Multi-protein marker combination and machine learning algorithm for intervertebral disc degeneration prediction and severity evaluation

Abstract

The invention belongs to the technical field of biological medicine, and particularly discloses a multi-protein marker combination for intervertebral disc degeneration prediction and severity evaluation and a machine learning algorithm. The marker combination comprises any three or more of COL6 [alpha] 3, REG1beta, ATF5, CAP1, MAGEA4 and LILRB3, and preferably six proteins are combined for use. And quantitatively detecting the expression level of the protein in the blood plasma through enzyme-linked immunosorbent assay (ELISA), constructing a prediction model in combination with a binary Logistic regression algorithm, and outputting an intervertebral disc degeneration grading result. Experiments show that the AUC (area under curve) of joint detection of six proteins reaches 0.870, which is averagely improved by 11.8% compared with that of single protein detection, and the AUC is obviously superior to that of an existing imaging grading method. According to the present invention, the intervertebral disc degeneration plasma protein diagnosis system is established based on the SOMAscan technology for the first time, the protein concentration threshold of the ELISA verification is provided, the efficient complementation with the MRI Pfirrmann grading is achieved, the advantages of noninvasive property, objective property and high sensitivity are provided, and the new method is provided for the clinical early screening and dynamic monitoring.

Description

Technical Field

[0001] The present invention belongs to the field of biomedicine technology, and specifically discloses a multi-protein marker combination and a machine learning algorithm for predicting and assessing the severity of intervertebral disc degeneration. Background Art

[0002] The intervertebral disc is a critical load-bearing structure in the human body, and its extracellular matrix plays a vital role in physiological functions, such as absorbing mechanical stress, maintaining joint stability, and protecting nerve roots. Intervertebral disc degeneration (IDD), a major cause of low back pain and joint dysfunction, is characterized by irreversible structural changes such as intervertebral disc stenosis, cartilage endplate calcification, proteoglycan loss, and nucleus pulposus dehydration, which severely impact patients' quality of life. Existing studies have shown that the development and progression of IDD are closely associated with genetic susceptibility, abnormal mechanical stress, and unhealthy lifestyle habits such as smoking and nutritional imbalance, but the underlying molecular mechanisms remain largely undefined. In clinical diagnosis, the Pfirrmann grading system based on lumbar MRI is widely used. However, this technique relies on subjective grading based on T2-weighted signal intensity and disc morphology, resulting in inherent limitations such as high operational complexity and inconsistent evaluation criteria. Although subsequent studies have optimized the grading criteria for elderly patients, the objectivity of imaging assessment is still limited by human interpretation errors. It is worth noting that plasma samples have shown important value in the screening of biomarkers for diseases such as tumors due to their advantages such as stable components and convenient collection. In existing technologies, although high-throughput proteomics methods based on mass spectrometry can effectively enrich low-abundance functional proteins, they still face the technical bottleneck of high-abundance proteins interfering with detection sensitivity. Currently, there are a series of new blood proteomics methods on the market, such as the aptamer-based protein detection technology developed by somalogic, which achieves quantitative analysis by specifically binding to target proteins, effectively improving the detection efficiency of low-abundance proteins. However, there are few cases of transformation that combine emerging technologies with clinical testing, and there is a lack of related scientific research transformation products. Summary of the Invention

[0003] To address the above problems, the present invention discloses a multi-protein marker combination and a machine learning algorithm for predicting and assessing the severity of intervertebral disc degeneration.

[0004] In order to achieve the above purpose, the present invention adopts the following technical solutions:.

[0005] A multi-protein marker combination for predicting and assessing the severity of intervertebral disc degeneration, the marker combination comprising any three or more of the following six plasma proteins:

[0006] type VI collagen α3 chain (COL6α3), regenerating islet-derived protein 1β (REG1β), transcription activator 5 (ATF5), adenylate cyclase-associated protein 1 (CAP1), melanoma antigen family A4 (MAGEA4), and leukocyte immunoglobulin-like receptor subfamily B member 3 (LILRB3);

[0007] The marker combination is used to predict the progression of intervertebral disc degeneration or assess its severity by quantitatively detecting the expression levels of the proteins in plasma;

[0008] The severity was divided into grades II to V based on the Pfirrmann grading system.

