Lipid metabolism subtypes reflects the prognosis and immune profiles of bladder cancer patients by NMF clustering and building related signature

Abstract

Background

Lipid metabolism is crucial in tumor formation and progression. However, the role of lipid metabolism genes (LMGs) in bladder cancer (BLCA) are unknown. The purpose of this study was to construct a LMGs-related subtypes that predicted the treatment and prognosis of BLCA patients.

Methods

The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were used for this study. The gene set enrichment analysis (GSEA) was utilized to distinguish functional differences between high-risk (HR) and low-risk (LR) groups. Single-sample GSEA (ssGSEA) was employed to determine potential associations between prognostic outcomes and immune status.

Results

First, BLCA patients were divided into two subtypes by non-negative matrix factorization (NMF) clustering, and there were substantial variations in survival status, immune cell infiltration and immune classification between the two subtypes. Next, a prognostic signature involving 8 LMGs was identified (AKR1B1, SCD, CYP27B1, UGCG, SGPL1, FASN, TNFAIP8L3, PLA2G2A). HR patients exhibited worse outcome than LR patients. Multivariate Cox regression analysis confirmed that LMGs-related signature was an independent prognostic indicator of BLCA patients’ survival. Compared with clinicopathological variables, LMGs-related signature showed higher prognostic predictive ability, with an area under curve of 0.720 at 5 years of follow-up. Through immunotherapy analysis, drug sensitivity analysis, TIDE score and immune cell infiltration characteristics, LMGs-related signature was confirmed to accurately predict the prognosis and treatment response of BLCA patients.

Conclusion

Our newly established prognostic signature, which involved eight LMGs, can give prognostic distinction for BLCA and may eventually lead to novel targets for treatment.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-024-01631-8.

Introduction

Bladder cancer (BLCA) is a kind of urological malignancy with a high prevalence and an annual morbidity and fatality rate that is gradually growing [1]. It is classified clinically into two types: non-muscle-invasive bladder cancer (NMIBC) and muscle-invasive bladder cancer (MIBC). Among them, NMIBC is the most common, it will not endanger life, but it is easy to relapse after surgery [2]. Currently, the 5-year overall survival (OS) rate of NMIBC patients treated with a first-line cisplatin-based combination chemotherapy regimen surpasses 90%. Unfortunately, 50–70% of these individuals may relapse [3], and an even greater number will advance to a more hazardous MIBC subtype [4]. As tumor immunology progressed, it became obvious that immunotherapy and immune-related variables were substantially correlated with BLCA patient prognosis [5]. Given the significant increase in BLCA incidence and susceptibility to relapse and progression, we urgently need to uncover novel predictive markers and build risk prediction models that are more robust than conventional prognostic tools.

Lipids are critical elements of bio-membranes and chemical messengers in cellular processes [6]. Lipid metabolism is necessary for maintaining tissue homeostasis [6, 7]. Many studies have revealed that lipid metabolism had a role in the development, relapse, and tumor cells of BLCA [8, 9]. Multiple tumors, including kidney tumor, bladder tumor, and colorectal tumor, have also exhibited dysregulated lipid metabolism [10–12]. Nonetheless, there has been little investigation into the characterization and risk profiles of lipid metabolism genes (LMGs).

The non-negative matrix factorization (NMF) method, which is based on the “NMF” package to extract biological correlation coefficients of data within a gene expression matrix, grasps internal structural features of data by organizing genes and samples, and thus groups samples, is currently widely used in disease typing. The purpose of this work is to combine NMF clustering and building related signature to uncover a function for LMGs in estimating the prognosis and immune microenvironment for patients with BLCA.

Materials and methods

Transcriptome data downloading and processing

The transcriptome data and clinical information of BLCA were retrieved via The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/repository). Through matching, a total of 395 samples with expression matrices and clinical data were found. The downloaded transcriptome data is HTseQ-FPKM. The Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) was taken and it served as an independent external validation cohort (GSE13507) [13]. This data set was released in 2010, including 165 primary BLCA samples, and the sequencing method is “expression profiling by array”. After log2 transformation, transcriptome data were used for subsequent research.

NMF clustering

The LMGs were obtained from lipid metabolism-related pathways in Molecular Signature Database v7.0 (http://www.gsea-msigdb.org/gsea/index.jsp). Next, BLCA-related LMGs were processed through NMF clustering algorithm using the “NMF” R package. The NMF analysis and 50 iterations were carried out with the standard “brunet” pattern. The optimal number of clusters was indicated by k values ranging from 2 to 10. The average contour width of the common member matrix was determined with the “NMF” R package, and the minimum member of each subclass was set to 10. The optimal k value was established using cophenetic, residual sum of squares, and silhouette indications.

