the target variable, a "clinical-biomarker combined predictive model" (CO-CPM) was constructed. ROC curves for the 11 algorithms were plotted, and the core performance indicators of each model were extracted for comprehensive comparison. Based on the comprehensive performance of the indicators, the optimal predictive model was selected.

[0136] (3) Confusion matrix construction and comprehensive model performance evaluation:

[0137] For the optimal model (NN model) obtained through screening, a confusion matrix was calculated to quantify the true positive, true negative, false positive, and false negative results of the model in the low-to-medium risk / high-risk classification, further verifying the classification reliability and clinical application value of the model.

[0138] By comparing the core performance indicators (AUC, sensitivity, specificity, Youden index, and accuracy) of the combined prediction model of exosome UDG and clinical multidimensional indicators with those of the separate prediction model of clinical multidimensional indicators, we can clarify the performance advantages of the combined model and provide data support for the non-invasive clinical application of the model.

[0139] The predictive model constructed in this invention, which combines exosome UDG with multiple clinical indicators (using the optimal NN neural network algorithm), exhibits excellent comprehensive performance: the area under the receiver operating characteristic (AUC) of the model reaches 0.904, with a corresponding 95% confidence interval of 0.736-1, and the ROC curve closely follows the upper left corner of the coordinate system, indicating a strong ability to distinguish between the two types of patients. Figure 17 Based on confusion matrix analysis, the training set model achieved a sensitivity of 88.9%, a specificity of 100.0%, and an overall accuracy of 90.0%, with positive predictive value (PPV) and negative predictive value (NPV) of 100.0% and 95.0%, respectively. The validation set model achieved a sensitivity of 85.7%, a specificity of 100.0%, and an overall accuracy of 92.8%, with positive predictive value (PPV) and negative predictive value (NPV) of 100.0% and 87.5%, respectively, fully meeting the accuracy requirements for clinical stratification. In terms of comparative advantages, the AUC of this combined model improved by 5.2% compared to the best prediction result of the combined prediction model of individual clinical indicators, with significant improvements in both sensitivity and specificity, fully demonstrating the synergistic value of multi-indicator combination. Figure 18 ).

[0140] Application example: Use of the NB non-invasive risk grading prediction model

[0141] To facilitate the use of this model, we have built the "NB Non-invasive Risk Classification Prediction Platform" (website: zhoaliang123.shinyapps.io / shiny_app2 / ), and the specific operation method is as follows.

[0142] exist Figure 19 On the homepage of the "NB Non-invasive Risk Grading Prediction Platform" shown in Figure A, enter the child's real information in the data input field and click "Start Calculation". The reference values ​​for the five constructed indicators are shown in Table 3.

[0143] Table 3 Reference values ​​for 5 construction indicators

[0144]

[0145] Interpretation of prediction results:

[0146] The calculation will generate a corresponding prediction report, as shown below:

[0147] Prediction results:

[0148] Redicted HR probability: 1 (This indicates that the probability of the child having a high risk of NB is 100%).

[0149] Best cutoff: 0.057

[0150] Risk classification: Hisk Risk (indicating that the child's NB risk level is high).

[0151] Interpretation: The patient is predicted to be at high risk and may require closer monitoring.

[0152] Example 1:

[0153] Taking actual data on low-, medium-, and high-risk NB patients as an example:

[0154] 1) Data Input

[0155] The data input is as follows:

[0156] ①Plasma MYCN Gene amplification: Amplification 1;

[0157] ② Whether the tumor has metastasized: 0 for non-metastasis;

[0158] ③ Neuron-specific enolase: 33.98 ng / mL;

[0159] ④ Lactate dehydrogenase: 261 U / L;

[0160] ⑤ Plasma exosomal uracil DNA glycosylation enzyme: 0.087 U / mL.

[0161] 2) Click Calculate

[0162] Click "Start Calculation" within the red box.

[0163] 3) Generate report

[0164] After clicking "Start Calculation" in the red box, a "Prediction Results" report will be automatically generated, and the results will be as follows: Figure 19 As shown in B:

[0165] The results showed that the patient with NB was predicted to have low or intermediate risk.

[0166] II. Example 2:

[0167] Taking actual high-risk NB patients as an example:

[0168] 1) Data Input

[0169] The data input is as follows:

[0170] ①Plasma MYCN Gene amplification: Amplification 1;

[0171] ② Whether the tumor has metastasized: Metastasis 1;

[0172] ③ Neuron-specific enolase: 326 ng / mL;

[0173] ④ Lactate dehydrogenase: 581 U / L;

[0174] ⑤ Plasma exosomal uracil DNA glycosylation enzyme: 0.2 U / mL.