[0009] Furthermore, the above-mentioned multi-protein marker combination for prediction and severity assessment of intervertebral disc degeneration consists of COL6α3, REG1β, ATF5, CAP1, MAGEA4 and LILRB3.

[0010] Furthermore, the expression level of the above-mentioned multi-protein marker combination for prediction and severity assessment of intervertebral disc degeneration is verified by enzyme-linked immunosorbent assay (ELISA), wherein the ELISA uses monoclonal antibodies to quantitatively detect COL6α3, REG1β, ATF5, CAP1, MAGEA4 and LILRB3.

[0011] Furthermore, the predictive efficacy of the above-mentioned multi-protein marker combination for prediction and severity assessment of intervertebral disc degeneration was evaluated by the receiver operating characteristic (ROC) curve, wherein the area under the curve (AUC) of the combined detection of the six proteins was ≥0.870.

[0012] The present invention also discloses the use of the multi-protein marker combination in preparing a reagent for predicting intervertebral disc degeneration and assessing its severity.

[0013] The present invention also discloses the use of the multi-protein marker combination in preparing a reagent that is complementary to the Pfirrmann grading result of MRI imaging and is used for early screening or dynamic monitoring of intervertebral disc degeneration.

[0014] The present invention also discloses a system for predicting intervertebral disc degeneration based on the above-mentioned multi-protein marker combination, the system comprising:

[0015] (a) a plasma sample collection device for obtaining anticoagulated whole blood from a subject and separating plasma;

[0016] (b) a protein detection module, used to quantitatively analyze the expression levels of the six proteins in the marker combination;

[0017] (c) Data processing module, which builds a prediction model based on machine learning algorithm and outputs the intervertebral disc degeneration grading results.

[0018] Furthermore, in the above system, the machine learning algorithm is a binary logistic regression algorithm, the input of which is the protein concentration value, and the output is the probability of intervertebral disc degeneration and the grading result.

[0019] Furthermore, in the above system, the construction of the binary logistic regression model includes the following steps:

[0020] (i) Determine the optimal cutoff value of a single protein using the ROC curve and screen for proteins with significant differences;

[0021] (ii) The multi-protein combination was optimized using stepwise regression and the AUC value of the joint prediction was calculated;

[0022] (iii) Model parameters were calibrated based on the ELISA data of the validation cohort to improve prediction accuracy.

[0023] Furthermore, in the above system, the data processing module further includes a visualization interface for displaying the ROC curve, AUC value and protein concentration threshold, wherein the cutoff value of COL6α3 is ≥22.73 mg / ml, MAGEA4 is ≥3.9 mg / ml, REG1β is ≥87.5 mg / ml, ATF5 is ≥1.67 mg / ml, LILRB3 is ≥10.12 mg / ml, and CAP1 is ≥3.838 mg / ml.

[0024] Compared with the prior art, the present invention has the following outstanding beneficial effects:

[0025] The present invention discloses a multi-protein marker combination and machine learning algorithm for the prediction and severity assessment of intervertebral disc degeneration, and innovatively proposes a dual-modal evaluation system that combines plasma multi-omics analysis with imaging features. By screening differentially expressed proteins that are closely related to the process of intervertebral disc degeneration, a grading prediction model based on a machine learning algorithm is constructed, aiming to break through the subjective limitations of existing imaging grading methods. The embodiment experiments show that the area under the curve (AUC) of the combined detection of six proteins reached 0.870, which is an average increase of 11.8% over the AUC of single protein detection, and is significantly better than the existing imaging grading method. The present invention is the first to establish an intervertebral disc degeneration plasma protein diagnostic system based on SOMAscan technology, provide a protein concentration threshold verified by ELISA, and achieve efficient complementarity with MRIPfirrmann grading. It has the advantages of being non-invasive, objective, and highly sensitive, and provides a new method for early clinical screening and dynamic monitoring.

