Dedifferentiated liposarcoma patient prognosis and immunotherapy curative effect prediction model and construction method thereof

Abstract

The invention discloses a prognosis and immunotherapy curative effect prediction model for a dedifferentiated liposarcoma patient and a construction method of the prognosis and immunotherapy curative effect prediction model. The construction method comprises the following steps: acquiring transcriptome and genome data of the DDLPS patient from a TCGA database; calculating GSP through ssGSEA, and performing survival analysis in combination with OS and DFS of the patient; the method comprises the following steps: screening genes related to GSP and patient survival states through WGCNA, constructing DDLPS patient prognosis and immunotherapy curative effect prediction by using LASSO regression analysis and a random forest algorithm, dividing DDLPS patients into a high-risk group and a low-risk group according to an RFRS optimal truncation value, and drawing an ROC survival curve to predict prognosis and immunotherapy curative effects. The prediction model provided by the invention not only can accurately predict the OS of a DDLPS patient, but also has immunotherapy curative effect prediction performance, and provides important reference for individualized application of DDLPS immunotherapy.

Description

Technical Field

[0001] The present invention belongs to the field of biomedicine technology, and specifically relates to a prognosis and immunotherapy efficacy prediction model for dedifferentiated liposarcoma patients and a construction method thereof. Background Art

[0002] Dedifferentiated liposarcoma (DDLPS) is a highly malignant soft tissue sarcoma with a poor prognosis and low sensitivity to traditional treatment. Although immunotherapy has shown significant efficacy in various tumor types in recent years, its application in DDLPS is still in the initial exploratory stage. At present, the characteristics of the immune microenvironment of DDLPS and its relationship with prognosis have not been fully elucidated, especially in the regulation of pyroptosis-related mechanisms and their potential impact on the efficacy of immunotherapy. There are still many unknowns. In addition, the prediction models constructed in the existing technology mainly rely on R language and public database data, and have not been verified according to population characteristics. It is unclear whether they are actually applicable. Therefore, it is of great clinical significance to construct a model that can accurately predict the prognosis of DDLPS patients and the efficacy of immunotherapy. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the main purpose of the present invention is to provide a method for constructing a prediction model for the prognosis and immunotherapy efficacy of DDLPS patients, and to construct a prediction model based on cell pyroptosis-related genes through transcriptomics for the prediction of the prognosis and immunotherapy efficacy of DDLPS patients.

[0004] Another object of the present invention is to provide a DDLPS patient prognosis and immunotherapy efficacy prediction model, which comprises one or a combination of two or more gene markers selected from B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] The present invention provides an application of a reagent for detecting the expression of a gene marker in the preparation of a product for the prognosis of a DDLPS patient and the efficacy of immunotherapy, wherein the gene marker is selected from one or a combination of two or more of B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

[0007] Preferably, the product is a kit.

[0008] The present invention also provides a kit for detecting the prognosis of DDLPS patients and the efficacy of immunotherapy, which comprises reagents for detecting the expression of gene markers and instructions for use, wherein the gene markers are selected from one or a combination of two or more of B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

[0009] The present invention also provides a method for constructing a DDLPS patient prognosis and immunotherapy efficacy prediction model, comprising the following steps:

[0010] Step 1: Obtain transcriptome and genome data of DDLPS patients from the TCGA database;

[0011] Step 2: Calculate the enrichment score (GSP) of pyroptosis-related genes by single-sample gene set enrichment analysis (ssGSEA), and perform survival analysis in combination with the overall survival (OS) and disease-free survival (DFS) of DDLPS patients;

[0012] Step 3. Use weighted gene co-expression network analysis (WGCNA) to screen genes related to the survival status of GSP and DDLPS patients. First, perform LASSO regression analysis based on the expression data of the above genes and the survival data of DDLPS patients to obtain a variable expression matrix with a regression coefficient that is not 0; then perform random forest modeling based on the survival data of DDLPS patients to obtain the random forest risk score (RFRS) of the sample, that is, the prediction model for the prognosis and immunotherapy efficacy of DDLPS patients. The optimal cutoff value of RFRS is 7.9341. According to the optimal cutoff value of RFRS, DDLPS patients are divided into high-risk and low-risk groups, and ROC survival curves are drawn. The prognosis and immunotherapy efficacy are evaluated based on the ROC survival curves.

