Prediction marker and prediction kit for curative effect of neoadjuvant therapy of esophageal squamous carcinoma based on proteomics

Abstract

The invention belongs to the technical field of biology, and particularly relates to an esophageal squamous cell carcinoma neoadjuvant therapy curative effect prediction marker based on proteomics and a prediction kit. The invention discloses a marker for predicting the curative effect of neoadjuvant therapy of esophageal squamous carcinoma based on proteomics. The marker comprises STAT2, NQO1, CBR1, HLAB, PSMB9 and AKR1B10. By analyzing a protein expression profile in a biopsy sample of a patient before treatment and combining actual treatment effect information of the patient, differential protein remarkably related to the treatment effect is screened out, and a scientific basis is provided for clinically formulating an individualized treatment scheme.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology, specifically relating to proteomics-based biomarkers and reagent kits for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma. Background Technology

[0002] Esophageal cancer is one of the most common malignant tumors of the digestive system, ranking sixth in mortality worldwide. More than half of all cases occur in China, and over 90% of the subtypes are esophageal squamous cell carcinoma (ESCC). Due to its insidious early symptoms and rapid progression, most patients are diagnosed at an advanced stage, making surgical resection a limited treatment option. To improve patient survival, a comprehensive treatment approach combining neoadjuvant chemoradiotherapy (nCRT) with surgical resection is widely adopted in clinical practice. Numerous studies have shown that neoadjuvant therapy can reduce tumor size, lower pathological stage, increase R0 resection rate, and improve long-term survival in some patients.

[0003] However, the efficacy of neoadjuvant therapy varies significantly among different patients. Some patients achieve pathological complete remission (pCR) with a good prognosis; while others are insensitive to treatment, not only failing to benefit but also experiencing a worse prognosis due to delayed surgery, accumulated toxic side effects, or drug resistance. Accurately predicting a patient's response to neoadjuvant therapy before treatment has become a pressing issue in clinical treatment decision-making.

[0004] Current research on efficacy prediction mainly focuses on imaging indicators, pathological features, gene mutations, and transcriptomics. For example, metabolic parameter analysis based on radiomics or PET-CT can indicate efficacy trends to some extent, but its accuracy is insufficient. While genomic or transcriptomic studies have revealed some molecular events related to efficacy, the transcriptional level is regulated by multiple factors and cannot fully reflect changes in functional state. In contrast, proteomics can directly reflect the terminal effects of cellular physiological activities and is an important means of understanding tumor biology and drug response mechanisms. With the rapid development of mass spectrometry and bioinformatics, tissue-based proteomics analysis has become an effective way to discover disease biomarkers.

[0005] For esophageal squamous cell carcinoma, current research on proteomics in predicting the efficacy of neoadjuvant therapy remains relatively limited. Some reports focus only on a few signaling pathways or single candidate proteins, lacking systematic screening and validation. Furthermore, most studies rely on post-treatment surgical specimens rather than pre-treatment biopsy tissues, limiting the applicability of predictive models in clinical decision-making.

[0006] Therefore, there is an urgent need to develop a proteomics analysis method based on pre-treatment biopsy samples from patients. This method, combined with known information on the efficacy of neoadjuvant therapy, can systematically screen and identify differentially expressed proteins closely related to treatment efficacy, and establish a combination of biomarkers that can be used for preclinical prediction. This approach can not only enable early assessment of treatment response and patient stratification management, but also provide a solid foundation for subsequent molecular mechanism research and the development of precision treatment strategies. Summary of the Invention

[0007] To address the current limitations of existing technologies in achieving systematic analysis and biomarker screening for accurate prediction of neoadjuvant therapy efficacy based on pre-treatment tissue samples from patients, this invention establishes a proteomics-based method for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma and its biomarkers. By analyzing the protein expression profiles in pre-treatment biopsy samples from patients and combining them with actual efficacy information, differentially expressed proteins that are significantly correlated with efficacy are screened out, providing a scientific basis for developing individualized treatment plans in clinical practice.