[0026] This invention simultaneously achieves the following technical advantages: (1) Utilizing high-throughput proteomics technology to capture the dynamic changes of low-abundance functional proteins, revealing early molecular events in IDD; (2) Establishing a quantifiable biomarker combination to improve the objectivity and repeatability of degeneration grading; and (3) Reducing clinical screening costs through non-invasive detection methods, providing technical support for dynamic monitoring of disease progression. This invention provides a new technical path for the precise diagnosis and treatment of intervertebral disc degeneration and the development of personalized intervention strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 : The main technical route of the present invention;

[0028] Figure 2 :The impact of grading on patients with intervertebral disc degeneration;

[0029] Figure 3 : Proteomics Overview;

[0030] Figure 4 : Differential protein pathway enrichment analysis;

[0031] Figure 5 : relative expression values ​​of differentially expressed proteins in proteomics;

[0032] Figure 6 : ELISA verification of differentially expressed proteins;

[0033] Figure 7 : ROC curve was used to evaluate the prediction effect. DETAILED DESCRIPTION

[0034] like Figure 1 As shown, the present invention mainly includes three parts: 1) collection of patient samples; 2) plasma proteomics identification using Soma Scan7k assay blood proteomics technology; 3) differential protein analysis and enrichment analysis based on the plasma proteomics results, and further verification of the candidate proteins with the most significant differences in a large cohort, and application of machine learning to evaluate the predictive effect of the candidate proteins.

[0035] 1. Sample Collection and Processing

[0036] (I) Patient samples were collected from the Department of Spine Surgery at Nanjing Gulou Hospital. After ethical review and informed consent, 5–6 ml of fresh whole blood was collected from hospitalized patients in anticoagulant tubes. Approximately 2 ml of plasma was obtained by centrifugation and stored in a -80°C ultra-low temperature freezer. Inclusion criteria were: 10 patients with grade II and grade V intervertebral disc degeneration, based on the Pfirrmann MRI grading system, were selected to form 10 case-control pairs.

[0037] 2. Proteomic Detection

[0038] (2) Plasma protein identification using SOMAscan 7k assay high-throughput proteomics technology, the technical features of which are:

[0039] ① SOMAmer probe modification: nucleic acid aptamers (SOMAmers) are labeled with fluorescent groups and biotin to capture target proteins through specific binding;

[0040] ② Target separation and purification: The SOMAmer-protein complex is released by ultraviolet cleavage of the photosensitive linker, and then solid-phase extraction is performed using streptavidin magnetic beads;

[0041] ③ Quantitative analysis: After denaturation and dissociation, SOMAmers were hybridized to a custom microarray, and the expression levels of 7289 proteins were quantified by fluorescence signal intensity;

[0042] ④ Differential protein screening: Using |Fold Change|>1.2 and p-value<0.05 as the threshold, 213 stable differentially expressed proteins were identified (109 upregulated and 104 downregulated).

[0043] 3. Bioinformatics Validation and Model Construction

[0044] (3) Data analysis process includes:

[0045] ①Quality control: intraclass correlation coefficient>0.95, principal component analysis showed significant separation between groups;

[0046] ② Functional enrichment analysis: The differentially expressed proteins were significantly enriched in immune regulation in the GO / KEGG database (p = 3.2 × 10 -5 ) and extracellular matrix remodeling pathway (p = 1.8 × 10 -4 );

[0047] ③ Target validation: ELISA was used to verify the significant differences in expression of six proteins, including COL6α3, REG1β, ATF5, CAP1, MAGEA4, and LILRB3, in the expanded cohort (p < 0.05);

[0048] ④ Machine learning modeling: Binary logistic regression was used to construct a multi-protein joint prediction model. Among them, the 6-protein joint model (AUC = 0.870) had the best prediction efficiency. Some specific performances are listed in Table 1:

[0049] Table 1 Some AUC values

[0050] Protein Combination AUC value 6 Protein Combination 0.8700 MAGEA4 single indicator 0.8243 ATF5 single indicator 0.8672

[0051] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0052] Intervertebral Disc Degeneration (IDD) is defined as the process of structural and functional damage to the intervertebral disc caused by aging or other factors (such as mechanical stress, genetic predisposition, etc.).