[0013] Preferably, in step three, the DDLPS patients whose RFRS is greater than 7.9341 are high-risk patients with poor prognosis and immunotherapy efficacy; the DDLPS patients whose RFRS is not greater than 7.9341 are low-risk patients with good prognosis and immunotherapy efficacy.

[0014] Preferably, the method further includes verifying the predictive performance of the DDLPS patient prognosis and immunotherapy efficacy prediction model through immunotherapy efficacy verification cohorts of three cancer types, namely melanoma, urothelial carcinoma and renal cell carcinoma, drawing the ROC curve of RFRS and PD-L1 expression level, and evaluating its accuracy in predicting immunotherapy efficacy.

[0015] More preferably, the immunotherapy efficacy validation cohorts are the Melanoma-phs000452 dataset for melanoma, the MUC-GSE176307 dataset for metastatic urothelial carcinoma (MUC), and the RCC-Braun_2020 dataset for renal cell carcinoma (RCC).

[0016] Preferably, in step three, the gene is selected from one or a combination of two or more of B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

[0017] The present invention also provides a DDLPS patient prognosis and immunotherapy efficacy prediction model, which comprises one or a combination of two or more gene markers selected from B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2; and also comprises the RFRS of the DDLPS patient; when the RFRS of the DDLPS patient is greater than 7.9341, it is a high-risk population with poor prognosis and immunotherapy efficacy; the RFRS of the DDLPS patient is not greater than 7.9341, it is a low-risk population with good prognosis and immunotherapy efficacy.

[0018] The present invention constructs a prediction model based on cell pyroptosis-related genes through transcriptomics, and evaluates its application value in the prognosis of DDLPS patients and the prediction of immunotherapy efficacy. First, the transcriptome and genomic data of DDLPS patients were obtained from the TCGA database, the GSP was calculated by ssGSEA, and the survival analysis was performed in combination with the OS and DFS of DDLPS patients. Subsequently, the key genes related to GSP and the survival status of DDLPS patients were screened out by WGCNA, and the prognosis prediction model was constructed using LASSO regression and random forest algorithms. Finally, the predictive performance of the model was verified by immunotherapy cohorts of three different cancer types, melanoma, urothelial carcinoma and renal cell carcinoma, and the ROC curve of RFRS and PD-L1 expression level was drawn to evaluate its accuracy in the prediction of immunotherapy efficacy. Compared with the prior art, the beneficial effects of the present invention are:

[0019] 1. The present invention screened out genes related to cell pyroptosis enrichment score and survival status of DDLPS patients through WGCNA analysis, and further constructed a DDLPS patient prognosis prediction model through LASSO regression analysis and random forest modeling, which not only verified its ability to distinguish cell pyroptosis score groups, but also could accurately predict the OS of DDLPS patients, providing a basis for the prognostic stratification of DDLPS patients.

[0020] 2. The present invention further verifies the accuracy and reliability of the model through immune correlation analysis, and applies the model to other immunotherapy cohort data, confirming that the prediction model has the performance of predicting the efficacy of immunotherapy, providing an important reference for the individualized application of immunotherapy, and providing new ideas for the precision treatment of DDLPS. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The prognosis and immunotherapy efficacy prediction model for DDLPS patients in the embodiment divides DDLPS patients into high-risk and low-risk groups according to the optimal RFRS cutoff value for survival analysis results.

[0022] Figure 2 : The ROC curves of the DDLPS patient prognosis and immunotherapy efficacy prediction model in the embodiment predicting 1-year, 3-year, and 5-year OS.

[0023] Figure 3 This is a visual diagram of the importance ranking of the 9 genes in the DDLPS patient prognosis and immunotherapy efficacy prediction model in the embodiment.

[0024] Figure 4 The hazard ratio (HR), 95% confidence interval (CI) and P value analysis results of the 9 genes in the DDLPS patient prognosis and immunotherapy efficacy prediction model in the example in the univariate Cox regression analysis.