[0008] The technical solution of the present invention is as follows: In one aspect, a proteomics-based biomarker for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma is provided, the biomarker including STAT2, NQO1, CBR1, HLA_B, PSMB9, and AKR1B10.

[0009] Secondly, a prediction method for the aforementioned biomarkers for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma is provided, characterized by the following steps: In the Tumor region, proteins with high connectivity in the PPI network are subjected to stepwise regression analysis using the stepAIC function in the MASS package of R language, with the efficacy of neoadjuvant therapy as the dependent variable and the protein expression level as the independent variable, and the model with the smallest AIC value is selected as the optimal model.

[0010] As a specific embodiment of the present invention, the method for screening proteins with high connectivity in the PPI network includes: S1 protein extraction and enzymatic hydrolysis; S2 DIA mass spectrometry detection; S3 DIA Data Analysis: Import the mass spectrometry data from step S2 into Spectronau software for protein identification and quantification. S4 Low-quality sample removal: Qualitative assessment is performed by plotting violin plots of protein identification counts for Tumor and Stroma sample points, and quantitative screening is performed using CV-Mean scatter plots to remove outlier samples. S5 differentially expressed protein screening: using FDR < 0.05 and Fold Change > 1.5 as screening criteria, differentially expressed proteins associated with the efficacy of neoadjuvant therapy were identified; S6 Protein-protein interaction network analysis: PPI analysis was performed on the differentially expressed proteins screened in step S5, with a confidence threshold set to medium; core proteins with significantly higher connectivity than the average were identified in each network as potential key functional proteins. S7 stepwise regression analysis, with neoadjuvant therapy efficacy as the dependent variable and protein expression level as the independent variable, selected the model with the smallest AIC value as the optimal model; The scoring formula derived from the model constructed by S8 is: Score = -7.877524 + 1.616224 × STAT2 - 0.358442 × NQO1 - 0.608091 × CBR1 - 1.161029 × HLA_B + 2.187545 × PSMB9 - 0.186361 × AKR1B10; when the predicted score Score > 0, the therapeutic effect is considered good, and when the score < 0, the therapeutic effect is considered poor.

[0011] Thirdly, another proteomics-based biomarker for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma is provided, which includes CAV1, TNC, PDLIM3, PDLIM7, CORO1A, ACTN1, ACTA2, CALD1, FBLN1, CALML5, and FLNC.

[0012] Fourthly, a method for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma is provided, which includes the following steps: In the two types of Stroma regions, proteins with high connectivity in the PPI network are subjected to stepwise regression analysis using the stepAIC function in the MASS package of R language, with the efficacy of neoadjuvant therapy as the dependent variable and the protein expression level as the independent variable, and the model with the smallest AIC value is selected as the optimal model.

[0013] As a specific embodiment of the present invention, the method for screening proteins with high connectivity in the PPI network includes: L1 protein extraction and enzymatic digestion; L2 DIA mass spectrometry detection; L3 DIA Data Analysis: Import the mass spectrometry data from step S2 into Spectronau software for protein identification and quantification. L4 Low-quality sample removal: Qualitative assessment was performed by plotting violin plots of protein identification counts for Tumor and Stroma sample points, and quantitative screening was performed using CV-Mean scatter plots to remove outlier samples. L5 differentially expressed protein screening: using FDR < 0.05 and Fold Change > 1.5 as screening criteria, differentially expressed proteins associated with the efficacy of neoadjuvant therapy were identified; L6 Protein-protein interaction network analysis: PPI analysis was performed on the differentially expressed proteins screened in step S5, with a confidence threshold set to medium; core proteins with significantly higher connectivity than the average were identified in each network as potential key functional proteins. L7 stepwise regression analysis was performed, with neoadjuvant therapy efficacy as the dependent variable and protein expression level as the independent variable. The model with the smallest AIC value was selected as the optimal model. The scoring formula derived from the L8 model is: Score = 14.416440 - 0.984800 × CAV1 - 0.638632 × TNC - 1.193191 × PDLIM3 + 2.703675 × PDLIM7 + 1.199359 × CORO1A - 3.182702 × ACTN1 - 1.448841 × ACTA2 + 2.945543 × CALD1 - 1.486126 × FBLN1 + 0.327034 × CALML5 + 0.837418 × FLNC; Judgment criteria: Score > 0 predicts good efficacy, Score < 0 predicts poor efficacy.