Construction and validation of the prognostic model

First, univariate COX regression was used to identify LMGs associated with prognosis. After that, the prognosis-related LMGs were investigated using least absolute shrinkage and selection operator (LASSO) regression, with family set to “Cox” and Maxit set to 1000. The survival differences between the high-risk (HR) group and the low-risk (LR) group were investigated in the training and validation cohorts, as well as whether LMGs-related signature could more properly classify patients’ risk categories.

The construction of nomogram

To more correctly assess the prognosis of BLCA patients, clinical data and LMGs-related signature were integrated using the “Regplot” R package and then we constructed a nomogram. The C-index value and calibration curves were conducted to assess the accuracy of survival prediction.

Gene set enrichment analysis

The gene set enrichment analysis (GSEA) was performed to determine potential pathways and molecular mechanisms of differentially expressed genes (DEGs) between the HR group and the LR group in TCGA cohorts. Significance was defined as a two-tailed p-value less than 0.05 and logFC greater than 1. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment were analyzed using GSEA methods and we visualized the first five pathways of each project.

Tumor mutational burden, immunotherapy analysis, drug sensitivity analysis

Tumor mutational burden (TMB) in BLCA patients was assessed and calculated using the “maftools” R package. Tumor Immune Dysfunction and Exclusion (TIDE) was utilized to predict the efficacy of immunotherapy. The half-maximal inhibitory concentration (IC50) of each BLCA patient compared to a common antineoplastic drug was calculated and visualized using the Genomics of Drug Sensitivity in Cancer (GDSC) platform and “pRRophetic” (version 0.5) R package.

Immunohistochemical analysis

The Human Protein Atlas (HPA) website (https://www.proteinatlas.org) provided IHC staining data, which was a database based on proteomic, transcriptome, and systems biology data that could map tissues, cells, and organs. IHC staining of AKR1B1, SCD, UGCG, SGPL1, FASN, TNFAIP8L3 and PLA2G2A in BLCA tumor tissues and normal bladder tissues were extracted from HPA. However, the IHC staining of CYP27B1 was lacking in the HPA database.

Statistical analysis

The data in this study was analyzed using R software (version 4.1.3). The Chi-square test, Fisher’s exact test, and Wilcoxon rank sum test were performed to compare the differences across groups. The link between risk score and tumor-infiltrating immune cells was investigated using Spearman’s rank correlation. The Kaplan–Meier curves with log-rank test, as well as univariate and multivariate Cox proportional hazard methods were used in the survival comparisons across groups. In the current study, p < 0.05 indicated statistical significance.

Results

Identifying lipid metabolism subtypes in BLCA patients

The analytical process of this study was shown in Fig. 1. We obtained 829 genes from lipid metabolism-related pathways, 733 of which were expressed in TCGA-BLCA tissues. Then, according to the thresholds of logFC ≥ 1 and P < 0.05, we found 75 DEGs related with lipid metabolism (Fig. 2A). The cophenetic coefficients, which shown the stability of classified cluster, were used to calculate the optimal k value, and we selected k = 2 as the optimal value (Fig. 2B). Consequently, 2 molecular subtypes (cluster 1 and cluster 2) were identified based on LMGs. Notably, the matrix heatmap had clear borders based on a k value of 2, suggesting that the molecular subtypes classification was stable.

Fig. 1.

The analysis process of this study

Fig. 2.

Identification of lipid metabolism subtypes in BLCA patients by NMF analysis. A Heatmap of DEGs related with lipid metabolism; B Two molecular subtypes (cluster 1 and cluster 2) were identified based on K = 2 as the optimal value; OS (C) and PFS (D) between cluster 1 and cluster 2

Survival analysis and immune cell infiltration among two clusters

Survival analysis revealed that cluster 1 had significantly shorter OS (P < 0.001) and progression-free survival (PFS) (P = 0.002) than cluster 2 (Fig. 2C, D). The MCPcounter method was used to compare the immune scores between the 2 clusters. Except for neutrophil (P = 3.1e−05), the immune scores of NK cell (P < 2.22e−16), monocytic lineage (P < 2.22e−16), fibroblasts (P < 2.22e−16), cytotoxic lymphocytes (P < 2.22e−16), and CD8+ T cells (P = 5e−08) in clusters 1 were higher than that of cluster 2 (Fig. 3A). Immunophenotyping analysis showed that most of BLCA patients in cluster 1 belonged to immune C2 (IFN-gamma Dominant), while most of BLCA patients in cluster 2 belonged to immune C1 (Wound Healing) (Fig. 3B). Collectively, these results revealed that LMGs could classify BLCA into distinct molecular subtypes and was associated with clinical prognosis.

