Biomarker for prognosis evaluation and immunotherapy effect prediction of urothelial carcinoma

Abstract

The invention discloses a combined predictive gene composition for evaluating the prognosis of urothelial carcinoma and predicting the immunotherapy effect. The combined predictive gene composition comprises 12 genes, namely NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1 and HSPA13. The invention discloses a comprehensive construction method and system of a combined predictive gene composition based on machine learning. The method comprises the following steps: analyzing depletion CD8 + T immune infiltration related genes based on a specific BLCA data set; carrying out cross analysis on the exhaustion CD8 + T immune infiltration related gene and an SLC16A1 expression related pathway gene to obtain M genes; through single-variable Cox analysis, identifying N prognosis genes from the M genes; a CD8 immune-related prognostic spectrum based on the SLC16A1 is determined based on a machine learning ensemble learning algorithm, and a combined predictive gene composition is determined based on the CD8 immune-related prognostic spectrum based on the SLC16A1. The invention discloses a combined predictive gene composition detection kit based on SLC16A1 related exhaustive CD8 + T cell polygene characteristics and application of the detection kit in predicting the lifetime of a patient with urothelial carcinoma and predicting the immunotherapy effect of the patient with urothelial carcinoma.

Description

Technical Field

[0001] This invention relates to the field of molecular diagnostics, and in particular to a biomarker for prognostic assessment and immunotherapy efficacy prediction of urothelial carcinoma, based on a multi-gene signature (SCPS) detection kit for SLC16A1-associated exhausted CD8+ T cells and its application. Background Technology

[0002] Bladder cancer (BLCA) refers to malignant tumors arising from the urothelial lining of the bladder. More than 90% are urothelial carcinomas, with the remainder being rare types such as squamous cell carcinoma and adenocarcinoma. BLCA causes approximately 150,000 deaths worldwide each year. The TNM staging system in the American Joint Committee on Cancer (AJCC) staging manual helps assess the prognosis of BLCA based on cystectomy specimens. However, increasing data shows that radical cystectomy does not necessarily improve survival and significantly reduces patients' quality of life. Furthermore, despite advances in traditional treatments such as surgical resection and adjuvant chemotherapy, the prognosis of BLCA is poor, with a high recurrence rate and a 5-year survival rate of less than 50%. Therefore, developing diagnostic kits with high sensitivity and specificity is of significant practical importance for more accurately predicting patient prognosis.

[0003] Immunotherapy has rapidly emerged as a transformative treatment approach in oncology, making significant strides in the management of human cancers. A key determinant of the effectiveness of immunotherapy is the immune composition of the tumor microenvironment (TME), which plays a central role in tumorigenesis and influences the response to immune checkpoint blockade (ICB) therapy. Evidence suggests that the TME is highly heterogeneous, with CD8+ cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells being the main immune components involved in eliminating cancer cells. The antitumor activity of both is influenced by the TME, which can either enhance or inhibit their function. For example, dendritic cells (key antigen-presenting cells) are recruited to the tumor site by intratumoral NK cells. M1 macrophages produce pro-inflammatory cytokines, indirectly enhancing the antitumor activity of CTLs and NK cells. On the other hand, certain tumor stromal cells within the TME, such as M2 tumor-associated macrophages, cancer-associated fibroblasts, and bone marrow-derived suppressor cells, possess immunosuppressive properties. These cells can suppress antitumor responses by releasing immunosuppressive cytokines (such as IL-10 and transforming growth factor-β (TGF-β)) or by directly inhibiting CTL activity through modulation of cell surface immune checkpoint molecules (such as PD-L1 and TIGIT). Therefore, a deeper understanding of the mechanisms driving immunosuppressive TMEs is crucial for better addressing immunotherapy resistance and developing more effective therapeutic strategies.

[0004] Activation of immune cells leads to significant metabolic reprogramming to support the energy and biosynthetic needs of rapid cellular growth and proliferation. A key aspect of this reprogramming involves glucose metabolism, which is converted to pyruvate via glycolysis and then transported to the mitochondria into the tricarboxylic acid (TCA) cycle. Despite the presence of oxygen, many cells, particularly tumor cells, prefer aerobic glycolysis, a phenomenon known as the "Warburg effect." This metabolic pathway results in the production of large amounts of lactate in the cytoplasm. To sustain glycolysis, the lactate produced in the cytoplasm must be exported to the extracellular microenvironment. The transport of lactate across the cell membrane is primarily facilitated by monocarboxylic acid transporters (MCTs). MCTs belong to the SLC16A gene family and consist of 14 members. MCTs, particularly MCT1-4, promote the passive transport of monocarboxylate salts (such as lactate, pyruvate, and ketone bodies) and protons across the cell membrane. These transporters play a crucial role in the metabolic reprogramming of tumor cells by transporting lactate, shuttling between glycolytic cancer cells, cancer-associated fibroblasts, and oxidative tumor cells, contributing to the survival and energy balance of tumor cells. Lactate signaling drives tumor angiogenesis, immunosuppression, multidrug resistance, migration, metastasis, and counteracts ferroptosis and autophagy—all of which contribute to tumor progression and evasion of cell death. A previous study showed that MCT1 expression in the cell membrane and cytoplasm accounted for approximately 40% of 124 BLCA cases and was associated with poor prognosis and metastasis. However, the exact relationship between SLC16A1 expression, its genetic signature, and its underlying molecular mechanisms remains unclear, requiring further investigation to fully understand its role in tumor biology and its potential as a therapeutic target. Summary of the Invention

[0005] The purpose of this invention is to provide a gene biomarker for prognostic assessment and immunotherapy efficacy prediction in urothelial carcinoma, a combined predictive gene composition detection kit based on SLC16A1-related exhaustive CD8+ T cell multi-gene signature (SCPS), and its application. This invention emphasizes the application of machine learning (ML) in multi-omics data analysis of urothelial carcinoma prognosis and immunotherapy. Furthermore, using ML algorithms, a prognostic feature based on SLC16A1 expression and exhaustive CD8+ T cell immune infiltration (SCPS) has been developed. This feature can predict patient prognosis and response to immunotherapy, providing value for clinical decision-making.