[0025] Figure 5 The correlation between the risk score and grouping and GSP and its score grouping for the prediction model of prognosis and immunotherapy efficacy of DDLPS patients in the embodiment; A: ROC curve drawn according to the patient GSP grouping and risk score; B: the overall level of GSP in different risk groups; C: correlation analysis between the risk score RFRS and GSP; D: distribution of the proportion of risk patients in different risk groups.

[0026] Figure 6 These are the analysis results of the three cohort data sets of Melanoma-phs000452, MUC-GSE176307, and RCC-Braun_2020 in the embodiment for validation of the efficacy of immunotherapy; A: differential analysis results; B, C, D: survival analysis results; E, F, G: ROC curves.

[0027] Figure 7 This is a nomogram for predicting the 1-year, 3-year, and 5-year OS probabilities of patients using the DDLPS patient prognosis and immunotherapy efficacy prediction model in the embodiment. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in combination with the embodiments and drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them, and all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Example 1

[0030] 1. Materials and methods

[0031] 1.1 Data Collection and Processing

[0032] On November 2, 2024, TCGA biolinks (version 2.28.4) of R language 4.3.1 was used to download DDLPS data from the TCGA (https: / / cancergenome.nih.gov) database, and 59 DDLPS patients with complete clinical information and expression spectrum matrix were selected, and TPM expression data and gene symbol tables corresponding to ENSID were obtained respectively. When there were multiple matches, the expression data took the median value, and genes that were not expressed (TPM = 0) in more than half of the samples were removed for subsequent transcriptomics-related analysis. The same package was used to download the Masked somatic mutation data in the SNV of the TCGA database for subsequent genomics-related analysis.

[0033] 1.2ssGSEA and survival analysis

[0034] 33 key genes of cell pyroptosis were obtained from published literature. Then, the GSP of 58 samples in the above TCGA-DDLPS database was calculated by ssGSEA method using the GSVA package (version 1.48.3) in R language.

[0035] Before survival analysis, patients with incomplete survival time or survival status were removed, resulting in a sample size of 59 and 57 for OS analysis and DFS analysis, respectively. Survival curves were plotted using the ggsurvplot function of the survminer package (version 0.4.9), where the optimal cutoff value was calculated using the surv_cutpoint function, and patients were divided into high and low GSP score groups based on this value, and univariate survival analysis was performed using the survival package (version 3.5-7). Then, the differentially expressed genes (DEGs) between the high and low score groups were screened using the limma package (version: 3.56.2), and the expression matrix of these DEGs and the clinical characteristics of the patients were extracted for subsequent WGCNA.

[0036] 1.3WGCNA analysis

[0037] After the above genes and their expression matrices that have an impact on both OS and DFS were extracted, they were incorporated into the WGCNA package (version 1.72-5) for analysis in combination with the patients’ clinical baseline characteristics and GSP. First, the hclust function was used to perform sample clustering analysis, and method = "average" was set; then, the pick Soft Threshold function was used to select the optimal power value for the construction of a scale-free network; in addition, the TOM similarity function was used to calculate TOM, and genes were clustered according to 1-TOM; then, the cutree Dynamic function was used to identify dynamic shear modules, deepSplit = 2, min Module Size = 5, and plotDendroAndColors function was used for plotting; subsequently, the moduleEigengenes function was used to observe the correlation between clinical characteristics and modules, and merge Close Modules was used to cluster similar modules, and the Hmisc package (version number: 5.1-1) was used to visualize the correlation between modules; next, the labeledHeatmap function was used to display the correlation between modules and clinical characteristics, and the MM and GS values ​​were calculated, and genes with negative correlation with OS.status and P < 0.05 were selected for subsequent modeling.

[0038] 1.4 LASSO regression analysis and random forest modeling

[0039] After sorting the expression data of the above genes and the patient survival data, the glmnet package (version 4.1-8) was used for LASSO regression analysis, where family was set to Cox, alpha was 1, and nfolds was 10. Then, the lambda value was selected according to the minimum position of the partial likelihood deviation to obtain variables with regression coefficients not equal to 0, which were included in the subsequent modeling. After sorting the above variable expression matrix and patient survival data, the random Forest SRC package (version 3.2.3) was used for random forest modeling, with ntree = 1000 and nodesize = 15. After the model was built, the predict.rfsrc function was used to calculate the RFRS of the sample. The univariate cox analysis results of the genes included in the model were sorted, and the forestploter package (version 1.1.1) was used to draw the forest map. According to the optimal cutoff value of RFRS, the patients were divided into high-risk and low-risk groups, and the survival curves were drawn.

