Prediction of the Prognosis of Clear Cell Renal Cell Carcinoma by Cuproptosis-Related lncRNA Signals Based on Machine Learning and Construction of ceRNA Network

Background Clear cell renal cell carcinoma's (ccRCC) occurrence and development are strongly linked to the metabolic reprogramming of tumors, and thus far, neither its prognosis nor treatment has achieved satisfying clinical outcomes. Methods The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, respectively, provided us with information on the RNA expression of ccRCC patients and their clinical data. Cuproptosis-related genes (CRGS) were discovered in recent massive research. With the help of log-rank testing and univariate Cox analysis, the prognostic significance of CRGS was examined. Different cuproptosis subtypes were identified using consensus clustering analysis, and GSVA was used to further investigate the likely signaling pathways between various subtypes. Univariate Cox, least absolute shrinkage and selection operator (Lasso), random forest (RF), and multivariate stepwise Cox regression analysis were used to build prognostic models. After that, the models were verified by means of the C index, Kaplan–Meier (K-M) survival curves, and time-dependent receiver operating characteristic (ROC) curves. The association between prognostic models and the tumor immune microenvironment as well as the relationship between prognostic models and immunotherapy were next examined using ssGSEA and TIDE analysis. Four online prediction websites-Mircode, MiRDB, MiRTarBase, and TargetScan-were used to build a lncRNA-miRNA-mRNA ceRNA network. Results By consensus clustering, two subgroups of cuproptosis were identified that represented distinct prognostic and immunological microenvironments. Conclusion A prognostic risk model with 13 CR-lncRNAs was developed. The immune microenvironment and responsiveness to immunotherapy are substantially connected with the model, which may reliably predict the prognosis of patients with ccRCC.