[0006] A first aspect of the present invention provides a combined predictive gene composition (SCPS) for assessing the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy, the combined predictive gene composition comprising 12 genes: NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

[0007] A second aspect of the present invention provides a method for comprehensively constructing a machine learning-based Joint Predictive Gene Composition (SCPS), comprising:

[0008] S1, based on a specific BLCA dataset, analyze genes related to exhaustion CD8+ T immune infiltration;

[0009] S2, cross-analysis was performed on the depleted CD8+T immune infiltration-related genes and the pathway genes related to SLC16A1 expression to finally obtain M genes;

[0010] S3, using univariate Cox analysis, N prognostic genes were identified from the M genes, p < 0.05, N <M;

[0011] S4. Determine the CD8 immune-related prognostic profile based on SLC16A1 using a machine learning ensemble learning algorithm, and determine a combined predictive gene composition based on the CD8 immune-related prognostic profile of SLC16A1.

[0012] Preferably, the specific BLCA dataset includes datasets GSE128702, GSE48276, and GSE87304.

[0013] Preferably, the method for obtaining the SLC16A1 expression-related pathway genes includes:

[0014] The effects of SLC16A1 tissue-specific expression on patient survival and immune regulation were further validated on multiple BLCA datasets.

[0015] Patients were divided into early T group and late T group according to T stage. The expression of SLC16A1 in the tissues of patients in the late group was higher than that in the early group, and the results of TCGA, GSE87304, GSE13507 and GSE48276 showed significant differences.

[0016] Based on SLC16A1 expression levels, patient samples were divided into high and low expression groups, and GSEA analysis was performed to identify pathway genes related to SLC16A1 expression.

[0017] Preferably, the GSE48276, GSE87304 and GSE128702 datasets are enriched with 11 pathway genes related to SLC16A1 expression, including cell cycle, focal adhesion and nod-like receptor signaling pathways.

[0018] Preferably, S4 includes:

[0019] S41, In the TCGA training queue dataset, 101 prediction models were built using 10 machine learning algorithms, and the C-index of each model in the training queue and validation dataset GSE13507 was calculated.

[0020] S42, Based on the comparison of the C-index, the optimal model corresponding to the highest C-index is determined to be the combination of Step-Cox[forward] + RSF;

[0021] S43, based on the optimal combination algorithm and Step-Cox [forward] proportional hazards regression, further RSF regression was performed to obtain a model composed of 12 genes including NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

[0022] A third aspect of the present invention is to provide a machine learning-based joint predictive gene composition (SCPS) synthesis system, comprising:

[0023] The immune infiltration-related gene analysis module (101) is used to analyze exhaustive CD8+T immune infiltration-related genes based on a specific BLCA dataset;

[0024] The gene cross-analysis module (102) is used to cross-analyze the depleted CD8+T immune infiltration-related genes and the pathway genes related to SLC16A1 expression, and finally obtain M genes.

[0025] The prognostic gene identification module (103) is used to identify N prognostic genes from the M genes using univariate Cox analysis, where p < 0.05, N <M;

[0026] The joint predictive gene composition determination module (104) is used to determine the CD8 immune-related prognostic profile based on SLC16A1 based on a machine learning ensemble learning algorithm, and to determine the joint predictive gene composition based on the CD8 immune-related prognostic profile of SLC16A1.

[0027] A fourth aspect of the present invention provides a combined predictive gene composition detection kit based on SLC16A1-associated exhausted CD8+ T cell polygenic signatures (SCPS) for assessing the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy. The kit is used to assess the prognosis of urothelial carcinoma and predict the efficacy of immunotherapy by means of the combined predictive gene composition (SCPS) of the first aspect and the detection of SLC16A1-associated exhausted CD8+ T cell polygenic signatures.

[0028] A fifth aspect of the present invention is to provide the application of the combined predictive gene composition detection kit in predicting the survival of patients with urothelial carcinoma.

[0029] A sixth aspect of the present invention is to provide the application of the combined predictive gene composition detection kit in predicting the efficacy of immunotherapy in patients with urothelial carcinoma.

[0030] The combined predictive gene composition, integrated construction method and system, kit, and application provided by this invention have the following beneficial technical effects:

[0031] 1. Clinical significance of SCPS as a combined predictive gene composition:

[0032] (1) Multidimensional assessment can be performed: SCPS integrates SLC16A1-related metabolic genes (such as lactate level and glycolytic activity) and exhaustion markers of CD8+ T cells (such as PD-1 and LAG3) to form a comprehensive score, which can accurately predict the sensitivity of BLCA patients to PD-1 / PD-L1 inhibitors.

[0033] (2) Treatment optimization strategy: Inhibitors targeting SLC16A1 (such as MCT1 blockers) can reduce TME lactate concentration, restore the metabolic function of exhausted CD8+ T cells, thereby improving SCPS score and enhancing the efficacy of immunotherapy.

[0034] 2. Through pan-cancer screening in public databases such as TCGA and GEO, 12 gene features related to the assessment of exhausted CD8+ T cells based on SLC16A1 gene expression were discovered. By comparing and analyzing with published detection models, it was demonstrated that the 12-gene joint detection provided by this invention has the highest specificity and sensitivity in assessing prognosis and immunotherapy efficacy, and this was also verified in real cases.

[0035] 3. Based on multi-cohort data of BLCA patients, a 12-gene prognostic feature (SCPS) associated with SLC16A1 expression in exhaustive CD8+ T cell immune infiltration was developed using machine learning methods, demonstrating potential diagnostic and prognostic value. SCPS showed significant advantages compared to existing clinical methods and published biomarkers for predicting BLCA prognosis.

[0036] 4. Immunotherapy evaluation showed that the high-risk group had a poorer response compared to the low-risk group. Therefore, SCPS can serve as a predictive biomarker for BLCA immunotherapy. Simultaneously, analysis of MCT1 protein expression in 86 formalin-fixed paraffin-embedded urothelial carcinoma specimens revealed that patients in the MCT1 membrane / plasma (MC) localization group had higher immune infiltration, while patients classified as low-risk (N group) had lower immune infiltration.

[0037] In summary, high-risk patients based on SCPS and MCT1MC localization are associated with immune escape and poor prognosis. The combination of MCT1-IHC analysis and SCPS may be a practical diagnostic approach for screening potential immunotherapies and predicting prognosis. Attached Figure Description

[0038] Figure 1 is a schematic diagram illustrating the effects of SLC16A1 on tissue expression, patient survival, and immune regulation in a pan-cancer study according to an embodiment of the present invention.