Fig. 3.

The immune cell infiltration among two clusters (cluster 1 and cluster 2). A The immune scores of cluster 1 vs cluster 2 using MCPcounter algorithm; B Immunophenotyping analysis of cluster 1 and cluster 2

Construction and validation of the prognostic model

We randomly divided TCGA-BLCA patients into training dataset (279 patients) and testing dataset (116 patients) at a ratio of 7:3. Basic clinical information of each group was uploaded to Supplementary Table 1. LASSO Cox regression analysis (Fig. 4A, B) and univariate Cox analysis (Table 1) was done with P < 0.05, and 16 LMGs related with prognosis in the training dataset were screened. 16 LMGs were subjected to a multivariate Cox analysis, and 8 LMGs were included in the model, and coefficients of these genes was showed in Table 2. The calculation formula of the model was: risk score = AKR1B1*(0.1732) + SCD*(0.1391) + CYP27B1*(− 0.4124) + UGCG*(− 0.2133) + SGPL1*(− 0.2579) + FASN*(0.2319) + TNFAIP8L3*(0.5419) + PLA2G2A*(− 0.1171). The BLCA patients were then separated into the HR group and the LR group according to the median value of risk score. As shown in Fig. 4C–J, in the training cohort, testing cohort, overall cohort (TCGA-BLCA) and external validation cohort (GSE13507), it was observed that the prognosis of the HR group was poorer than that of the LR group (P < 0.05). The receiver operating characteristic (ROC) curve, as shown in Fig. 4C–J, revealed that this model can discriminate BLCA patient’s prognosis well in each cohort. Subgroup analysis of clinical indicators showed that the risk score exhibited stable predictive ability in different clinical situations (Fig. 5A–L). String database was utilized to investigate the potential protein-related interactions of eight LMGs, as shown in the Supplementary Fig. 1.

Fig. 4.

Construction and validation of LMGs-related signature. A, B LASSO regression analysis and regression coefficient; C Kaplan–Meier curve of OS in training dataset; D AUC values of 1-, 3-, 5-year OS in training dataset; E Kaplan–Meier curve of OS in testing dataset; F AUC values of 1-, 3-, 5-year OS in testing dataset; G Kaplan–Meier curve of OS in overall dataset; H AUC values of 1-, 3-, 5-year OS in overall dataset; I Kaplan–Meier curve of OS in GEO dataset; J AUC values of 1-, 3-, 5-year OS in GEO dataset

Table 1.

Univariate analysis for LMGs in the training dataset

Gene	HR	HR.95%CI	P-value
MGLL	1.2685	1.0223–1.5739	0.0307
AKR1B1	1.3296	1.1318–1.5619	0.0005
SCD	1.2174	1.0458–1.4171	0.0111
TRIB3	1.2864	1.0620–1.5581	0.0100
SQLE	1.3001	1.0239–1.6507	0.0312
CYP27B1	0.6584	0.4658–0.9305	0.0179
UGCG	0.7228	0.5349–0.9766	0.0345
FADS1	1.3407	1.1244–1.5987	0.0011
PLA2G2F	0.8583	0.7682–0.9591	0.0069
ACP6	0.7296	0.5722–0.9303	0.0110
SGPL1	0.7712	0.6027–0.9868	0.0389
FASN	1.3585	1.0989–1.6792	0.0046
DHCR7	1.3041	1.0313–1.6491	0.0266
TNFAIP8L3	1.3304	1.0732–1.6493	0.0092
PLA2G2A	0.9115	0.8321–0.9985	0.0463
CPT1B	0.6669	0.4764–0.9337	0.0183

HR Hazard ratio, CI confidence interval

Table 2.

Coefficient of eight LMGs included by multivariate analysis

Gene	Coefficient
AKR1B1	0.1732
SCD	0.1391
CYP27B1	− 0.4124
UGCG	− 0.2133
SGPL1	− 0.2579
FASN	0.2319
TNFAIP8L3	0.5419
PLA2G2A	− 0.1171

Fig. 5.