[0175] 2) Click Calculate

[0176] Click "Start Calculation" within the red box.

[0177] 3) Generate report

[0178] After clicking "Start Calculation" in the red box, a "Prediction Results" report will be automatically generated, and the results will be as follows: Figure 19 As shown in C:

[0179] The results indicate that the predicted risk for this NB child is high. The patient is expected to be at higher risk and may require closer monitoring.

[0180] This invention integrates plasma exosome UDG data with multidimensional clinical indicators to construct a non-invasive risk grading prediction model for neonatal nephropathy (NB), achieving a breakthrough improvement in the predictive efficacy of NB risk grading.

[0181] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A probe for non-invasive risk grading of neuroblastoma, characterized in that, The preparation method is as follows: (1) Four single-stranded DNAs S1, S2, S3 and S4 were uniformly mixed in TEM buffer, denatured and cooled, and assembled into DNA tetrahedral tFNA; (2) DNA single strand S5 and S6 are mixed in TEM buffer, denatured and cooled to form S5-S6 complex; (3) Mix single-stranded DNA S7 with TEM buffer, and then denature and cool to prepare HP2; (4) After mixing tFNA, S5-S6 complex and HP2, incubate to obtain probe UUU-DZ-tFNA.

2. The probe according to claim 1, characterized in that: In step (1), the sequence of DNA single strand S1 is shown in SEQ ID No. 1, the sequence of S2 is shown in SEQ ID No. 2, the sequence of S3 is shown in SEQ ID No. 3, and the sequence of S4 is shown in SEQ ID No. 4; the concentrations of S1, S2, S3 and S4 are all 1-100 μM and the molar ratio is 1:1:1:

1.

3. The probe according to claim 2, characterized in that: In step (2), the sequence of DNA single strand S5 is shown in SEQ ID No. 5; the sequence of S6 is shown in SEQ ID No. 6, and the nucleotides at positions 7, 10, 11, 14 and 18 from the 5' end are deoxyuridine dU; the concentrations of S5 and S6 are both 1-100 μM, and the molar ratio is 1:

1.

4. The probe according to claim 3, characterized in that: In step (3), the DNA single-strand S7 sequence is shown in SEQ ID No.

7. The sequence has a Cy5 fluorescent group linked between the 24th and 25th nucleotides from the 5' end, the 39th nucleotide is riboadecanoside rA, and the 3' end is modified with a BHQ3 fluorescent quencher group; the concentration of S7 is 3-300 μM.

5. The probe according to claim 4, characterized in that: The TEM buffer solution contains 20 mM Tris-HCl, 2 mM EDTA, and 12.5 mM MgCl2·6H2O; denaturation is performed at 95°C for 10 min, cooling is performed at a rate of -1°C / min to 25°C, and incubation is performed at 37°C for 30 min.

6. A reagent or kit, characterized in that: It includes the probe UUU-DZ-tFNA as described in any one of claims 1-5.

7. A combination of markers for non-invasive risk grading of neuroblastoma, characterized in that: The marker combination includes plasma MYCN Whether amplification occurs, whether the tumor has metastasized, the content of neuron-specific enolase, the content of lactate dehydrogenase, and the content of plasma exosomal uracil DNA glycosylation enzyme; the content of plasma exosomal uracil DNA glycosylation enzyme is obtained by the reagent or kit described in claim 6.

8. A method for predicting non-invasive risk grading of neuroblastoma using the marker combination described in claim 7, characterized in that, The steps are as follows: 1) Collect feature data of the combination of markers in the sample to be tested to form a joint non-invasive feature set; 2) Input the combined non-invasive feature set into the neural network model, and calculate the cut-off value corresponding to the maximum value of the Youden index based on the ROC curve of the neural network model to determine the risk level of the sample to be tested.

9. The prediction method according to claim 8, characterized in that: The Youden index, by simultaneously considering sensitivity and specificity, reflects the overall discriminative power of the model at different thresholds. Within the ROC curve framework, the maximum value of the Youden index can determine the threshold at which the sum of sensitivity and specificity reaches its maximum value. Therefore, it is used to determine the optimal cut-off value for continuous risk predictions, where Youden index = sensitivity + specificity - 1.

10. The prediction method according to claim 9, characterized in that: If the risk score of the sample output by the neural network model is higher than the cut-off value corresponding to the maximum value of the Youden index, it is identified as a high-risk sample; if the risk score of the sample output by the neural network model is lower than the cut-off value corresponding to the maximum value of the Youden index, it is identified as a medium- or low-risk sample.

Images