[0039] Figure 1A This diagram illustrates the relative expression differences of SLC16A1 in 32 types of cancerous normal and cancerous tissues according to embodiments of the present invention. Human figures and expression level heatmaps illustrate the differential expression of SLC16A1 between normal and cancerous tissues, as well as among various pathological subgroups. Color intensity on the human figure corresponds to expression level, with darker colors indicating higher expression levels. Tumor samples are shown in red, while normal tissue samples are shown in green. The color gradient of the heatmap from red to blue represents the relative expression level from high to low. *p<0.05, **p<0.01, ***p<0.005.

[0040] Figure 1B This is a schematic diagram illustrating the effect of SLC16A1 expression on overall survival (OS), DSS, DFI, and RFI according to an embodiment of the present invention; the forest plot shows the effect of SLC16A1 expression level on patient overall survival (OS).

[0041] Figure 1C This diagram illustrates the correlation between SLC16A1 expression and immune infiltration based on an algorithmic analysis according to an embodiment of the present invention. Using the TIMER algorithm, the correlation between SLC16A1 expression and immune scores, as well as immune cell infiltration, was analyzed in different cancer types. The circle colors represent correlation coefficients, with orange indicating a positive correlation and blue indicating a negative correlation. Statistically insignificant correlations were excluded from the figure.

[0042] Figure 1D This illustration shows the synergistic effect of SLC16A1 expression and immune infiltration on patient survival according to an embodiment of the present invention.

[0043] To analyze the synergistic effect of SLC16A1 expression and immune cell infiltration on overall survival (OS) in patients.

[0044] Figure 2 is a schematic diagram illustrating the effects of SLC16A1 on centralized tissue expression, patient survival, and immune regulation in BLCA datasets according to an embodiment of the present invention.

[0045] Figure 2AThis diagram illustrates further validation of the previous results on multiple BLCA datasets according to embodiments of the present invention; SLC16A1 expression in BLCA patient tissues varies across different T stages. If the number of T1 stage patient samples is less than three, they are merged with T2 stage patient samples and compared with late-stage samples (T3 and T4).

[0046] Figure 2B This diagram illustrates the GSEA analysis results for classifying patient samples into high and low expression groups based on SLC16A1 expression according to an embodiment of the present invention. GSEA analysis was performed on SLC16A1 samples from the high and low expression groups to identify pathways associated with their expression. The Venn diagram shows the intersection of three BLCA datasets, representing pathways associated with SLC16A1 expression in BLCA.

[0047] Figure 2C This is a schematic diagram illustrating the correlation between SLC16A1 expression and CD8+ T cell immune infiltration in patients according to an embodiment of the present invention; the dot plot shows the correlation between SLC16A1 expression level and CD8 immune cell infiltration. Red dots indicate statistically significant correlations (p<0.05 and r>0.2), black dots indicate significant correlations (p<0.05), and gray dots indicate no significant correlations (p>0.05).

[0048] Figure 2D This is a schematic diagram illustrating the significantly higher SLC16A1 expression level in patients with high levels of exhausted CD8+ T cell infiltration compared to patients with low levels of exhausted CD8+ T cells, according to an embodiment of the present invention. The violin plot shows the difference in exhausted CD8+ T cell immune infiltration between the high and low SLC16A1 expression groups. *p<0.05, **p<0.01, ***p<0.005, ns indicates no significant difference.

[0049] Figure 3 is a schematic diagram of a machine learning integration program developed based on SLC16A1 and depleted CD8+ T cell immune infiltration of SCPS according to an embodiment of the present invention.

[0050] Figure 3A This illustration shows the results of cross-analysis of CD8 immune infiltration-related genes in the BLCA datasets GSE128702, GSE48276, and GSE87304 according to an embodiment of the present invention, and the results of cross-analysis with SLC16A1-related pathway genes obtained in the previous step.

[0051] Figure 3B This is a schematic diagram showing the results of univariate Cox analysis identifying 96 prognostic genes according to an embodiment of the present invention; the heatmap shows the expression profiles of the top genes identified in the gene pool in the Venn diagram.

[0052] Figure 3C This is a schematic diagram illustrating the process of determining the optimal model according to an embodiment of the present invention; a total of 101 algorithm combination models were predicted using 10 different machine learning algorithms. The C-index, used to evaluate the model prediction accuracy, was calculated for each model in the validation dataset.

[0053] Figure 3D To illustrate the Kaplan-Meier curves and 1-, 3-, and 5-year ROC curves of the TCGA cohort based on SCPS according to an embodiment of the present invention, in the training set TCGA cohort, the survival rate of patients in the low-risk group was significantly higher than that of patients in the high-risk group (HR = 6.6, P < 0.001), and the AUC values ​​at 1 year, 3 years, and 5 years were 0.842, 0.876, and 0.866, respectively.

[0054] Figure 3E Kaplan-Meier curves and 1-, 3-, and 5-year ROC curves for the GSE13507 cohort were plotted based on SCPS according to an embodiment of the present invention. The results show that the survival rate of low-risk patients was significantly higher than that of high-risk patients.

[0055] Figure 3F The Kaplan-Meier curve and ROC curves for the GSE31684 cohort at 1, 3 and 5 years were plotted based on SCPS according to an embodiment of the present invention. The AUC values ​​for the GSE31684 cohort at 1 year, 3 years and 5 years are 0.735, 0.793 and 0.799, respectively.

[0056] Figure 3G The Kaplan-Meier curve and ROC curves for 1, 3 and 5 years of the CJUH cohort were plotted based on SCPS according to an embodiment of the present invention. The AUC values ​​for 1 year, 3 years and 5 years of the CJUH cohort were 0.912, 0.961 and 1.034, respectively.

[0057] Figure 4 is a schematic diagram comparing the predictive performance of SCPS with clinical indicators and reported prognostic features according to an embodiment of the present invention.

[0058] Figures 4A-4D This is a schematic diagram illustrating the results of comparing subgroup differences in RS based on clinical indicators according to an embodiment of the present invention, and evaluating the correlation between SCPS and clinical staging indicators and patient survival in different cohorts; wherein:

[0059] Figure 4AThe differences in hazard coefficients between the high-risk and low-risk groups, as well as the distribution of clinical indicators between the high-risk and low-risk groups, were analyzed in the TCGA cohort. If the number of T1 stage patients was less than three, they were pooled with T2 stage patients and compared with late-stage samples (T3 and T4). Statistical analysis was performed using the Wilcoxon test to determine whether the differences between groups were significant.