Introduction
CCRCC is the most prevalent subtype of renal malignancy, accounting for nearly 70% of all cases [1]. In addition, it exhibits higher rates of recurrence, metastasis, and mortality when compared to chromophobe cell renal carcinoma (cRCC) and papillary renal cell carcinoma (pRCC) [2,3]. Due to the insidious nature of ccRCC, 30% of patients have metastases when they are frst diagnosed [4]. Currently, partial or radical nephrectomy is the best treatment option for nonmetastatic ccRCC patients, but this procedure has a postoperative recurrence rate that can range from 20 to 40%, which has a substantial impact on patient prognosis [5]. In addition, radiation and chemotherapy frequently have poor results for patients with metastatic ccRCC, and drug resistance brought on by prolonged medication frequently results in a terrible prognosis. Despite the fact that immunotherapies such as programmed death-1 (PD-1) and programmed death ligand 1 (PD-L1) have been employed in the treatment of ccRCC recently and have demonstrated some therapeutic results, some patients still do not respond well to this course of action [6,7].
According to previous research, copper can induce tumor angiogenesis, which aids in the progression of cancer, as well as be directly linked to the occurrence and growth of a variety of malignancies [8][9][10]. Besides that, certain outcomes have been attained previously based on the use of copper ion chelators in the therapy of cancer [11,12]. Te key to pathological and physiological processes is cell death, and cuproptosis is the most recent type of death that difers from previous cell deaths such as apoptosis [13], necrosis [14], and ferroptosis [15]. According to the research, iron-sulfur cluster protein loss and fatty acylated protein aggregation are induced by copper binding to tricarboxylic acid (TCA) cycle fatty acyl proteins, which results in death from toxic protein stress [16]. In this regard, it is worth noting that studies have shown that the occurrence and development of ccRCC frequently involve reprogramming of the TCA cycle. Tis is primarily accomplished by afecting the upregulation of the VHL/HIF pathway, which results in the inhibition of the TCA cycle, thereby promoting the occurrence and development of ccRCC [17][18][19]. In view of this, the cuproptosis theory may provide a novel approach to the therapy of ccRCC.
Long noncoding RNA (LncRNA) is a subclass of noncoding RNAs that can take part in and control a number of pathophysiological processes. lncRNA is a noncoding RNA whose biological function is more than 209 bases long. Similar to coding genes, lncRNAs can be chromatin reprogrammed. Dysregulation and posttranscriptional regulation of enhancers are widely involved in biological, physiological, and pathological processes. As a newly discovered class of RNA molecules, several lncRNAs have been identifed as biomarkers of cancer, which control tumor proliferation, immune evasion, cell death resistance, and regional or distant metastasis. Terefore, lncRNA represents an important improvement in our understanding of copper worm disease and evidence that lncRNA is a therapeutic target that can induce GC copper Fibrobacteres. However, the specifc role of lncRNA in the adjustment of aeruginosa is largely unknown. By controlling metabolic reprogramming, lncRNA can regulate carcinogenesis [20]. Additionally, studies have shown that lncRNAs play a variety of roles in the development of ccRCC, including upregulating lncRNA PVT1 and activating the HIF2α pathway to promote the growth and progression of ccRCC cells, as well as lncRNA HCG18, which promotes ccRCC migration and transfer by modulating the miR-152-3/RAB14 axis [21,22]. LncRNA can also be used to predict the progression of ccRCC [23,24]. In a recent study, it was found that CRGS is linked to immune infltration and the immune checkpoint PD-1, which can help predict how well ccRCC patients would fare and ofer new information about how to treat the disease [25]. Nevertheless, there is still a lack of knowledge about the mechanism of action of CR-lncRNA in ccRCC, particularly its infuence on prognosis. Tis study investigated the function of CR-lncRNA in ccRCC and developed a new prognostic model based on CR-lncRNA, which may ofer fresh perspectives for future studies on ccRCC and patient-specifc management.
Te proposed CR-lncRNA-based prognostic model includes the following advantages. (1) Due to the discrete Fourier transform data, its main information components are concentrated in the low-frequency part of the frequency domain, and the high-frequency part is mainly secondary information or noise. Terefore, the lengthening lncRNA sequence can be truncated into a fxed-length vector by intercepting the fxed-length part of the low frequency. (2) Two traditional convolutional models were used vgg16_bn build task models with Resnet18. Firstly, to adapt the data dimension, the commonly used two-dimensional convolution and pooling are adjusted to one-dimensional convolution and pooling. At the same time, since the label data are a twenty-four-dimensional data, the task model is extended to a multioutput model. LncRNA tissue-specifc analysis was performed on multiple output regression, multiple output classifcation, and multilabel classifcation, respectively.

Data Collection.
A recent signifcant study investigated the cuproptosis subtypes and built a predictive model to improve the prognosis of patients with CRC. Gene expression data were downloaded from the TCGA database to identify distinct molecular subtypes using a nonnegative matrix factorization algorithm [16]. Samples with a survival time of less than 30 days were disregarded as we downloaded the gene expression profle data, clinical information data, mutation data, and copy number variation (CNV) data of ccRCC from the TCGA ofcial website. Finally, 71 normal samples and 511 tumor samples were comprised. Te GEO database provided the CRGS gene expression profle in ccRCC. Gene count values were employed for diferential analysis, and for downstream analysis, count values were converted to log2 (TPM +1) values.

Analysis of Genetic Mutation Data of CRGS.
Te TCGA Kidney Renal Clear Cell Carcinoma (TCGA-KIRC) cohort was used to investigate the diferences in CRGS expression between normal and malignant samples. Tese discrepancies in gene expression were then re-examined in the GSE53757 and GSE40435 cohorts. We further confrmed them using the immunohistochemistry results of proteins in the Human Protein Atlas (HPA) database to assess their alterations in protein expression. Te location of these genes on diferent chromosomes was visualized using the "RCircos" package, the "maftools" package was used to plot the mutational landscape of these genes, and fnally, univariate COX regression analysis and log-rank test were performed to investigate the impact of these genes on the prognosis of ccRCC patients.