[0053] Biomarker: refers to any biological characteristic that can be objectively measured and evaluated, which can reflect normal biological processes, pathological states or responses to therapeutic interventions.

[0054] Magnetic resonance imaging (MRI) is a technique that uses strong magnetic fields and radio waves to non-invasively image the internal structures of the human body.

[0055] High-throughput proteomics: A technique that uses large-scale analysis to identify and quantify all proteins in cells, tissues, or body fluids.

[0056] Differentially Expressed Proteins: A class of proteins whose expression levels change significantly under different conditions (such as between disease and healthy states).

[0057] SOMAscan technology: A protein detection platform based on a lentiviral vector library that can quantify thousands of proteins simultaneously.

[0058] ROC curve (Receiver Operating Characteristic Curve): A visualization tool used to show the relationship between the true positive rate and the false positive rate in a binary classification problem.

[0059] AUC value (Area Under the Curve): The area under the ROC curve is an indicator used to quantify the model's ability to distinguish.

[0060] Example 1

[0061] Proteomics research

[0062] (I) Patient samples were collected from the Department of Spine Surgery at Nanjing Gulou Hospital. Following strict ethical review procedures and informed consent from patients and their families, 5-6 ml of fresh whole blood was collected from hospitalized patients in EDTA tubes. Approximately 2 ml of plasma was obtained by centrifugation (3000 rpm, 10 min, 4°C). The plasma was immediately aliquoted and stored in a -80°C ultra-low temperature freezer to ensure sample bioactivity. Sample selection was based on the internationally recognized Pfirrmann grading system. Two experienced spine surgeons independently evaluated MRI images using a double-blind method. In cases of discrepancies, a third expert arbitrated. Ultimately, representative cases of intervertebral disc degeneration progressing to grade II (early degeneration) and grade V (end-stage degeneration) were selected for comparative study.

[0063] like Figure 2 As shown, according to the Pfirrmann classification, MRI images of patients with grade II degeneration demonstrate mildly heterogeneous changes in the nucleus pulposus structure, with characteristic gray horizontal bands visible. However, the boundary between the nucleus pulposus and the annulus fibrosus remains clear, the nucleus pulposus signal intensity is high, and the intervertebral disc height remains normal. In contrast, patients with grade V degeneration exhibit severe nucleus pulposus structural disorganization, significantly reduced signal intensity, and complete loss of the nucleus pulposus-annulus boundary, accompanied by irreversible structural changes such as intervertebral disc collapse. All grade II cases included in this study met the criteria for partial nucleus pulposus fibrosis with normal intervertebral disc height, while grade V cases strictly met the terminal characteristics of loss of nucleus pulposus signal and intervertebral disc collapse. Furthermore, histopathological H&E staining revealed that compared with grade II degeneration tissue, grade V degeneration specimens exhibited reduced staining, significantly lighter eosin staining, and abundant inflammatory cell infiltration due to severe loss of extracellular matrix. These histological features further validated the reliability of the sample grading.

[0064] Through standardized sample collection procedures and strict grading criteria, we have established a representative sample library for intervertebral disc degeneration research. Using double-blind physician evaluation combined with pathological verification, we effectively ensure the objectivity and accuracy of sample grouping, laying a solid foundation for subsequent investigations into the molecular mechanisms underlying different stages of intervertebral disc degeneration. A multi-dimensional verification system based on MRI imaging and histopathology significantly enhances the scientific nature and reliability of our research findings.

[0065] (II) SomaScan technology is a high-throughput proteomics assay that uses modified nucleic acid aptamers (SOMAmers) to rapidly scan proteins. Its core process is as follows: First, the SOMAmer reagent is modified with a fluorescent group and biotin, specifically binding to the target protein in the sample; unbound proteins are washed away, and bound proteins are labeled with biotin. Subsequently, ultraviolet light cleaves the photocleavable linker, releasing the SOMAmer-protein complex, which is then captured by streptavidin affinity. Finally, after protein denaturation, the SOMAmer is hybridized to a microarray, and protein expression levels are quantified by fluorescence intensity. Based on the high-throughput SomaScan proteomics platform, we performed a systematic proteomic analysis of plasma samples from 20 patients with intervertebral disc degeneration (10 grade II and 10 grade V). A total of 7,289 quantifiable proteins were detected. By setting strict differential expression screening criteria (|Fold Change|>1.2 and p-value<0.05), 213 statistically significant and stably differentially expressed proteins were successfully identified. Further analysis showed that these differentially expressed proteins included 109 significantly upregulated proteins and 104 significantly downregulated proteins, showing obvious differentiation characteristics in expression patterns.