[0060] Figure 4B In the GSE13507 cohort, the differences in hazard coefficients between the high-risk and low-risk groups, as well as the distribution of clinical indicators between the high-risk and low-risk groups, were analyzed. If the number of T1 stage patients was less than three, they were pooled with T2 stage patients and compared with late-stage samples (T3 and T4). Statistical analysis was performed using the Wilcoxon test to determine whether the differences between groups were significant.

[0061] Figure 4C In the GSE31684 cohort, the differences in hazard coefficients between the high-risk and low-risk groups, as well as the distribution of clinical indicators between the high-risk and low-risk groups, were analyzed. If the number of T1 stage patients was less than three, they were pooled with T2 stage patients and compared with late-stage samples (T3 and T4). Statistical analysis was performed using the Wilcoxon test to determine whether the differences between groups were significant.

[0062] Figure 4D In the CJUH cohort, the differences in hazard coefficients between the high-risk and low-risk groups, as well as the distribution of clinical indicators between the high-risk and low-risk groups, were analyzed. If the number of T1 stage patients was less than three, they were pooled with T2 stage patients and compared with late-stage samples (T3 and T4). Statistical analysis was performed using the Wilcoxon test to determine whether the differences between groups were significant.

[0063] Figure 4E The diagram illustrates the results of univariate Cox regression analysis on each feature set of all datasets according to an embodiment of the present invention; the forest plot shows the results of univariate and multivariate Cox regression analysis for different cohorts.

[0064] Figure 4F-4G This is a schematic diagram illustrating the comparison of the C-index of SCPS with other characteristics according to an embodiment of the present invention; wherein:

[0065] Figure 4F Univariate Cox regression analysis was performed on different datasets to evaluate the performance of SCPS on prognostic features compared to 11 other reports. *p<0.05; **p<0.01;

[0066] Figure 4GFor different datasets, the C-index was calculated to evaluate the performance of SCPS on prognostic features compared to 11 other reports. *p<0.05; **p<0.01.

[0067] Figure 5 is a schematic diagram illustrating the relationship between SCPS and the patient's immune infiltration and immunotherapy response according to an embodiment of the present invention.

[0068] Figure 5A To assess the TME of patients in the TCGA and GSE13507 cohorts using the ESTIMATE algorithm according to embodiments of the present invention, a schematic diagram of the differences in immune function between the high-risk and low-risk groups was obtained using the SCPS risk assessment system; the violin plot shows the ESTIMATE, immune, matrix score and tumor purity between the high-risk and low-risk groups in the TCGA and GSE13507 cohorts.

[0069] Figure 5B A heatmap illustrating the correlation between immune assessment and immune cell infiltration according to an embodiment of the present invention is provided. The heatmap displays a correlation coefficient matrix between the infiltration abundance of 22 immune cell types and various scores, including immune scores, matrix scores, and ESTIMATE scores. In the heatmap, red indicates a positive correlation, meaning that when one variable increases, the other variable also increases, while blue indicates a negative correlation, meaning that an increase in one variable is related to a decrease in the other.

[0070] Figure 5C This diagram illustrates the positive correlation between SCPS gene expression and the infiltration abundance of exhausted CD8+ T cells, macrophages, Tregs, CAFs, fibroblasts, etc., according to embodiments of the present invention; the heatmap shows the correlation coefficient matrix between the infiltration abundance of 22 immune cells and SCPS expression level.

[0071] Figure 5D To compare the differences in tumor immune cell infiltration between the high-risk group and the low-risk group according to embodiments of the present invention, the infiltration of B memory cells, exhausted CD8+ T cells, macrophages, monocytes and neutrophils in the low-risk group was significantly lower than that in the high-risk group, as shown in the schematic diagram; the difference in the abundance of tumor immune cell infiltration between the high-risk group and the low-risk group.

[0072] Figure 5E This diagram illustrates the differences in immunosuppressive factor expression between high- and low-risk populations according to an embodiment of the present invention; it also shows the analysis of immunosuppressive gene expression levels in the high- and low-risk groups.

[0073] Figure 5FAccording to embodiments of the present invention, in the TCGA and GSE13507 cohorts, the difference in TIDE scores between the high-risk and low-risk groups was statistically significant, with the high-risk group having a higher TIDE score (p<0.01). The TIDE score is used to assess the likelihood of response to ICI treatment in high-risk and low-risk patients. (Note: The last part of the original text appears to be a separate, unrelated statement: ns, no statistical significance, *p<0.05, **p<0.01, ***p<0.005.)

[0074] Figure 6 is a schematic diagram illustrating how the subcellular localization of MCT1 in BLCA affects gene expression, patient survival, and immune infiltration according to an embodiment of the present invention.

[0075] Figure 6A According to embodiments of the present invention, MCT1 has different subcellular localization types, mainly including 54 nuclear staining (N) groups and 19 membrane / cytoplasm staining (MC) groups, with a schematic diagram of the results of 13 N / MC groups in between; H&E staining and IHC are used to determine the subcellular localization type of MCT1 in BLCA samples.

[0076] Figure 6B According to embodiments of the present invention, the primary diagnosis was stage II-IV in MC (89.5%) and N / MC (84.6%) patients (p<0.0001). SLC16A1 expression was also significantly elevated in N, N / MC and MC patients (p<0.05). Schematic diagram of results; differences in MCT1 expression levels among patients with three different subcellular localization types of MCT1.

[0077] Figure 6C The patient survival according to the embodiments of the present invention shows a gradually worsening trend, with median survival of 80.5 months, 43 months and 68 months in the N, MC and N / MC groups, respectively. The effect of MCT1 expression on the survival of patients with specific MCT1 subcellular localization types is illustrated in the diagram.

[0078] Figure 6D This is a schematic diagram of the sequencing results of transcriptomics and proteomics for patients with N and MC types according to an embodiment of the present invention; the PCA diagram shows the transcriptomic and proteomic classification of patients with N and MC types.

[0079] Figure 6E This diagram illustrates the good agreement between the sequencing results of transcriptomics and proteomics of patients with N and MC types according to an embodiment of the present invention; the circular heatmap shows the significant DEGs in the transcriptome and proteome of patients in the N and MC groups. Green indicates genes downregulated in MC patients, while orange indicates genes upregulated compared to N patients.

[0080] Figure 6FThis diagram illustrates the analysis results of differentially expressed genes in patients in groups N and MC according to an embodiment of the present invention; KEGG pathway enrichment analysis of DEGs.