Consensus Clustering Analysis Based on CRGS.
Using the "ConsensusClusterPlus" R package, we performed unsupervised clustering of the ccRCC samples based on the expression patterns of the 19 CRGS. To ensure the stability of the clusters, 1000 random repeated samplings were carried out on 80% of the samples and all genes. Te Euclidean distance clustering algorithm was selected. Te appropriate number of clusters was established through using cumulative distribution function (CDF) and intra-group correlation. To confrm the discriminating of various subtypes, principal component analysis (PCA) was utilized. Te variations in survival among various subtypes were then shown using K-M survival curves, and the log-rank test was used to determine whether the diferences were statistically signifcant.

Identifcation of Molecular Characteristics, Immune
Infltration Characteristics, and Immunotherapeutic Response Based on Diferent Subtypes. Te "GSVA" package was used to study the pathways implicated in various subtypes through gene set variation analysis (GSVA). To further investigate the diferences in immune infltration features across distinct subtypes, the infltration abundance of diverse immune cells in diferent subtypes was estimated using the single sample genes enrichment analysis (ssGSEA) algorithm of the aforementioned R package. Te tumor immune dysfunction and exclusion (TIDE) approach, which was developed in recent years, can be used to anticipate whether immunotherapy will beneft tumor patients. Tis research thorough investigation of hundreds of distinct tumor expression profles looked for indicators to predict whether patients would respond to immune checkpoint blockade (ICB) therapy, i.e., a higher TIDE score indicates a lower likelihood of responding to immunotherapy [26]. Te website it created (https://tide.dfci.harvard.edu/) was subsequently used to forecast the immunotherapy response in patients from diferent subtypes. Te results of the analysis were visualized using the R packages "tinyarray," "pheatmap," and "ggplot2." Statistical signifcance was set at a P-value <0.05.

Diferential
Analysis of mRNA, lncRNA, and MicroRNA. Te three approaches of "Edger," "DESeq2," and "limma" were utilized to produce the overlapping mRNAs, lncRNAs, and microRNAs (miRNAs), which were then employed as the diferential genes. |Log2fold change| >1 and false discovery rate (FDR) <0.05 were the screening criteria thresholds for the approaches described above.

WGCNA Identifes Cuproptosis-Related Modules.
Te "WGCNA" R package was used to conduct the WGCNA analysis of the lncRNAs from the ccRCC samples. According to the scale-free network criteria, the best soft threshold was chosen. Te modules with distances of less than 0.25 were then combined, and the minimum number of genes for the modules was set at 30. Te modules having the strongest association with cuproptosis were chosen for further analysis after a correlation analysis between the modules and cuproptosis phenotypic data was completed.

Construction and Evaluation of Prognostic Risk Scoring
Models. Te R package "Caret" was used to frst randomly divide the ccRCC samples in the TCGA queue into training and test sets in a ratio of 7 : 3. Te training set and test set are used to train and test the model, respectively, to create a stable model. After intersecting the diferential lncRNA of the ccRCC with the lncRNA in the module most associated with cuproptosis discovered by the aforementioned WGCNA, a univariate Cox regression analysis and log-rank test were carried out. Subsequently, lncRNAs with a P value <0.05 obtained by both of the above two test methods were considered as candidate lncRNAs. We selected characteristic genes using two machine learning approaches, namely, Lasso regression and RF, to avoid the overftting of the model. A well-known machine learning technique called lasso regression decreases the dimensionality of highdimensional data by assigning each feature a penalty coefcient that makes the coefcient of unimportant features 0 and therefore eliminates collinearity across features. Is frequently employed to tune the COX proportional hazards model (CPH) [27]. Studies have proven that RF can also be used to model survival analyses, and a minimal depth (MD) strategy has also been developed to identify key prognostic characteristics, and recent studies have pointed out that treebased machine learning methods outperform deep learning in dealing with tabular data [28][29][30]. Te overlapping lncRNAs chosen by the two machine learning methods above were subjected to multivariate stepwise Cox regression analysis, and the best CPH model was found according to the Akaike information criterion (AIC) criteria, which states that the smaller the AIC value, the better the model's performance [31]. Te performance and precision of the model in the training set were assessed by using the timedependent ROC curve and the C-index, and the performance of the risk score as an independent prognostic indicator was confrmed utilizing univariate and multivariate Cox regression analysis. In order to more thoroughly assess the performance of our prognostic model, we gathered a number of lncRNA-based prognostic risk scoring models developed based on the TCGA-KIRC cohort in recent years, computed the time-dependent ROC curve and C index of the lncRNA model based on the entire TCGA-KIRC cohort, and compared them with our developed prognostic model [32][33][34][35][36][37].