[0014] Fifthly, a proteomics-based kit for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma is provided, wherein the biomarkers include the aforementioned biomarkers. These biomarkers include STAT2, NQO1, CBR1, HLA-B, PSMB9, and AKR1B10.

[0015] Sixthly, the method of using the above-mentioned predictive kit is provided. The scoring formula is Score=-7.877524+1.616224×STAT2-0.358442×NQO1-0.608091×CBR1-1.161029×HLA_B+2.187545×PSMB9-0.186361×AKR1B10. When the predictive score Score > 0, the efficacy is considered good, and when the score < 0, the efficacy is considered poor.

[0016] Seventhly, another proteomics-based neoadjuvant therapy efficacy prediction kit for esophageal squamous cell carcinoma is provided, characterized in that the biomarkers are as described above. The biomarkers include CAV1, TNC, PDLIM3, PDLIM7, CORO1A, ACTN1, ACTA2, CALD1, FBLN1, CALML5, and FLNC.

[0017] Eighthly, the method of using the above-mentioned predictive kit is provided. The scoring formula used is Score=14.416440-0.984800×CAV1-0.638632×TNC-1.193191×PDLIM3+ 2.703675×PDLIM7+1.199359×CORO1A-3.182702×ACTN1-1.448841×ACTA2+2.945543×CALD1-1.486126×FBLN1+0.327034×CALML5+0.837418×FLNC. When the predictive score Score > 0, the efficacy is considered good, and when the score < 0, the efficacy is considered poor.

[0018] Beneficial effects of the present invention This invention establishes a systematic method for predicting the efficacy of neoadjuvant therapy by performing proteomic analysis on biopsy samples from esophageal squamous cell carcinoma (ESCC) patients before neoadjuvant therapy, combined with actual patient treatment information. Compared with existing technologies, this invention can predict efficacy before treatment, avoiding ineffective or overtreatment and significantly improving the accuracy of clinical decision-making. By systematically screening and validating protein biomarkers closely related to efficacy, the accuracy and stability of the prediction results are improved. This method can be implemented based on routine pathological samples, possessing good operability and clinical translation potential. Furthermore, the key proteins identified in this invention provide an important molecular basis for in-depth analysis of the sensitivity or resistance mechanisms of esophageal squamous cell carcinoma to radiotherapy and chemotherapy, which is of great significance for promoting personalized precision medicine. Attached Figure Description