[0081] Figure 6G The diagram illustrates how genes significantly associated with SLC16A1 expression levels were found in various cell death pathways, including VEGF signaling, glycolysis, apoptosis, autophagy, and ferroptosis, suggesting that SLC16A1 may be involved in influencing these pathways, according to embodiments of the present invention. Significantly enriched genes are positively correlated with SLC16A1 expression levels.

[0082] Figure 6H This diagram illustrates how patients in groups N and MC are categorized into risk classes according to a risk formula, based on an embodiment of the present invention. Patients in groups N and MC are categorized into risk classes according to the risk formula.

[0083] Figure 6I This is a schematic diagram showing the significant differences in the expression of the eight SCPS genes between the high-risk and low-risk groups according to an embodiment of the present invention; the heatmap shows the expression levels of SCPS genes in N-type and MC-type patient samples.

[0084] Figure 6J This is a schematic diagram showing that the fluorescence expression intensity of CD8, CD68, PD1, and PDL1 in the MC group patients was higher than that in the N group patients, according to an embodiment of the present invention; the mIHC image shows the expression of multiple immune markers in the samples of the N and MC groups patients.

[0085] Figure 6K This is a schematic diagram showing the downregulation of DNA mismatch repair genes in group N (NES = -1.52, p = 0.04, FDRq = 0.26) according to an embodiment of the present invention; downregulation of mismatch repair gene pathways and related genes in patients in group N.

[0086] Figure 6L This is a schematic diagram showing the upregulation of the expression of potential oncogenes MYC and E2F1 in patients in the MC group according to an embodiment of the present invention. The stereo diagram shows significant differences in expression among the groups, *p<0.05, **p<0.01, ***p<0.005.

[0087] Figure 7 This is a schematic diagram of the integrated construction method of machine learning-based joint predictive gene composition (SCPS) according to an embodiment of the present invention.

[0088] Figure 8 This is a schematic diagram of the principle structure of the integrated construction system of machine learning-based joint predictive gene composition (SCPS) according to an embodiment of the present invention.

[0089] Figure 9 This is a schematic diagram of the structure of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0090] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0091] Example 1

[0092] A combined predictive gene composition (SCPS) for assessing the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy, comprising 12 genes: NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

[0093] Example 2

[0094] A combined predictive gene composition assay kit based on SLC16A1-associated exhausted CD8+ T cell polygenic signature (SCPS) is provided for assessing the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy. The kit is used to assess the prognosis of urothelial carcinoma and predict the efficacy of immunotherapy by means of the combined predictive gene composition (SCPS) and the detection of SLC16A1-associated exhausted CD8+ T cell polygenic signature.

[0095] Example 3

[0096] Application of a combined predictive gene composition detection kit in predicting the survival of patients with urothelial carcinoma.

[0097] Example 4

[0098] Application of a combined predictive gene composition detection kit in predicting the efficacy of immunotherapy in patients with urothelial carcinoma.

[0099] Working principle:

[0100] Pan-oncogene analysis showed that SLC16A1 expression was positively correlated with the immune infiltration of exhausted CD8+ T cells in BLCA, and both factors affected patient survival.

[0101] Application Examples and Verification Examples

[0102] 1. The impact of SLC16A1 tissue-specific expression on patient survival and immune regulation

[0103] This invention analyzes the relative expression differences of SLC16A1 in normal and cancerous tissues of 32 types of cancer, such as Figure 1AAs shown, SLC16A1 expression was significantly upregulated in most cancer types. Subgroup analysis of patient expression differences based on different pathological stage indicators revealed significant differences in SLC16A1 expression in a few cancer types. These results indicate that SLC16A1 generally exhibits tumor specificity, but its expression varies significantly across different tumor types and pathological stages.

[0104] Simultaneously, the effects of SLC16A1 expression on survival OS, DSS, DFI, and RFI were analyzed, such as... Figure 1B As shown in the results, SLC16A1 expression, as a poor prognostic factor, affects the overall survival (OS), disease-specific survival (DSS), disease-free interval (DFI), and recurrence-free interval (RFI) in patients with PAAD (Pancreatic Adenocarcinoma), KIRP (Kidney Renal Papillary Cell Carcinoma), and ACC (Adrenocortical Carcinoma). It also affects LGG (Lower Grade G16A1). Patients with renal chromophobe carcinoma (low-grade glioma), kidney chromophobe carcinoma (KICHKidneyChromophobe), lung adenocarcinoma (LUAD), bladder urothelial carcinoma (BLCA), and sarcoma (SARC) have overall survival (OS), disease-free interval (DSE), and recurrence-free survival (RFI).

[0105] Subsequently, the correlation between SLC16A1 expression and immune infiltration was analyzed using TIMER (Tumor Immune Estimation Resource, a tool for analyzing tumor immune infiltration), MCP (Microenvironment Cell Populations, a multi-cell procedure (algorithm for immune infiltration analysis)), and quantum-seq (Quantitative Immunophenotyping by Sequencing, a sequencing-based quantitative immunophenotyping algorithm (algorithm for quantifying immune cells)). Figure 1CAs shown in the figure. The results showed that SLC16A1 expression was positively correlated with BLCA, GMLGG (Glioblastoma and Lower Grade Glioblastoma), and KIRP immune cell infiltration, and negatively correlated with SARC (Sarcoma), SKCM (Skin Cutaneous Melanoma), and TGCT (Testicular Germ Cell Tumor). Further analysis was conducted on BLCA and TGCT, which showed significant correlations and consistent results across different algorithms, to analyze the synergistic effect of SLC16A1 expression and immune infiltration on patient survival. The results are as follows. Figure 1D As shown, SLC16A1 expression is positively correlated with immune infiltration of exhausted CD8+ T cells in BLCA patients, both influencing patient survival. Immune cell infiltration in TGCT patients has no significant impact on survival.

[0106] in:

[0107] (1) Cancer type abbreviations (such as LGG, ACC) and prognostic indicators (such as OS, DSS) are common terms in oncology.

[0108] (2) TIMER and MCP are commonly used algorithms for immune infiltration analysis.

[0109] (3) GBMLGG is a combined term for glioblastoma and low-grade glioma, which is commonly used in brain tumor research.

[0110] 2. SLC16A1 expression in the BLCA dataset is associated with T stage, survival, and CD8+ immune infiltration.

[0111] The results of the previous step were further validated on multiple BLCA datasets. Patients were divided into early-stage and late-stage T groups based on T staging. It was found that SLC16A1 expression in the tissues of patients in the late-stage group was higher than that in the early-stage group, and significant differences were observed in the results from TCGA, GSE87304, GSE13507, and GSE48276 datasets. Figure 2A As shown.