[0066] (III) We conducted an overall evaluation of the plasma proteomics data, which showed good intra-group reproducibility and significant inter-group differences.

[0067] like Figure 3 As shown in Figure 2, this study used proteomics to comprehensively analyze plasma samples from 20 patients with intervertebral disc degeneration (10 each of grade II and grade V). Using highly sensitive detection techniques, a total of 7,289 proteins were identified. After rigorous statistical screening (|Fold Change|>1.2 and p-value<0.05), 213 significantly differentially expressed proteins were identified, including 109 upregulated and 104 downregulated proteins. Principal component analysis (PCA) results showed that grade II and grade V samples formed independent clusters within the dimensionality reduction space. Within-group samples showed high consistency (small intra-group differences), while the two groups showed a clear trend of separation (significant inter-group differences). This finding was further validated by a volcano plot, which clearly displays the distribution of the 213 differentially expressed proteins (109 upregulated and 104 downregulated) that met the screening criteria. To more intuitively visualize the expression patterns of the differentially expressed proteins, a heat map analysis was constructed. Heat map results showed that the differentially expressed proteins screened exhibited stable expression patterns in both groups: upregulated proteins were consistently highly expressed in the stage V group, while downregulated proteins showed the opposite expression trend. These data collectively confirm that significant differences in protein expression profiles exist at different stages of intervertebral disc degeneration.

[0068] like Figure 4As shown in the figure, after quality control was met, we subsequently performed differential protein enrichment analysis. In the GO and KEGG annotations, we found that the differentially expressed proteins were significantly enriched in pathways such as human immunity and extracellular matrix assembly. This is closely related to the progression of intervertebral disc degeneration and is consistent with academic reports, further validating the credibility of our omics results. While focusing on the enriched pathways, we also noticed a cluster of differentially expressed proteins with a high fold change among the stably upregulated proteins as the patient's intervertebral disc degeneration progressed. These proteins include human type VI collagen α3 (COL6α3), human regenerating islet-derived protein 1β (REG1β), human transcription activator 5 (ATF5), human adenylate cyclase-associated protein 1 (CAP1), human melanoma antigen family A4 (MAGEA4), and human leukocyte immunoglobulin-like receptor subfamily B member 3 (LILRB3).

[0069] Functional enrichment analysis of differentially expressed proteins revealed significant immune regulation and extracellular matrix remodeling during disc degeneration. GO functional annotation analysis (Gene Ontology) revealed that upregulated proteins were primarily enriched in pathways such as immune system regulation and immune response activation. This finding was further validated in the Kyoto Encyclopedia of Genes and Genomes (Kyoto Encyclopedia of Genes and Genomes / REACTOME databases, with significant enrichment in complement activation and innate immune response pathways, suggesting that late-stage disc degeneration is accompanied by altered immune microenvironments. Meanwhile, downregulated proteins were significantly enriched in structural pathways such as cell-matrix adhesion and extracellular matrix organization. REACTOME database analysis also confirmed that these proteins are closely associated with extracellular matrix assembly, suggesting progressive disruption of disc tissue structure. Furthermore, all differentially expressed proteins showed some degree of enrichment in the platelet degranulation pathway. These results systematically reveal the synergistic mechanism of immune activation and matrix degradation in the process of intervertebral disc degeneration, which not only deepens the understanding of the occurrence and development of the disease, but more importantly provides new molecular targets and theoretical basis for the development of targeted treatment strategies, especially for intervention measures at different stages of degeneration, which have important guiding value.