The survival outcomes of HR and LR groups were stratified by various clinicopathological features including age (A, B), gender (C, D), stage (E, F), T stage (G, H), N stage (I, J) and M stage (K, L)

Nomogram construction and clinical impact

Univariate Cox analysis revealed that age (HR = 1.0382, P = 1.44E−05), stage (HR = 1.8251, P = 2.14E−08) and risk score (HR = 1.2697, P = 6.99E−13) were prognostic variables in BLCA patients (Table 3). Multivariate Cox analysis showed that age (HR = 1.0327, P = 0.0002), stage (HR = 1.6553, P = 4.38E−06) and risk score (HR = 1.1918, P = 5.57E−07) could be considered as independent prognostic indicators (Table 3). To accurately predict patients’ OS, we developed a nomogram including risk score and other clinical characteristics (Fig. 6A), and visualized the performance of nomogram with ROC (Fig. 6B) and calibration plots at 1, 3, 5 years (Fig. 6C). The decision curve demonstrated that the nomogram outperformed other clinical indicators in predicting the prognosis of BLCA patients (Fig. 6D). The C index (Fig. 6E) and root mean square (RMS) curve (Fig. 6F) showed that our model was more stable than several models currently under development [14–17]. Clinical analysis identified substantial differences in patients’ stage, grade, T stage, N stage, M stage and response between the HR group and the LR group (Fig. 7).

Table 3.

Univariate and multivariate analysis of clinical indicators and risk score

Items	Univariate analysis				Multivariate analysis
	HR	HR.95L	HR.95H	P-value	HR	HR.95L	HR.95H	P-value
Age	1.0382	1.0207	1.0559	1.44E-05	1.0327	1.0152	1.0504	0.0002
Gender	0.8891	0.6259	1.2629	0.5115	NA
Grade	9612415	0	Inf	0.9915	NA
Stage	1.8251	1.4785	2.2529	2.14E−08	1.6553	1.3349	2.0526	4.38E−06
Risk score	1.2697	1.1895	1.3552	6.99E−13	1.1918	1.1127	1.2766	5.57E−07

Fig. 6.

Nomogram construction in TCGA database. A Construction of a prognostic nomogram to predict 1-, 3-, 5-year OS of BLCA patients; B AUC values of nomogram, risk and clinicopathological features; C the calibration curve of nomogram; D decision curve of nomogram and other clinical indicators; C-index (E) and RMS curves (F) analysis of our model and several models

Fig. 7.

Clinical influences of risk score for BLCA patients including age (A), gender (B), stage (C), grade (D), T stage (E), N stage (F), M stage (G) and response (H)

Immunological checkpoints, immunological reaction, and anticancer drug responsiveness

We analyzed the effect of risk score and signature associated LMGs on immune checkpoints and discovered that their effect on immune checkpoints was highly significant (Fig. 8A). Researchers discovered that two groups had significant variations in their responses to ctla-4_neg_pd1_neg, ctla-4_pos_pd1_neg, ctla-4_pos_pd1_pos, and ctla-4_neg_pd1_pos, indicating that individuals with distinct risk ratings had substantially varying immunological responses (Fig. 8B). Research on the susceptibility of BLCA individuals to anti-cancer agents might aid in the growth of therapeutic therapies. Since seen in Fig. 8C, the HR group proved particularly susceptible to Cyclopamine, Sorafenib, Pazopanib, and Parthenolide, while the LR group proved particularly susceptible to SB590885, Gefitinib, Erlotinib, and BIBW2992. These findings are instructive for us to select specific drugs based on anti-tumor drug sensitivity. Besides, we aggregated the results of 395 BLCA patients computed using various algorithms and compared all immune cell subtypes in two groups, and the proportion of partial cell subtypes infiltrated was significantly different between the two groups (Supplementary Fig. 2).

Fig. 8.

Immune checkpoints (A), immunotherapy response (B) and anti-tumor drug sensitivity (C) between the LR group and the HR group

Tumor mutational burden

The frequency of somatic mutations was directly related to tumor immune cells infiltration. Therefore, we used the “maftools” package to examine the changes in the frequency of somatic mutations in two groups. The total somatic mutation rate differed considerably between the two groups, with the top 10 most commonly mutated genes being TP53, TTN, KMT2D, MUC16, ARID1A, KDM6A, PIK3CA, SYNE1, RYR2 and KMT2C (Fig. 9A, B). The high-TMB was positively associated with a longer OS (P < 0.001) (Fig. 9C). BLCA patients with higher TMB had a better outcome in both the LR and HR groups (Fig. 9D). Figure 10A showed the close relationship between risk score, TMB score and immune cells, indicating that LMGs-related signature may play a role via regulating immunity and mutation. The correlation analysis identified that risk score was significantly related with CD8+ T cells, Monocytic lineage and Fibroblasts, suggesting that LMGs-related signature may affect the prognosis of BLCA patients by influencing immune functions (Fig. 10B).