[0019] Figure 1 This is a distribution diagram of the total number of peptides and proteins in the QC sample in Example 1, used to demonstrate the stability of the mass spectrometry detection system; Figure 2 The Spearman correlation coefficient heatmap between QC samples in Example 1 shows the repeatability between different batches of tests; Figure 3 The violin plots of protein identification counts for Tumor and Stroma sample points in Example 1 are used to qualitatively screen out low-quality samples with insufficient protein extraction. Figure 4 The CV-Mean scatter plot for protein quantification in Example 1 is used to quantitatively screen out low-quality samples with insufficient protein extraction. Figure 5 The graph shows the expression of protein quantification data from the screened samples in Example 1, indicating that the sample data conforms to a normal distribution and has no obvious bias. Figure 6This is a clustering distribution map of the samples after dimensionality reduction using the UMAP algorithm in Example 1, showing the clustering trend of the same patient samples in the dimensionality reduction space; Figure 7 This is a clustering distribution map of the samples after dimensionality reduction using the UMAP algorithm in Example 1, showing the separation trend between Tumor and Stroma samples; Figure 8 This is a UMAP clustering result diagram of the Good and Poor groups within the Tumor region in Example 1; Figure 9 This is a UMAP clustering result diagram of the Good and Poor groups within the Stroma region in Example 1; Figure 10 This is a volcano map of DEPs in the Tumor region of Example 1, showing the upregulated and downregulated proteins associated with the efficacy of neoadjuvant therapy; Figure 11 This is a volcano map of DEPs in the Stroma region of Example 1, showing the upregulated and downregulated proteins associated with the efficacy of neoadjuvant therapy; Figure 12 This is a graph showing the enrichment results of the KEGG signaling pathway in the Good group of the Tumor region in Example 1; Figure 13 This is a graph showing the KEGG signaling pathway enrichment results of the Poor group upregulated proteins in the Tumor region in Example 1; Figure 14 This is a graph showing the enrichment results of the KEGG signaling pathway for the upregulated proteins in the Good group of the Stroma region in Example 1; Figure 15 This is a graph showing the enrichment results of the KEGG signaling pathway for the Poor group upregulated proteins in the Stroma region in Example 1; Figure 16 This is a PPI diagram of the upregulated proteins in the Good group of the Tumor region in Example 1; Figure 17 This is a PPI diagram of the proteins upregulated in the Poor group of the Tumor region in Example 1; Figure 18 This is a PPI diagram of the Good group upregulated proteins in the Stroma region in Example 1; Figure 19 This is a PPI diagram of the Poor group upregulated proteins in the Stroma region in Example 1; Figure 20 The efficacy prediction model constructed based on the six characteristic proteins obtained by Tumor region screening in Example 1 is shown with the ROC curve of each protein. Figure 21The image shows the efficacy prediction model constructed based on the 11 characteristic proteins obtained from Stroma region screening in Example 1, along with the ROC curve for each protein. Detailed Implementation

[0020] Example 1 I. Clinical Sample Collection and Sample Site Acquisition This study included 35 patients with esophageal squamous cell carcinoma (ESCC) from Zhejiang Cancer Hospital. All patients had primary ESCC and underwent preoperative biopsy before receiving neoadjuvant therapy. Biopsy samples were collected before neoadjuvant therapy.

[0021] Tissue samples were fixed in 10% neutral buffered formalin (10% NBF) and then sequentially dehydrated with ethanol, cleared with xylene, infiltrated with paraffin, and embedded to prepare formalin-fixed paraffin-embedded (FFPE) tissue blocks. The tissue blocks were then serially sectioned, each 3 μm thick, and stained with hematoxylin and eosin (HE).

[0022] The locations of the tumor parenchyma and tumor stroma were determined and marked under a microscope by a pathologist. Samples were then excised from these two regions using a laser microdissection system (Leica LMD7, Leica Microsystems), with each sample point having an area of ​​approximately 0.05 mm². 2 Four samples were collected from each of the two regions for each patient. All samples were immediately transferred to low-adsorption tubes for storage after cutting, and were subsequently used for protein extraction.

[0023] II. Data Acquisition via Label-Free Quantitative Proteomics 1. Protein extraction and enzymatic hydrolysis Due to the extremely small amount of tissue obtained from laser cutting, this embodiment employs a micro-sample-appropriate lysis enzymatic digestion system. The sample spots were placed in a lysis buffer containing ammonium bicarbonate (NH4HCO3) and subjected to cross-linking lysis in a metal bath. After lysis, trypsin was added at a 1:50 enzyme-to-protein ratio for enzymatic digestion, and the reaction was incubated in a water bath at 37°C. After the reaction was completed, formic acid (FA) at a final concentration of 10% was added to terminate the reaction.

[0024] After freeze-drying, the samples were reconstituted using mobile phase A (0.1% FA aqueous solution) and iRT calibration peptides to ensure consistent protein concentration. Both the ammonium bicarbonate lysis buffer and the FA termination system are mass spectrometry compatible and can be directly used for mass spectrometry detection without affecting electrospray ionization efficiency or signal response.