[0070] Example 2

[0071] Clinical validation

[0072] 1. Based on the plasma proteomic analysis results of Example 1, we focused on screening and verifying 6 protein molecules with significant differential expression, including type VI collagen α3 chain (COL6α3) involved in the composition of the extracellular matrix, regenerative islet-derived protein 1β (REG1β) related to tissue repair, transcription activator 5 (ATF5) that regulates stress response, adenylate cyclase-associated protein 1 (CAP1) involved in cytoskeleton regulation, tumor-associated antigen melanoma antigen family A4 (MAGEA4), and immune regulatory molecule leukocyte immunoglobulin-like receptor subfamily B member 3 (LILRB3).

[0073] like Figure 5 As shown, in the plasma proteomics data, these proteins showed statistical significance between the two groups by independent sample t-test analysis, among which ATF5 (p=0.0367), CAP1 (p=0.0404) and LILRB3 (p=0.0382) reached significant levels (*), while COL6α3 (p=0.0053), MAGEA4 (p=0.0021) and REG1β (p=0.0444) showed higher significance (**). Notably, these differentially expressed proteins are not only involved in extracellular matrix remodeling but also in multiple key biological processes, including stress response regulation and immune microenvironment modulation. COL6α3, a key component of the extracellular matrix, showed particularly significant changes in expression (p = 0.0053), potentially directly related to the structural destruction of intervertebral disc degeneration. Furthermore, differential expression of immune-related proteins such as MAGEA4 (p = 0.0021) and LILRB3 (p = 0.0382) further confirms the crucial role of immune regulation in intervertebral disc degeneration. These findings not only validate the reliability of the previous proteomic screening results but, more importantly, provide specific biomarkers and potential therapeutic targets for a deeper understanding of the molecular mechanisms of intervertebral disc degeneration. The discovery of highly significantly differentially expressed proteins, such as COL6α3 and MAGEA4, in particular, lays a crucial foundation for the subsequent development of molecular diagnostic methods and targeted therapeutic strategies for intervertebral disc degeneration.

[0074] 2. We subsequently performed ELISA validation on these proteins in plasma samples from a larger cohort of patients. The validation cohort consisted of 50 patients in grade II and 50 patients in grade V, a total of 100 patients. It was verified that these 6 proteins had stable and significant differences in the two grades. Further, we used machine learning combined with ROC curves to evaluate the effectiveness of these 6 proteins in predicting the progression of intervertebral disc degeneration. At the same time, we constructed different combinations of the 6 proteins through binary logistic regression for joint analysis and found a method with higher prediction accuracy. Specifically: in the single protein prediction model, MAGEA4 (AUC = 0.8243), ATF5 (AUC = 0.8672), REG1β (AUC = 0.7512), CAP1 (AUC = 0.7288), COL6α3 (AUC = 0.7840), LILRB3 (AUC = 0.7120), in the joint analysis of protein combinations, REG1β + COL6α3 + ATF5 + CAP1 + MAGEA4 + LILRB3 The combined analysis combination of 6 proteins (AUC=0.8700), the combined analysis combination of 5 proteins REG1β+COL6α3+ATF5+MAGEA4+LILRB3 (AUC=0.7989), the combined analysis combination of 5 proteins REG1β+COL6α3+CAP1+MAGEA4+LILRB3 (AUC=0.7700), the combined analysis combination of 4 proteins REG1β+COL6α3+MAGEA4+LILRB3 (AUC=0.7011), and the combined analysis combination of 3 proteins REG1β+COL6α3+MAGEA4 (AUC=0.6978). According to the prediction results, the combined prediction of six proteins, including human type VI collagen α3 (COL6α3), human regenerating islet-derived protein 1β (REG1β), human transcription activator 5 (ATF5), human adenylate cyclase-associated protein 1 (CAP1), human melanoma antigen family A4 (MAGEA4), and human leukocyte immunoglobulin-like receptor subfamily B member 3 (LILRB3), has the highest credibility.