[0040] Next, the discrimination and goodness of fit of the model were evaluated. First, the pec package (version 2023.04.12) was used to analyze the trend of the model C-statistic with survival time and draw a graph; the 1-year, 3-year, and 5-year ROC curves of the model were drawn using the timeROC package (version 0.4); the precision-recall curve (PRC) of the model was drawn using the modEvA package (version 3.13.3). In addition, the rms package (version 6.7-1) was used to draw the calibration curve and nomogram of the model.

[0041] 1.5 Correlation analysis between RFRS, GPS and OS

[0042] The GSP, survival status and overall survival of patients in the TCGA-DDLPS database were integrated with the RFRS and risk grouping of the model for correlation analysis to evaluate the correlation between the model and cell pyroptosis and further analyze the predictive value of the model for patients' OS.

[0043] 1.6 Immune correlation analysis

[0044] First, the stromal score, immune score, estimate score and Tumor Purity of TCGA-DDLPS samples were calculated using the estimate package (version 1.0.13). At the same time, the TCGA pan-cancer immune infiltration data were downloaded from the TIMER2.0 (http: / / timer.cistrome.org) database, and the data of DDLPS patients were extracted from it. According to the high- and low-risk groups of patients, the inter-group differences were analyzed, and representative immune cell types were selected for display, with special attention to the immune score and tumor microenvironment score.

[0045] Since DDLPS lacks immunotherapy cohort data for analysis, expression profile data and clinical prognosis information of three different cancer cohorts that have received immunotherapy were obtained, including Melanoma-phs000452, MUC_GSE176307, and RCC-Braun_2020. Among them, the Melanoma-phs000452 cohort had 153 melanoma patients treated with anti-PD-1 drugs; the MUC_GSE176307 cohort had 89 patients with metastatic urothelial carcinoma (MUC) with complete immune response information, treated with anti-PD-1 or anti-PD-L1 drugs; and RCC-Braun_2020 had 172 patients with renal cell carcinoma (RCC) treated with anti-PD-1 drugs. The model was built using the same method and divided into high-risk and low-risk groups. The differences in RFRS between the groups were analyzed based on whether the patients in the validation cohort responded to immunotherapy. Survival analysis was then performed in combination with the patients' prognostic information to further verify the prognostic prediction function of the model. Finally, the pROC package (version 1.18.5) was used to draw the ROC curve of PDL1 and RFRS in predicting the efficacy of immunotherapy.

[0046] 1.7 Statistical analysis

[0047] All data were statistically analyzed and plotted using R 4.3.1 and GraphPad Prism (version 9.0.0, San Diego, California USA). The default parameters were used for the parameters not mentioned in the method, and the ggplot2 software package (version 3.4.4) was used for the visualization of the data not mentioned. Continuous variables were displayed as mean ± standard deviation; for data that did not conform to the normal distribution, the Mann Whitney rank sum test was used to analyze the difference between the two groups; for data that conformed to the normal distribution and had homogeneous variance, the Student's t test was used to compare the two groups; for data that conformed to the normal distribution but did not conform to the homogeneous variance, the Welch's t test was used; and the Spearman test was used for correlation analysis. The chi-square test was used for inter-group comparison of discrete variables. The difference between the two ROC curves was analyzed using DeLong's test. P < 0.05 was considered to be statistically significant. Among them, * indicates P < 0.05, ** indicates P < 0.01, *** indicates P < 0.001, and **** indicates P < 0.0001.