[0025] 2. DIA mass spectrometry detection The peptide solution after enzymatic hydrolysis was subjected to gradient separation using a nanoElute nano-level liquid chromatography system, and then ionized by an ESI electrospray ionization source before being scanned in DIA (Data Independent Acquisition) mode on a Tims TOF Pro2 tandem mass spectrometer (Bruker).

[0026] To ensure data stability and reproducibility, wash samples were inserted between sample tests from different tissue regions of each patient to clean the tubing, and QC samples were interspersed throughout daily testing to monitor system performance and signal drift. Raw data generated after mass spectrometry acquisition was saved as .d files for subsequent bioinformatics analysis.

[0027] III. Data Preprocessing 1. DIA Data Analysis Raw mass spectrometry data were imported into Spectronau software for protein identification and quantification. Retention time correction was performed using iRT peptides, and reliable identification results were screened using the software's built-in FDR control (False Discovery Rate < 0.05). The output included the number of identified proteins, protein types, and relative quantitative intensity values.

[0028] 2. Quality control and repeatability assessment The total number of peptides and proteins in the QC samples was statistically analyzed, and a distribution map of detection stability was plotted. Figure 1 Further calculate the Spearman correlation coefficient between QC samples (). Figure 2 The results were all greater than 0.94, indicating that the mass spectrometry system has good repeatability and stability.

[0029] 3. Removal of low-quality samples Considering the minute errors during tissue cutting and extraction, some samples may have insufficient protein extraction. A violin plot of protein identification counts for Tumor and Strom samples was plotted. Figure 3 Qualitative evaluation was performed, and CV-Mean scatter plots were used. Figure 4 Quantitative screening was performed. After removing outlier samples, the data distribution of the remaining samples conformed to a normal distribution ( ). Figure 5 The absence of significant offset indicates good overall data quality, suggesting that this batch of data can be used for subsequent analysis.

[0030] IV. Bioinformatics Analysis 1. Analysis of individual and organizational differences The protein quantification data of all samples were log2 transformed and normalized. Dimensionality reduction analysis was performed using the UMAP algorithm. The results showed that samples from the same patient significantly clustered in the reduced dimensionality space, indicating significant individual differences. Figure 6 ).

[0031] At the population level, the Tumor and Stroma samples also showed a clear trend of separation. Figure 7 After further removing proteins with significant individual variability, the protein expression distribution of different neoadjuvant therapy efficacy groups was compared. Results showed a clustering trend within the same efficacy group, but complete separation occurred between different efficacy groups. Figure 8 , Figure 9 ).

[0032] 2. Screening for differentially expressed proteins (DEPs) Using FDR < 0.05 and Fold Change > 1.5 as screening criteria, differentially expressed proteins associated with neoadjuvant therapy efficacy were identified. A total of 79 DEPs were found in the Tumor region, of which 48 were upregulated in the Good group and 31 were upregulated in the Poor group. Figure 10 A total of 92 DEPs were identified in the Stroma region, of which 40 were upregulated in the Good group and 52 were upregulated in the Poor group. Figure 11 ).

[0033] 3. KEGG signaling pathway enrichment analysis The DEPs were imported into the KEGG database for pathway enrichment analysis (screening criteria: p < 0.05). Results showed that the upregulated proteins in the Good group of the Tumor region were mainly enriched in immune response and inflammatory signaling pathways. Figure 12 The Poor group upregulated proteins were enriched in drug metabolism and oxidative stress-related pathways. Figure 13 ); while the Good group upregulated proteins in the Stroma region are involved in immune activation and cell migration pathways ( Figure 14 The upregulated proteins in the Poor group are mainly associated with matrix remodeling and immunosuppressive pathways. Figure 15 ) 4. Protein-protein interaction network (PPI) analysis PPI analysis was performed on the four groups of DEPs using the STRING online database (https: / / string-db.org / ), with a medium confidence threshold (interaction score > 0.4). The analysis results are presented in the form of a node network. Figure 16 – Figure 19Core proteins with significantly higher connectivity degrees than the average were identified in each network and identified as potential key functional proteins, providing a candidate feature set for subsequent modeling.