Fig. 9.

Tumor mutational burden between the LR group and the HR group. A, B The top 10 most frequently mutated genes; C low-TMB had a shorter OS than high-TMB; D high-TMB had a better OS in both the LR and HR groups than low-TMB

Fig. 10.

The relationship between risk score, TMB score and immune cells (A); the correlation of risk score and immune cells (B); the functional enrichment analysis of DEGs including GO analysis (C, D) and KEGG analysis (E–F)

Functional enrichment analyses

To understand the differences in functional enrichment of each group, we identified DEGs among them and performed enrichment analysis. The GO analysis revealed that the LR group was observably enriched in ADAPTIVE MMUNE RESPONSE, IMMUNOGLOBULIN COMPLEX, ANTIGEN BINDING and RNA BINDING INVOLVED IN POSTTRANSCRIPTIONAL GENE SILENCE (Fig. 10C); the HR group was enriched in EPIDERMAL CELL DIFFERENTIATION, EPIDERMIS DEVELOPMENT, KERATINIZATION, KERATINOCYTE DIFFERENTIATION and SKIN DEVELOPMENT (Fig. 10D). The KEGG analysis identified that the LR group was enriched in PEROXISOME, PORPHYRIN AND CHLOROPHYLL METABOLISM, PRIMARY BILE ACID BIOSYNTHESIS, RIBOSOME, RIG I LIKE RECEPTOR SIGNALING PATHWAY (Fig. 10E); the HR group was enriched in CELL CYCLE, DILATED CARDIOMYOPATHY, MELANOMA, NEUROACTIVE LIGAND RECEPTOR INTERACTION, REGULATION OF ACTIN CYTOSKELETON (Fig. 10F).

Validation of signature-related LMGs

Based on gene expression patterns from BLCA tissues and paracancerous tissues in the TCGA database, five genes (AKR1B1, SCD, CYP27B1, SGPL1, FASN) were significantly up-regulated in BLCA tissues and three genes (UGCG, TNFAIP8L3, PLA2G2A) were significantly down-regulated in BLCA tissues (Fig. 11). Meanwhile, we showed significant differences in IHC staining of AKR1B1, SCD, UGCG, SGPL1, FASN, TNFAIP8L3, and PLA2G2A in tumor and normal tissues (Fig. 12).

Fig. 11.

Validation expression patterns of signature-related eight LMGs including AKR1B1, SCD, CYP27B1, UGCG, SGPL1, FASN, TNFAIP8L3, PLA2G2A between BLCA tissues and paracancerous tissues in TCGA database

Fig. 12.

Verify the expression of AKR1B1, SCD, UGCG, SGPL1, FASN, TNFAIP8L3, PLA2G2A in BLCA and normal bladder tissues from HPA database

Discussion

The first-line therapy for BLCA is currently a combination of cisplatin chemotherapy, which can provide full remission to the majority of BLCA patients and a 5-year OS of 90% [3]. Nonetheless, BLCA patients are susceptible to relapse and infiltration after conventional treatment, so new prognostic markers are desperately needed [18].

In this study, 2 molecular subtypes (cluster 1 and cluster 2) in TCGA-BLCA patients were identified based on LMGs by NMF clustering. Survival analysis revealed that cluster 1 had significantly shorter OS and PFS than cluster 2. MCPcounter analysis identified that the immune score of NK cell, monocytic lineage, fibroblasts, cytotoxic lymphocytes and CD8+ T cells in clusters 1 was greater than that of cluster 2. Immunophenotyping analysis revealed that majority of BLCA patients in cluster 1 belonged to IFN-gamma Dominant, while the majority of BLCA patients in cluster 2 belonged to Wound Healing. Collectively, these findings revealed that LMGs could classify BLCA patients into distinct molecular subtypes which were associated with clinical prognosis.