[0048] 2. Results

[0049] The optimal cutoff value RFRS in the DDLPS patient prognosis and immunotherapy efficacy prediction model constructed in Example 1 was 7.9341. The DDLPS patients were divided into high-risk and low-risk groups for survival analysis. The results showed that the median OS of the high-risk group was 846 days, which was significantly lower than that of the low-risk group (the median time was not reached). The difference between the two groups was statistically significant (P<0.0001, Figure 1 In addition, ROC curve analysis showed that the area under the curve (AUC) of the model in predicting 1-year, 3-year, and 5-year OS was 0.8640, 0.8453, and 0.8599, respectively ( Figure 2 ). It can be seen that the RFRS of this model can well divide DDLPS patients into high-risk and low-risk groups, and has prognostic prediction value.

[0050] Compared with single sample enrichment analysis based on high-throughput sequencing, the nine genes B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2 are used to predict patient prognosis and immunotherapy efficacy, which is more cost-effective and has potential value for clinical promotion and use. The importance of these nine genes in the model is ranked as follows: Figure 3 The hazard ratio (HR), 95% confidence interval (CI) and P value of the above 9 genes in univariate Cox regression analysis are shown in Figure 4 shown.

[0051] In this example, a prognosis prediction model for DDLPS patients related to cell pyroptosis was developed, and the correlation between its risk score and grouping and GSP and its score grouping was tested. ROC curves were drawn according to the GSP grouping and risk score of the patients. The results showed that the model had excellent performance in distinguishing high and low GSP scores, with an AUC value of up to 0.8804 (95% CI: 0.7891-0.9718, P<0.0001, Figure 5 A). Further comparison of different risk groups revealed that the overall GSP level in the high-risk group was significantly lower than that in the low-risk group (P = 0.0003, Figure 5 B). In addition, there was a significant negative correlation between the RFRS and GSP of the model (R = -0.6119, P < 0.0001, Figure 5 C). It is worth noting that in the high GSP score group, the proportion of high-risk patients was relatively low (P = 0.0002, Figure 5 D). In summary, this model achieves a strong correlation with cell pyroptosis and can well replace GSP for risk grouping.

[0052] In view of the lack of immunotherapy cohort data for DDLPS, data sets from three different platforms were used as efficacy validation to verify the value of this prediction model in predicting the efficacy of immunotherapy. Patients in the three cohorts were all treated with anti-PD-1 drugs and had complete drug efficacy and patient prognosis information. The results of the differential analysis showed that the RFRS of patients in the responding group in the Melanoma-phs000452 cohort (P=0.0027) was significantly lower than that in the non-responding group. The same results were obtained in the MUC-GSE176307 cohort (P=0.0141) and the RCC-Braun_2020 cohort (P=0.0072). Figure 6 A). In addition, survival analysis showed that Melanoma-phs000452 (P<0.0001, Figure 6 B), MUC-GSE176307 (P<0.0001, Figure 6 C) and RCC-Braun_2020 (P<0.0001, Figure 6 D) The differences between the high-risk and low-risk groups in the three cohorts were statistically significant.

[0053] To further analyze the value of RFRS in predicting the efficacy of immunotherapy, we compared it with the commonly used clinical index PDL1. The ROC curve results showed that Melanoma-phs000452 ( Figure 6 E), MUC-GSE176307( Figure 6 F) and RCC-Braun_2020( Figure 6 G) In the three cohorts, the AUC values ​​of the ROC curves drawn with RFRS were all higher than those of PDL1, and the difference between the two ROC curves of the Melanoma-phs000452 cohort was statistically significant (P=0.03279). Therefore, the constructed prediction model has been preliminarily confirmed to have the performance of predicting the efficacy of immunotherapy, and the high-risk group is less likely to benefit from immunotherapy.

[0054] The DDLPS patient prognosis prediction model was constructed to predict the probability of 1-year, 3-year, and 5-year OS of the patient. Figure 7 As shown, according to the optimal cutoff value RFRS = 7.9341, DDLPS patients can be divided into high-risk and low-risk groups. This value can be used as the boundary value of DDLPS RNA sequencing data standardized by TPM to distinguish high-risk and low-risk groups. The high-risk group has poor sensitivity to immunotherapy, while the low-risk group should actively undergo immunotherapy to prolong patient survival.

[0055] The above is only a preferred embodiment of the present invention, but the present invention should not be limited to the contents disclosed in this embodiment. Any equivalent or modification completed without departing from the spirit disclosed in the present invention shall fall within the scope of protection of the present invention.