Construction of Competing Endogenous RNA (ceRNA)
Network. We predicted the target miRNAs of these lncRNAs by applying the mircode website (https://mircode. org/download.php) based on the overlapping lncRNAs identifed by the aforementioned univariate Cox analysis and log-rank test. Te resulting target miRNAs were crossed with the diferential miRNAs and then submitted to miRDB (https://mirdb.org/), miRTarBase (https:// mirtarbase.cuhk.edu.cn/), and targets can (https://www. targetscan.org/). Te target mRNAs of the aforementioned miRNAs were predicted by the three websites in turn. Te fnal target mRNAs were chosen based on predictions made simultaneously by the three websites' predicted targets. We created a lncRNA-miRNA-mRNA ceRNA network in the Cytoscape based on the above predicted results.

Functional Enrichment Analysis.
Using the LNCSEA online platform (https://bio.liclab.net/LncSEA/), the functional enrichment analysis of the aforementioned prognostic CR-lncRNAs was carried out [38]. Functional enrichment annotation of CR-lncRNA target miRNAs was performed using the miEAA online tool (https://ccb-compute2.cs.unisaarland.de/mieaa2/) [39]. In order to investigate the probable biological pathways of patients in diferent risk groups, we simultaneously performed gene set enrichment analysis (GSEA) on patients in high-and low-risk groups using the R package "clusterProfler." Te adjusted Pvalue <0.05 and q value <0.05 were used to identify statistically signifcant enriched pathways.

Immunotherapy Response Prediction and Drug
Sensitivity Analysis. ICB therapy has now been proven to be benefcial for some tumor patients, although the majority of patients do not gain from it, which may be partially attributed to tumor heterogeneity and varied immune checkpoint expression. In order to determine whether patients might beneft from immunotherapy, we compared the immune checkpoint expression between high-and low-risk patients. We then used the TIDE website to estimate the immunotherapy response for ccRCC patients in diferent risk groups. Sensitive anticancer drugs were examined using the "pRRophetic" R package for two risk groups. Te Wilcoxon rank test was used to analyze diferences between various risk groups, and a P value lower than 0.05 was regarded as statistically signifcant.
2.11. Statistical Analysis. R software was used to perform all analyses (version 4.1.2). Te association analysis of two categorical variables and the sample rate (composition ratio) of two or more groups were both compared using the chisquare test. To determine whether there were any diferences in the distribution of measurement data or grade data between the two groups, a Wilcoxon rank-sum test was utilized. Te Kruskal-Wallis test was used for nonparametric comparisons when there were three or more groups. For correlation testing in the correlation analysis, Spearman, and distance correlation tests were employed. Te statistical signifcance was defned as a P value of 0.05, where * denotes P value <0.05, * * denotes P value <0.01 and * * * denotes P value <0.001, and ns denotes no statistical signifcance.