[0112] Next, patient samples were divided into high and low expression groups based on SLC16A1 expression, and GSEA analysis was performed. The results showed that 11 pathways related to SLC16A1 expression were enriched in GSE48276, GSE87304, and GSE128702, such as... Figure 2B As shown, this includes the cell cycle, focal adhesion, and nod-like receptor signaling pathways.

[0113] Subsequently, the correlation between SLC16A1 expression and patient-acquired CD8+ T cell immune infiltration was analyzed. It was found that only one of the eight datasets showed no significant results, while six datasets showed a positive correlation between SLC16A1 expression and patient-acquired CD8+ T cell immune infiltration. Figure 2C As shown in the figure, in these six groups of data, the expression level of SLC16A1 was significantly higher in patients with high levels of exhausted CD8+ T cell infiltration than in patients with low levels of exhausted CD8+ T cells. Figure 2D As shown.

[0114] Furthermore, the expression levels of SLC16A1 in the TCGA and GSE13507 datasets, along with patient survival and clinical indicators, were analyzed. The results showed that SLC16A1 had a better prognosis in patients with early-stage disease (T1 / T2, N0, M0, AJCCI / II) and lower infiltration of exhausted CD8+ T cells. Its expression level had a significant impact on patient survival (HR>1, P<0.05, Table S3). These data strongly support the previous findings.

[0115] 3. Construction of a comprehensive method based on ML features

[0116] CD8 immune infiltration-related genes were analyzed in the BLCA datasets GSE128702, GSE48276, and GSE87304, and cross-analysis was performed with the SLC16A1-related pathway genes obtained in the previous step, ultimately yielding 163 genes, such as... Figure 3A As shown.

[0117] Univariate Cox regression analysis identified a total of 96 prognostic genes (p<0.05), such as... Figure 3B As shown.

[0118] Subsequently, a CD8 immune-related prognostic spectrum (SCPS, Step-Cox Prognostic Signature) based on SLC16A1 was developed using a machine learning ensemble program. In the TCGA training cohort dataset, we constructed 101 predictive models using 10 machine learning algorithms and calculated the C-index of each model in both the training cohort and the validation dataset GSE13507. Through this comparison, the optimal model was determined to be the combination of Step-Cox[forward] + RSF, which had the highest average C-index in the cohort at 0.815. Figure 3CAs shown. Therefore, by using the optimal combination algorithm and Step-Cox [forward] proportional hazards regression, further RSF regression was performed to obtain a model composed of 12 genes, namely NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

[0119] 4. SCPS Evaluation

[0120] The predictive efficacy of SCPS was evaluated using KM curve plotting and ROC analysis. A risk scoring formula was used to calculate individualized risk scores by multiplying the expression levels of the identified gene sets (12 SCPS genes) in each patient by regression coefficients. Patients were divided into high-risk and low-risk groups based on the optimal threshold determined by the R package survminer. In the training set TCGA cohort, the survival rate of patients in the low-risk group was significantly higher than that in the high-risk group (HR = 6.6, P < 0.001), with AUC values ​​of 0.842, 0.876, and 0.866 at 1, 3, and 5 years, respectively. Figure 3D As shown in the validation set GSE13507 cohort, the survival rate of low-risk patients was significantly higher than that of high-risk patients.

[0121] The survival rate of low-risk patients was significantly higher than that of high-risk patients (HR = 6.38, P < 0.001) (0.839, 0.878, and 0.865, respectively). Figure 3E As shown. In the external dataset, GSE31684 (HR=3.88, P<0.001) and CJUH (HR=9.78) were found.

[0122] Similar results were obtained in the P<0.001 cohort. The 1-year, 3-year, and 5-year AUC values ​​for the GSE31684 cohort were 0.735, 0.793, and 0.799, respectively. The 1-year, 3-year, and 5-year AUC values ​​for the CJUH cohort were 0.912, 0.961, and 1.034, respectively. Figure 3F and Figure 3G As shown in the figure. These results demonstrate that SCPS exhibits robust performance across multiple independent queues.

[0123] 5. Relationship between SCPS and clinical factors

[0124] The correlation between SCPS and clinical stage parameters and patient survival was assessed by comparing subgroup differences in RS based on clinical indicators. Results showed that in different cohorts, such as... Figures 4A-4DAs shown, the risk coefficient for early-stage patients is generally lower than that for late-stage patients. Similarly, late-stage patients account for a higher proportion of high-risk patients, while early-stage patients account for a higher proportion of low-risk patients. Comparing survival rates (OS / DSS / DFI / PFI) between high-risk and low-risk groups, significant differences were found in all subgroups except for DFI in the TCGA cohort. Therefore, the survival rate for low-risk patients is generally higher than that for high-risk patients.

[0125] 6. Comparison of prognostic features based on BLCA gene expression

[0126] To evaluate the superiority of the SCPS model developed in this invention, a comprehensive literature search was conducted on published features. Ultimately, 11 features were identified. These features are related to various biological processes, including immune responses, autophagy, ferroptosis, dehydration, and epithelial-mesenchymal transition.

[0127] Univariate Cox regression analysis was performed on each feature set of all datasets, and it was found that only the model of this invention could effectively determine the patient survival rate for each cohort, such as Figure 4E As shown, this demonstrates the stability of SCPS. Furthermore, comparing the C-index of SCPS with other characteristics, such as... Figure 4F and Figure 4G As shown. Notably, SCPS outperforms almost all models on every dataset.

[0128] Most features perform well on the training dataset TCGA, but poorly on the validation dataset. This is likely due to the poor generalization ability of these models obtained through fitting. Filtering with ML algorithms, combined with dimensionality reduction and processing, gives them better potential for external inference.

[0129] 7. The relationship between SCPS and immune landscape and immune cell infiltration

[0130] In the TCGA and GSE13507 cohorts, the tumor microelectroencephalopathy (TME) of patients was assessed using the ESTIMATE algorithm. The SCPS risk assessment system was used to compare differences in immune function between high-risk and low-risk groups. Results showed that the low-risk BLCA group had lower immune scores, ESTIMATE scores, and matrix scores than the high-risk group (there was no statistically significant difference in immune scores among the TCGA group), and tumor purity was significantly higher in the low-risk group than in the high-risk group. Figure 5A As shown in the figure, this indicates a higher degree of immune cell infiltration and lower tumor purity in high-risk samples. To systematically assess the differences in tumor immune cell infiltration, we used five algorithms (TIMER, CIBERSORT, QUANTISEQ, MCPcounter, and EPIC) to quantitatively assess the abundance of immune cell infiltration in patient samples.