Claims

1. Use of a reagent for detecting gene marker expression in the preparation of a product for the prognosis of patients with dedifferentiated liposarcoma and the efficacy of immunotherapy, characterized in that: The gene marker is selected from one or a combination of two or more of B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

2. The use according to claim 1, characterized in that: The product is a test kit.

3. A kit for detecting the prognosis of patients with dedifferentiated liposarcoma and the efficacy of immunotherapy, characterized in that: The method comprises reagents for detecting the expression of gene markers and instructions for use, wherein the gene markers are selected from one or a combination of two or more of B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

4. A method for constructing a model for predicting the prognosis and immunotherapy efficacy of patients with dedifferentiated liposarcoma, characterized in that: The following steps are involved: Step 1: Obtain transcriptome and genomic data of dedifferentiated liposarcoma patients from the TCGA database; Step 2: Calculate the enrichment score of pyroptosis-related genes through single-sample gene set enrichment analysis, and perform survival analysis based on the overall survival and disease-free survival of patients with dedifferentiated liposarcoma; Step three, screen the genes related to the enrichment score of the cell pyroptosis-related genes and the survival status of the dedifferentiated liposarcoma patients through weighted gene co-expression network analysis; first perform LASSO regression analysis based on the expression data of the above genes and the patient survival data to obtain a variable expression matrix with a regression coefficient that is not 0; then perform random forest modeling in combination with the patient survival data to obtain the random forest risk score of the sample - RFRS, that is, the prognosis and immunotherapy efficacy prediction model for dedifferentiated liposarcoma patients, where: the optimal cutoff value of RFRS is 7.9341, and the dedifferentiated liposarcoma patients are divided into high and low risk groups according to the optimal cutoff value of RFRS, and the ROC survival curve is drawn, and the prognosis and immunotherapy efficacy are predicted according to the ROC survival curve.

5. The method for constructing a prognosis and immunotherapy efficacy prediction model for dedifferentiated liposarcoma patients according to claim 4, characterized in that: The patients with dedifferentiated liposarcoma whose RFRS is greater than 7.9341 are high-risk patients with poor prognosis and immunotherapy efficacy; the patients with dedifferentiated liposarcoma whose RFRS is not greater than 7.9341 are low-risk patients with good prognosis and immunotherapy efficacy.

6. The method for constructing a prognosis and immunotherapy efficacy prediction model for dedifferentiated liposarcoma patients according to claim 4, characterized in that: It also includes validating the prognosis and immunotherapy efficacy prediction model for patients with dedifferentiated liposarcoma through immunotherapy efficacy validation cohorts for three cancer types: melanoma, urothelial carcinoma, and renal cell carcinoma, drawing the ROC curve of RFRS and PD-L1 expression level, and evaluating the accuracy of its immunotherapy efficacy prediction.

7. The method for constructing a prognosis and immunotherapy efficacy prediction model for dedifferentiated liposarcoma patients according to claim 6, characterized in that: The immunotherapy efficacy validation cohorts are the Melanoma-phs000452 dataset for melanoma, the MUC-GSE176307 dataset for metastatic urothelial carcinoma, and the RCC-Braun_2020 dataset for renal cell carcinoma.

8. The method for constructing a prognosis and immunotherapy efficacy prediction model for dedifferentiated liposarcoma patients according to claim 4, characterized in that: In step three, the gene is selected from one or a combination of two or more of B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2.

9. A model for predicting the prognosis and immunotherapy efficacy of patients with dedifferentiated liposarcoma, characterized in that: It comprises one or a combination of two or more gene markers selected from B3GALT4, B3GNT8, BST2, CFI, HNMT, LGALS9, LRMDA, MSRA, and STXBP2; Also included is the RFRS of the patient with dedifferentiated liposarcoma; When the RFRS of the dedifferentiated liposarcoma patient is greater than 7.9341, it is a high-risk group with poor prognosis and immunotherapy efficacy; when the RFRS of the dedifferentiated liposarcoma patient is not greater than 7.9341, it is a low-risk group with good prognosis and immunotherapy efficacy.

Images