[0075] The information of 6 protein marker ELISA-verified monoclonal antibodies is as follows:

[0076] COL6α3:sc-515335, Santa Cruz Biotechnology (Santa Cruz, CA, USA);

[0077] MAGEA4: mAb#82491, Cell Signaling Technology (Danvers, MA, USA);

[0078] ATF5:MA5-38080, Invitrogen (Carlsbad, CA, USA);

[0079] REG1β:MA5-29856, Invitrogen (Carlsbad, CA, USA);

[0080] LILRB3:MA5-29759, Invitrogen (Carlsbad, CA, USA);

[0081] CAP1:MA5-53318, Invitrogen (Carlsbad, CA, USA).

[0082] 3. If Figure 6 As shown in the figure, in independent validation experiments, we used ELISA to quantitatively detect six key candidate proteins. All target molecules showed significantly differential expression patterns in plasma samples from different stages of intervertebral disc degeneration (grade II vs. grade V). Specific test results showed that the expression level of the stress regulator ATF5 in the grade V group (2.495±0.8256ng / ml) was extremely significantly increased compared to the grade II group (1.502±0.5074ng / ml) (****); the expression level of the cytoskeleton regulatory protein CAP1 increased from 3.236±1.229ng / ml in grade II to 5.754±4.114ng / ml in grade V (**). Of particular note, the concentration of the core extracellular matrix component COL6α3 in the grade V group (23.73±4.591 ng / ml) was significantly higher than that in the grade II group (19.40±2.800 ng / ml) (***). The immune regulatory molecule LILRB3 (11.25±0.8256 vs 10.00±1.621 ng / ml) and the tumor-associated antigen MAGEA4 (6.236±5.889 vs 2.825±0.7773 ng / ml) also showed significant (**) and extremely significant (****) upregulation trends, respectively. Furthermore, the concentration of the tissue repair-related protein REG1β was also significantly increased in the grade V group (90.06±31.34 ng / ml) compared to the grade II group (69.59±12.01 ng / ml) (**). These results not only validated the reliability of previous proteomic findings using highly specific immunoassays, but more importantly, revealed the dynamic changes in these molecules during intervertebral disc degeneration, providing an experimental basis for establishing a degeneration staging diagnostic system based on peripheral blood biomarkers. Among them, MAGEA4 exhibited the largest fold change (2.2-fold) and the stable differential expression of COL6α3, making it particularly promising for development as a clinical diagnostic biomarker.

[0083] 4. If Figure 7As shown, we systematically evaluated the predictive efficacy of candidate proteins for the development of intervertebral disc degeneration using receiver operating characteristic (ROC) curve analysis. The core principle of ROC curve analysis is to calculate the sensitivity and specificity of biomarkers at different thresholds and quantify their diagnostic value using the area under the curve (AUC). AUC values ​​closer to 1 indicate higher predictive accuracy. In the single-protein prediction model, each marker demonstrated good predictive ability, with ATF5 performing the best (AUC = 0.8672), followed by MAGEA4 (AUC = 0.8243) and COL6α3 (AUC = 0.7840). REG1β (AUC = 0.7512), CAP1 (AUC = 0.7288), and LILRB3 (AUC = 0.7120) also exhibited above-average predictive value. Based on the principle of "improving diagnostic efficacy while balancing clinical feasibility," we constructed a multi-protein joint prediction model using binary logistic regression. During the model construction process, we comprehensively considered three key factors: 1) statistically significant differences in the validation cohort; 2) detectability of plasma protein concentrations; and 3) convenience of clinical application. The final 6-protein joint model was established.

[0084] The 5-protein combination (REG1β+COL6α3+ATF5+CAP1+MAGEA4+LILRB3) demonstrated the best predictive performance (AUC = 0.8700), while the streamlined 5-protein combination (REG1β+COL6α3+ATF5+MAGEA4+LILRB3, AUC = 0.7989) maintained high predictive accuracy while being more clinically practical. Notably, as the number of included proteins decreased, the model's predictive performance showed a gradient decline, with significant decreases in the predictive power of the 4-protein combination (AUC = 0.7011) and the 3-protein combination (AUC = 0.6978).

[0085] At the same time, we combined the protein concentration values ​​provided by the ELISA validation results of a large sample of patients with the optimal cutoff value in the ROC prediction results to provide a diagnostic standard:

[0086] COL6α3: ≥22.73 mg / ml;

[0087] MAGEA4: ≥3.9 mg / ml;

[0088] REG1β: ≥87.5 mg / ml;

[0089] ATF5: ≥1.67 mg / ml;

[0090] LILRB3: ≥10.12 mg / ml;

[0091] CAP1: ≥3.838mg / ml.