Landscape of Genetic Mutations in CRGS.
Te 19 CRGS were acquired through recent signifcant scientifc discoveries [16]. Following that, diferential analysis of the previously mentioned genes in the TCGA cohort revealed that, with the exception of LIPT1, LIPT2, and ATP7A, which revealed no statistically signifcant differences, the expression of the majority of CRGS difered signifcantly between normal and tumor samples and most of them were downregulated (Figure 1(a)). Next, the expression levels of these CRGS were checked again in the two GEO cohorts, GSE40435 and GSE53757, and although the results were slightly diferent from the TCGA cohort, the general results were similar (Figures 1(b) and 1(c)). It is common knowledge that proteins carry out the majority of biological processes in humans. To this end, we further assessed the variance in these genes protein expression in the HPA database between tumor and normal tissues. Te outcomes demonstrated that most genes expressed diferently at the protein level as well (Figure 1(f )). Te fndings from the K-M survival curve were similar to those from the univariate Cox analysis, with the exception that CDKN2A and GCSH were risk factors, whereas the remaining genes were protective (Figure 1(d)). Te somatic mutation rate of each CRGS was incredibly low, and just 23 (6.44%) of 357 ccRCC samples showed genetic alterations, according to our analysis of somatic mutations in these genes ( Figure 1(e)). Figure 1(f ) demonstrates that the majority of CRGS have low CNV frequencies. Te frequency of copy number deletions is almost 9% for only PDHB. On the chromosome, CRGS is located, as shown in Figure 1(c). We hypothesized that the genetic variation in ccRCC is largely stable because both somatic mutations and CNV frequencies had very small sample sizes. Of course, additional elements such as methylation and histone modifcations might also be at work. According to the aforementioned fndings, CRGS has a signifcant impact on the prognosis of ccRCC patients as well as the occurrence and progression of cancer.

Identifcation of Molecular Subtypes Based on CRGS.
We used the "ConsensusClusterPlus" R package, a consensus clustering method based on a machine learning algorithm, to perform unsupervised clustering of ccRCC patients-based on the expression levels of the 19 CRGS. Finally, as shown in Figures 2(a) and 2(b), we were able to distinguish the cuproptosis molecules into two optimum clusters, A and B, each of which had 335 and 176 samples, respectively. Based on the abovementioned results, we can infer that patients in clusters A and B refect two distinct cuproptosis phenotypes, with cluster A presenting the activating subtype of cuproptosis and cluster B representing the suppressing subtype. Te PCA results demonstrated good discrimination between the two distinct subtypes (Figure 2(c)). A subsequent study of survival analysis revealed that patients in cluster A had signifcantly higher overall survival (OS) than those in cluster B (Figure 2(d)). For the two subtypes, GSVA analysis identifed separate underlying biological processes and pathways (Figure 2(e)). Te pathways DNA repair, Myc targets, Reactive Oxygen Species pathway, and Kras Signaling pathway, which are typically linked to tumor development and tumor immune infammation, were signifcantly enriched in patients in cluster B compared with patients in cluster A. Terefore, the reason that cluster B patients have a poor prognosis may be due to the activation of the aforementioned pathways. However, the spermatogenesis, pancreas beta cells, heme metabolism, and androgen response of patients in cluster A were signifcantly enriched. In light of the variations in the biological pathways mentioned above, we investigated the immune infltration traits of the two subtypes. As can be seen in Figure 2(f ), cluster A had a larger concentration of infltrating neutrophil, mast, and eosinophil cells, whereas cluster B had a higher concentration of infltrating activated CD8 T cells, CD4 T cells, activated B cells, and myeloidderived suppressor cells (MDSC) cells. Ten, using the TIDE website, we predicted whether certain patient subgroups would respond to immunotherapy. According to Figure 2(g), patients in cluster A had lower TIDE scores, making them more likely to beneft from immunotherapy. Figure 2(h) compares the immunotherapy responses of diferent patient subgroups (cluster A, 89% vs. cluster B, 72%). Our fndings imply that therapeutic regimens developed for cuproptosis may be a potential anticancer target in ccRCC patients and may improve ccRCC patients' responsiveness to immunotherapy. TCGA **** **** *** ns **** **** * n s ns **** **** **** **** ** **** **** **** ** **** 0.0 Tumor ns **** ns ns **** **** **** ns **** **** **** **** **** ns **** **** ns **** Tumor GSE53757 ns **** ns ns **** **** ** *** **** **** **** **** * **** **** **** ****  Journal of Oncology