Herein, we developed a lipid metabolism-related signature involving 8 LMGs, and validated its predictability as a prognostic biomarker of BLCA patient’s outcome. To investigate the value of this model in BLCA, we conducted the following analysis: first, based on the OS and LMGs expression of BLCA patients, we identified 16 prognosis-related genes with lipid metabolism using the LASSO and univariate Cox analyses. Then, using multivariate Cox analysis, we produced a prognostic model containing 8 LMGs and classified BLCA patients into either HR or LR cohorts based on the median risk score. Moreover, using both survival and risk profile analyses, we proved that LMGs-related signature strongly differentiated OS between the HR and LR cohorts. Using univariate Cox, multivariate Cox, and ROC curves analyses, we identified this LMGs-related signature to be independent prognostic indicators for BLCA patients, with strong validity in predicting prognosis.

The role of lipid metabolism in BLCA has been much reported. Liu et al. identified that elevated lncDBET triggered the PPAR signaling pathway to increase the lipid metabolism of BLCA cells via direct interaction with FABP5, hence enhancing the development of BLCA in vitro and in vivo [8]. Chao et al. reported MEX3C as a novel oncogene that promoted BLCA by regulating lipid metabolism via MAPK/JNK pathway [19]. Our study combined NMF clustering and building related signature to adequately reveal the significant roles for eight LMGs (AKR1B1, SCD, CYP27B1, UGCG, SGPL1, FASN, TNFAIP8L3, PLA2G2A) in estimating the prognosis and immune microenvironment for BLCA patients.

In addition, we generated a nomogram based on age, clinical stage, and LMGs-related signature, and 1-, 3-, 5-year calibration curves demonstrated a good consistency between the predicted and actual OS of BLCA patients. Moreover, we further validated that our model can better predict the OS of BLCA patients with various clinicopathological variables. Subsequently, we performed GO and KEGG analysis of DEGs in HR and LR patients to explore possible mechanisms of action. GO analysis revealed that DEGs were primarily enriched in ADAPTIVE MMUNE RESPONSE, IMMUNOGLOBULIN COMPLEX, ANTIGEN BINDING and RNA BINDING INVOLVED IN POSTTRANSCRIPTIONAL GENE SILENCE, and KEGG analysis showed that DEGs had a remarkable enrichment in PEROXISOME, PORPHYRIN AND CHLOROPHYLL METABOLISM, PRIMARY BILE ACID BIOSYNTHESIS, RIBOSOME, RIG I LIKE RECEPTOR SIGNALING PATHWAY.

Finally, we observed that HR and LR patients had more diverse immune-related functions and immunotherapy responses based on the immunological checkpoints, TIDE score, and immune cell invasion assessments. According to their risk scores, we also predicted prospective therapeutic drugs for HR BLCA patients. Previous studies indicated that TMB was a possible indicator of reactions to immunotherapy and favorably linked to anti-PD-1 and anti-PD-L1 reactions [20]. Therefore, we hypothesized that immune checkpoint blockade (ICB) therapy would be more effective in the HR group. Furthermore, participants who had greater TIDE scores was more inclined to lose immunity against tumors and exhibited worse rates of responding to ICB treatment. TIDE outperformed TMB when assessing ICB-treated individuals’ survival results [21]. From our investigation, individuals participating in the HR group exhibited reduced TIDE scores, indicating that ICB treatment was more helpful for them. Like a result, we hypothesized that an LMG-related pattern may serve as an accurate marker for BLCA patients’ therapeutic reactions. Drug sensitivity research has provided a theoretical basis for the selection of chemotherapy medicines. Taken together, our findings suggested that eight LMGs-related signatures may be complementary to the pathological classification of BLCA, thereby contributing to the risk assessment and individualized therapy of BLCA patients.

Limitation

However, our study also faced certain limitations. Firstly, owing to limitations in data availability, we only employed data from public databases for modeling and validation. Therefore, we recommended additional clinical patients’ data to validate the performance and applicability of LMGs-related signature. Secondly, the underlying mechanism of LMGs in BLCA required additional experimental verification. Third, the expression of eight LMGs in BLCA samples required further analysis. Fourth, conducting drug sensitivity assays through manipulating the identified LMGs (overexpression or knockdown) would be helpful for the treatment of BLCA patients.

Conclusions

Our newly established prognostic signature, which involved eight LMGs, can give prognostic distinction for BLCA and may eventually lead to novel targets for treatment.

Author contributions

ZZ and WT constructed this study. ZZ, CY and LY performed the data analysis, figures plotted, and writing. ZY and WT were responsible for the critical reading of the manuscript. All authors contributed to the article and approved the submitted version.

Funding

None.

Data availability

All data are obtained in the article.

Declarations

Ethics approval and consent to participate

All patients were from public databases and have been approved by ethics.

Competing interests

The authors declare no competing interests.