[0131] like Figure 5BAs shown, the heatmap illustrates the correlation between immune assessment and immune cell infiltration. Analysis of the TCGA and GSE13507 cohorts revealed that immune scores, matrix scores, and ESTIMATE scores were positively correlated with the abundance of infiltration of exhausted CD8+ T cells, macrophages, monocytes, endothelial cells, Tregs, CAFs, and fibroblasts.

[0132] Similarly, SCPS gene expression is positively correlated with the infiltration abundance of exhausted CD8+ T cells, macrophages, Tregs, CAFs, fibroblasts, etc., of immune cells. Figure 5C As shown. Subsequently, the differences in tumor immune cell infiltration between the high-risk and low-risk groups were compared. The infiltration of B memory cells, exhausted CD8+ T cells, macrophages, monocytes, and neutrophils in the low-risk group was significantly lower than that in the high-risk group (p<0.05), as shown. Figure 5D As shown.

[0133] 8. SCPS as a predictive biomarker for immunotherapy in BLCA patients

[0134] In recent years, monoclonal antibodies targeting immune checkpoint molecules have provided a significant breakthrough in cancer treatment. This invention determines the differences in the expression of immunosuppressive factors in high- and low-risk populations. Results are as follows... Figure 5E As shown, the expression levels of PDCD1, CD274, and CTLA4 in the high-risk group were significantly higher than those in the low-risk group (p<0.05). Furthermore, the expression levels of numerous genes in the immune activating factors, chemokines, receptors, and MHC were found to be significantly higher in the high-risk group than in the low-risk group (p<0.05). Further investigation was conducted to determine whether SCPS could predict the response of BLCA patients to immunotherapy. The TIDE algorithm was used to predict the potential immune response of patients. In the TCGA and GSE13507 cohorts, there was a statistically significant difference in TIDE scores between the high-risk and low-risk groups, with the high-risk group having a higher TIDE score (p<0.01). Figure 5F As shown, the results suggest that immune checkpoint blockade therapy is less effective in high-risk patients. Therefore, SCPS can serve as a predictive biomarker for BLCA immunotherapy.

[0135] 9. Subcellular localization of SLC16A1 affects gene expression, pathological progression, patient survival, and immune infiltration.

[0136] Analysis of tissue samples from 86 BLCA patients revealed different subcellular localization types of MCT1, mainly including 54 nuclear staining (N) groups and 19 membrane / cytoplasmic staining (MC) groups, with 13 N / MC groups falling between the two. Figure 6AAs shown. Combining patient clinical data, the location of MCT1 was also correlated with the AJCC stage of disease progression, with N-type patients (86.8%) primarily in stage 0-I. In MC (89.5%) and N / MC (84.6%) patients, the main diagnosis was stage II-IV (p<0.0001). SLC16A1 expression was also significantly elevated in N, N / MC, and MC groups (p<0.05). Figure 6B As shown, patient survival showed a gradually worsening trend, with median survival of 80.5 months, 68 months, and 43 months in the N, N / MC, and MC groups, respectively. Figure 6C As shown.

[0137] Subsequently, the transcriptomics and proteomics of patients with both N and MC types were sequenced, such as... Figure 6D As shown, the results match well, as... Figure 6E As shown. Analysis of differentially expressed genes in the N and MC groups revealed that MC patients showed significant enrichment in pathways such as cytoplasm, endoplasmic reticulum, and protein-binding complex tissues, while N patients showed significant enrichment in gene transcription (negative) regulation, RNA biosynthesis pathways, and negative regulation of cellular (metabolic) processes, such as... Figure 6F As shown, genes significantly associated with SLC16A1 expression levels were found in various cell death pathways, including VEGF signaling, glycolysis, apoptosis, autophagy, and ferroptosis. Figure 6G As shown, this suggests that SLC16A1 may be involved in influencing these pathways.

[0138] Then, based on the risk formula, patients in the N group and MC group were divided into risk categories, such as... Figure 6H As shown, N-type patients are all low-risk patients, while MC-type patients are high-risk patients. Furthermore, the expression of eight SCPS genes differed significantly between the high-risk and low-risk groups, such as... Figure 6I As shown. mIHC analysis of patient samples from groups N and MC revealed that the fluorescence expression intensities of CD8, CD68, PD1, and PDL1 were higher in the MC group than in the N group. Figure 6J As shown. Furthermore, high microsatellite instability (MSI) may be beneficial for immune checkpoint drug therapy. DNA mismatch repair genes were downregulated in group N (NES = -1.52, p = 0.04, FDRq = 0.26), such as... Figure 6K As shown, this indicates that immune checkpoint therapy has potential benefits in group N patients. Figure 6L The 3D plot shows that the expression of potential oncogenes MYC and E2F1 was upregulated in patients in the MC group, and the 3D plot shows that there were significant differences in expression among the groups (*p<0.05, **p<0.01, ***p<0.005).

[0139] Example 5

[0140] like Figure 7 As shown, this embodiment provides a comprehensive method for constructing a machine learning-based Joint Predictive Gene Composition (SCPS), including:

[0141] S1, based on a specific BLCA dataset, analyze genes related to exhaustion CD8+ T immune infiltration;

[0142] As a preferred embodiment, the specific BLCA dataset includes the GSE128702, GSE48276, and GSE87304 datasets.

[0143] S2, cross-analysis was performed on the depleted CD8+T immune infiltration-related genes and the pathway genes related to SLC16A1 expression to finally obtain M genes.

[0144] In a preferred embodiment, the method for obtaining the SLC16A1 expression-related pathway genes includes:

[0145] The effects of SLC16A1 tissue-specific expression on patient survival and immune regulation were further validated on multiple BLCA datasets.

[0146] Patients were divided into early T group and late T group according to T stage. The expression of SLC16A1 in the tissues of patients in the late group was higher than that in the early group, and the results of TCGA, GSE87304, GSE13507 and GSE48276 showed significant differences.

[0147] Based on SLC16A1 expression levels, patient samples were divided into high and low expression groups, and GSEA analysis was performed to identify pathway genes related to SLC16A1 expression.