[0092] These results not only confirmed the diagnostic value of a single biomarker, but more importantly, significantly improved the prediction accuracy through multi-indicator combined analysis, providing an important basis for the development of a clinical diagnostic model for intervertebral disc degeneration.

[0093] The above are only a few preferred embodiments of the present invention, and their description is relatively specific and detailed, but it should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art may make various modifications and improvements without departing from the scope of the present invention, and such modifications and improvements are within the scope of protection of the present invention.

Claims

1. A multi-protein marker combination for predicting and assessing the severity of intervertebral disc degeneration, characterized in that: The marker combination includes any three or more of the following six plasma proteins: type VI collagen α3 chain (COL6α3), regenerating islet-derived protein 1β (REG1β), transcription activator 5 (ATF5), adenylate cyclase-associated protein 1 (CAP1), melanoma antigen family A4 (MAGEA4), and leukocyte immunoglobulin-like receptor subfamily B member 3 (LILRB3); The marker combination is used to predict the progression of intervertebral disc degeneration or assess its severity by quantitatively detecting the expression levels of the proteins in plasma; The severity was divided into grades II to V based on the Pfirrmann grading system.

2. The multi-protein marker combination for prediction and severity assessment of intervertebral disc degeneration according to claim 1, characterized in that: The marker combination consists of COL6α3, REG1β, ATF5, CAP1, MAGEA4 and LILRB3.

3. The multi-protein marker combination according to claim 1, characterized in that The expression level of the marker combination was verified by enzyme-linked immunosorbent assay (ELISA), wherein the ELISA used monoclonal antibodies to quantitatively detect COL6α3, REG1β, ATF5, CAP1, MAGEA4, and LILRB3.

4. The multi-protein marker combination according to claim 1, characterized in that The predictive efficacy of the marker combination was evaluated by the receiver operating characteristic (ROC) curve, in which the area under the curve (AUC) of the combined detection of the six proteins was ≥0.

870.

5. Use of the multi-protein marker combination according to any one of claims 1 to 4 in the preparation of a reagent for predicting and assessing the severity of intervertebral disc degeneration.

6. Use of the multi-protein marker combination according to any one of claims 1 to 4 in the preparation of a reagent complementary to the Pfirrmann grading results of MRI imaging for early screening or dynamic monitoring of intervertebral disc degeneration.

7. A system for predicting intervertebral disc degeneration based on the multi-protein marker combination according to any one of claims 1 to 4, characterized in that: The system comprises: (a) a plasma sample collection device for obtaining anticoagulated whole blood from a subject and separating plasma; (b) a protein detection module, used to quantitatively analyze the expression levels of the six proteins in the marker combination; (c) Data processing module, which builds a prediction model based on machine learning algorithm and outputs the intervertebral disc degeneration grading results.

8. The system according to claim 7, characterized in that The machine learning algorithm is a binary logistic regression algorithm, the input of which is the protein concentration value, and the output is the probability of intervertebral disc degeneration and the grading result.

9. The system according to claim 8, characterized in that The construction of the binary logistic regression model includes the following steps: (i) Determine the optimal cutoff value of a single protein using the ROC curve and screen for proteins with significant differences; (ii) The multi-protein combination was optimized using stepwise regression and the AUC value of the joint prediction was calculated; (iii) Model parameters were calibrated based on the ELISA data of the validation cohort to improve prediction accuracy.

10. The system according to claim 9, characterized in that The data processing module further includes a visualization interface for displaying ROC curves, AUC values, and protein concentration thresholds, wherein the cutoff values ​​for COL6α3 are ≥22.73 mg / ml, MAGEA4 is ≥3.9 mg / ml, REG1β is ≥87.5 mg / ml, ATF5 is ≥1.67 mg / ml, LILRB3 is ≥10.12 mg / ml, and CAP1 is ≥3.838 mg / ml.

Images