[0034] 5. Protein Feature Screening and Therapeutic Effect Classifier Construction In the Tumor and Stroma regions, proteins with high connectivity in the PPI network were analyzed using the stepwise logistic regression function in the MASS package of R language. The neoadjuvant therapy efficacy (Good vs. Poor) was used as the dependent variable and the protein expression level was used as the independent variable. The model with the smallest AIC value was selected as the optimal model.

[0035] Stepwise regression screening in the Tumor region yielded six optimal characteristic proteins (STAT2, NQO1, CBR1, HLA_B, PSMB9, AKR1B10). The scoring formula for the model constructed from these proteins is: Score = -7.877524 + 1.616224 × STAT2 - 0.358442 × NQO1 - 0.608091 × CBR1 - 1.161029 × HLA_B + 2.187545 × PSMB9 - 0.186361 × AKR1B10. A predicted score > 0 indicates good efficacy, while a score < 0 indicates poor efficacy. Based on ROC curve analysis, the model's AUC value is 0.90, demonstrating excellent discriminative performance. Figure 20 ).

[0036] Stepwise regression screening in the Stroma region yielded 11 optimal characteristic proteins (CAV1, TNC, PDLIM3, PDLIM7, CORO1A, ACTN1, ACTA2, CALD1, FBLN1, CALML5, FLNC). The scoring formula derived from the model constructed using these proteins is: Score = 14.416440 - 0.984800 × CAV1 - 0.638632 × TNC - 1.193191 × PDLIM3 + 2.703675 × PDLIM7 + 1.199359 × CORO1A - 3.182702 × ACTN1 - 1.448841 × ACTA2 + 2.945543 × CALD1 - 1.486126 × FBLN1 + 0.327034 × CALML5 + 0.837418 × FLNC. The judgment criterion is the same as above: Score > 0. The model predicts good therapeutic efficacy, while a score < 0 indicates poor predictive efficacy. The model's ROC curve AUC is 0.89, demonstrating high sensitivity and specificity. Figure 21 ).

Claims

1. A proteomics-based biomarker for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma, characterized in that, The markers include STAT2, NQO1, CBR1, HLA_B, PSMB9, and AKR1B10.

2. The prediction method using the predictive biomarkers for neoadjuvant therapy efficacy in esophageal squamous cell carcinoma as described in claim 1, characterized in that, Includes the following steps: In the Tumor region, proteins with high connectivity in the PPI network were subjected to stepwise regression analysis using the stepAIC function in the MASS package of R language. The efficacy of neoadjuvant therapy was used as the dependent variable and the protein expression level was used as the independent variable. The model with the smallest AIC value was selected as the optimal model.

3. The prediction method according to claim 2, characterized in that, The screening method for proteins with high connectivity in the PPI network includes: S1 protein extraction and enzymatic hydrolysis; S2 DIA mass spectrometry detection; S3 DIA Data Analysis: Import the mass spectrometry data from step S2 into Spectronau software for protein identification and quantification. S4 Low-quality sample removal: Qualitative assessment is performed by plotting a violin plot of protein identification counts for Tumor sample points, and quantitative screening is performed using CV-Mean scatter plots to remove outlier samples. S5 differentially expressed protein screening: using FDR < 0.05 and Fold Change > 1.5 as screening criteria, differentially expressed proteins associated with the efficacy of neoadjuvant therapy were identified; S6 Protein-protein interaction network analysis: PPI analysis was performed on the differentially expressed proteins screened in step S5, with a confidence threshold set to medium; core proteins with significantly higher connectivity than the average were identified in each network as potential key functional proteins. S7 stepwise regression analysis, with neoadjuvant therapy efficacy as the dependent variable and protein expression level as the independent variable, selected the model with the smallest AIC value as the optimal model; The scoring formula derived from the model constructed by S8 is: Score = -7.877524 + 1.616224 × STAT2 - 0.358442 × NQO1 - 0.608091 × CBR1 - 1.161029 × HLA_B + 2.187545 × PSMB9 - 0.186361 × AKR1B10; when the predicted score Score > 0, the therapeutic effect is considered good, and when the score < 0, the therapeutic effect is considered poor.