[0148] In this embodiment, the results showed that 11 pathway genes related to SLC16A1 expression were enriched in the GSE48276, GSE87304 and GSE128702 datasets, including cell cycle, focal adhesion and nod-like receptor signaling pathways.

[0149] S3, using univariate Cox analysis, N prognostic genes were identified from the M genes, p < 0.05, N <M;

[0150] S4. Determine the CD8 immune-related prognostic profile based on SLC16A1 using a machine learning ensemble learning algorithm, and determine a combined predictive gene composition based on the CD8 immune-related prognostic profile of SLC16A1.

[0151] In a preferred embodiment, S4 includes:

[0152] S41, In the TCGA training queue dataset, 101 prediction models were built using 10 machine learning algorithms, and the C-index of each model in the training queue and validation dataset GSE13507 was calculated.

[0153] S42, Based on the comparison of the C-index, the optimal model corresponding to the highest C-index is determined to be the combination of Step-Cox[forward] + RSF;

[0154] S43, based on the optimal combination algorithm and Step-Cox [forward] proportional hazards regression, further RSF regression was performed to obtain a model composed of 12 genes including NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

[0155] Example 6

[0156] like Figure 8 As shown, this embodiment provides a machine learning-based Joint Predictive Gene Composition (SCPS) comprehensive construction system, including:

[0157] The immune infiltration-related gene analysis module (101) is used to analyze CD8 immune infiltration-related genes based on a specific BLCA dataset;

[0158] The gene cross-analysis module (102) is used to cross-analyze the CD8 immune infiltration-related genes and the pathway genes related to SLC16A1 expression, and finally obtain M genes.

[0159] The prognostic gene identification module (103) is used to identify N prognostic genes from the M genes using univariate Cox analysis, where p < 0.05, N <M;

[0160] The joint predictive gene composition determination module (104) is used to determine the CD8 immune-related prognostic profile based on SLC16A1 based on a machine learning ensemble learning algorithm, and to determine the joint predictive gene composition based on the CD8 immune-related prognostic profile of SLC16A1.

[0161] The present invention also provides a memory that stores a plurality of instructions for implementing the method as described in Embodiment 1.

[0162] like Figure 9 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301. The memory 302 stores a plurality of instructions, which can be loaded and executed by the processor to enable the processor to perform the method as described in Embodiment 1.

[0163] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A combined predictive gene composition for evaluating the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy, characterized in that: The combined predictive gene composition includes 12 genes: NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

2. A comprehensive construction method for the joint predictive gene composition of claim 1 based on machine learning, characterized in that, include: S1, based on a specific BLCA dataset, analyze genes related to exhaustion CD8+ T immune infiltration; S2, cross-analysis was performed on the depleted CD8+T immune infiltration-related genes and the pathway genes related to SLC16A1 expression to finally obtain M genes; S3, using univariate Cox analysis, N prognostic genes were identified from the M genes, p < 0.05, N <M; S4. Determine the CD8 immune-related prognostic profile based on SLC16A1 using a machine learning ensemble learning algorithm, and determine a combined predictive gene composition based on the CD8 immune-related prognostic profile of SLC16A1.

3. The comprehensive construction method of the combined predictive gene composition according to claim 2, characterized in that, The specific BLCA datasets include GSE128702, GSE48276, and GSE87304.

4. The comprehensive construction method of the combined predictive gene composition according to claim 3, characterized in that, The methods for obtaining the SLC16A1 expression-related pathway genes include: The effects of SLC16A1 tissue-specific expression on patient survival and immune regulation were further validated on multiple BLCA datasets. Patients were divided into early T group and late T group according to T stage. The expression of SLC16A1 in the tissues of patients in the late group was higher than that in the early group, and the results of TCGA, GSE87304, GSE13507 and GSE48276 showed significant differences. Based on the expression of SLC16A1, patient samples were divided into high and low expression groups, and GSEA analysis was performed to identify pathway genes related to SLC16A1 expression.

5. The comprehensive construction method of the combined predictive gene composition according to claim 4, characterized in that, The GSE48276, GSE87304, and GSE128702 datasets were enriched with 11 pathway genes associated with SLC16A1 expression, including cell cycle, focal adhesion, and nod-like receptor signaling pathways.

6. The comprehensive construction method of the combined predictive gene composition according to claim 5, characterized in that, S4 includes: S41, In the TCGA training queue dataset, 101 prediction models were built using 10 machine learning algorithms, and the C-index of each model in the training queue and validation dataset GSE13507 was calculated. S42, Based on the comparison of the C-index, the optimal model corresponding to the highest C-index is determined to be the combination of Step-Cox[forward] + RSF; S43, based on the optimal combination algorithm and Step-Cox [forward] proportional hazards regression, further RSF regression was performed to obtain a model consisting of 12 genes: NRP2, SOD2, NCBP1, FKBP5, DEGS1, ME2, LAP3, CCDC88A, SLC16A1, ANXA5, ASAP1, and HSPA13.

7. A machine learning-based joint predictive gene composition (SCPS) synthesis system for implementing the construction method according to any one of claims 2-6, characterized in that, include: The immune infiltration-related gene analysis module (101) is used to analyze exhaustive CD8+T immune infiltration-related genes based on a specific BLCA dataset; The gene cross-analysis module (102) is used to cross-analyze the depleted CD8+T immune infiltration-related genes and the pathway genes related to SLC16A1 expression, and finally obtain M genes. The prognostic gene identification module (103) is used to identify N prognostic genes from the M genes using univariate Cox analysis, where p < 0.05, N <M; The joint predictive gene composition determination module (104) is used to determine the CD8 immune-related prognostic profile based on SLC16A1 based on a machine learning ensemble learning algorithm, and to determine the joint predictive gene composition based on the CD8 immune-related prognostic profile of SLC16A1.

8. A detection kit for a combined predictive gene composition based on the multi-gene signature of SLC16A1-associated exhausted CD8+ T cells for assessing the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy, for use in assessing the prognosis of urothelial carcinoma and predicting the efficacy of immunotherapy by means of the combined predictive gene composition of claim 1 and the detection of the multi-gene signature of SLC16A1-associated exhausted CD8+ T cells.

9. The use of the combined predictive gene composition detection kit of claim 8 in predicting the survival of patients with urothelial carcinoma.

10. The use of the combined predictive gene composition detection kit of claim 8 in predicting the efficacy of immunotherapy and MCT1 inhibitor therapy in patients with urothelial carcinoma.

Images