4. A proteomics-based biomarker for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma, characterized in that, The markers include CAV1, TNC, PDLIM3, PDLIM7, CORO1A, ACTN1, ACTA2, CALD1, FBLN1, CALML5, and FLNC.

5. The method for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma using the biomarkers described in claim 1, characterized in that, The steps include: In the Stroma region, the proteins with high connectivity in the PPI network are subjected to stepwise regression analysis using the stepAIC function in the MASS package of R language, with the efficacy of neoadjuvant therapy as the dependent variable and the protein expression level as the independent variable, and the model with the smallest AIC value is selected as the optimal model.

6. The prediction method according to claim 5, characterized in that, The screening method for proteins with high connectivity in the PPI network includes: L1 protein extraction and enzymatic digestion; L2 DIA mass spectrometry detection; L3 DIA Data Analysis: Import the mass spectrometry data from step S2 into Spectronau software for protein identification and quantification. L4 Low-quality sample removal: Qualitative assessment was performed by plotting a violin plot of protein identification counts for Stroma sample points, and quantitative screening was performed using CV-Mean scatter plots to remove outlier samples. L5 differentially expressed protein screening: using FDR < 0.05 and Fold Change > 1.5 as screening criteria, differentially expressed proteins associated with the efficacy of neoadjuvant therapy were identified; L6 Protein-protein interaction network analysis: PPI analysis was performed on the differentially expressed proteins screened in step S5, with a confidence threshold set to medium; core proteins with significantly higher connectivity than the average were identified in each network as potential key functional proteins. L7 stepwise regression analysis was performed, with neoadjuvant therapy efficacy as the dependent variable and protein expression level as the independent variable. The model with the smallest AIC value was selected as the optimal model. The scoring formula derived from the L8 model is: Score = 14.416440 - 0.984800 × CAV1 - 0.638632 × TNC - 1.193191 × PDLIM3 + 2.703675 × PDLIM7 + 1.199359 × CORO1A - 3.182702 × ACTN1 - 1.448841 × ACTA2 + 2.945543 × CALD1 - 1.486126 × FBLN1 + 0.327034 × CALML5 + 0.837418 × FLNC; Judgment criteria: Score > 0 predicts good efficacy, Score < 0 predicts poor efficacy.

7. A proteomics-based kit for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma, characterized in that, The marker includes the marker as described in claim 1.

8. The method of using the prediction kit according to claim 7, characterized in that, The scoring formula used is Score = -7.877524 + 1.616224 × STAT2 - 0.358442 × NQO1 - 0.608091 × CBR1 - 1.161029 × HLA_B + 2.187545 × PSMB9 - 0.186361 × AKR1B10. When the predicted score Score > 0, the therapeutic effect is considered good, and when the score < 0, the therapeutic effect is considered poor.

9. A proteomics-based kit for predicting the efficacy of neoadjuvant therapy for esophageal squamous cell carcinoma, characterized in that, The marker includes the marker as described in claim 4.

10. The method of using the prediction kit according to claim 9, characterized in that, The scoring formula used is Score = 14.416440 - 0.984800 × CAV1 - 0.638632 × TNC - 1.193191 × PDLIM3 + 2.703675 × PDLIM7 + 1.199359 × CORO1A - 3.182702 × ACTN1 - 1.448841 × ACTA2 + 2.945543 × CALD1 - 1.486126 × FBLN1 + 0.327034 × CALML5 + 0.837418 × FLNC. When the predicted score Score > 0, the therapeutic effect is considered good; when the score < 0, the therapeutic effect is considered poor.

Images