Identifcation of CR-lncRNAs.
As master regulators of gene expression, lncRNAs have been implicated in a number of malignancies in recent years. A notable illustration is PCAT-1 dysregulation, which is strongly linked to the development of prostate cancer [40]. Additionally, lncRNAs can be employed independently to forecast tumor prognosis, tumor progression, and disease diagnosis [41,42]. Terefore, we retrieved the lncRNA expression profles of ccRCC patients from the TCGA database and, after deleting the lncRNAs that were barely expressed, acquired 9024 lncRNAs for WGCNA analysis. Te WGCNA network was built using the one-step method, and Figure 3(a) shows that there were no outlier samples discovered and that the samples were well clustered. Te ideal soft threshold of 3 was identifed using the scale-free topology ftting index of 0.85 and network connectivity as the standard (Figure 3(b)). A hierarchical clustering dendrogram that obtained 10 modules is shown in Figure 3(c). As can be seen from Figure 3(d), the blue, green, and magenta modules are all signifcantly associated with tumor and cuproptosis, but the blue module has the strongest correlation with tumor (r � −0.5, P < 0.001). As a result, we chose the lncRNAs identifed in the blue modules to further develop the prognostic molecular characteristics of ccRCC patients.

Construction and Validation of Prognostic Risk Scoring
Model. We eventually discovered 4229 overlapping mRNAs, 2287 overlapping lncRNAs, and 181 overlapping miRNAs using the three methods of "EdgeR," "DESeq2," and "limma" for gene diferential analysis. Te diferential lncRNAs were intersected with the 1033 lncRNAs in the blue module above to provide 630 overlapping lncRNAs as candidate lncRNAs. Subsequently, univariate Cox regression analysis and the log-rank test yielded 116 lncRNAs with prognostic signifcance (P value <0.05). Figure 4(a) displays the optimum parameter (λ) interval for Lasso regression using 10-fold cross-validation. When we selected the λ value with the smallest mean error, we got 33 lncRNAs (Figure 4(b)). Te relationship between the number of trees and the error rate in the RF algorithm is illustrated in Figure 4(c), along with the characteristic genes the algorithm identifed. It is clear that as the tree expands, the error rate curve gradually fattens out, showing that the number of trees chosen was sound. Te MD approach yielded a threshold of 7.9681, and using this threshold, we were able to derive 44 signifcant eigengenes (Figure 4(d)). By intersecting the lncRNAs produced by the previous two approaches, we identifed 23 potential lncRNAs (Figure 4(e)). Based on the aforementioned potential lncRNAs, a multivariate stepwise CPH model was created, and with an AIC � 1234.71, we were able to generate the ideal CPH model for 13 lncRNA combinations in the training set ( Figure 4(f )). Te expression of lncRNA in the aforementioned model and the regression coefcient obtained by multivariate stepwise Cox regression analysis were used to generate the risk score for each patient. Te following is the calculating formula: risk score � (−0.3586 * AC007637. . Based on the median risk score, patients were separated into high-and low-scoring groups. Te risk score's area under the curve (AUC) at one year, three years, and fve years is 0.800, 0.793, and 0.819, respectively, according to the ROC curve of the training set ( Figure 5(a)). Te ROC curve of the test set also displays greater accuracy, with AUCs exceeding 0.75 at one year, three years, and fve years ( Figure 5(d)). Te C index, which was 0.77 in the training set ( Figure 5(b)) and 0.71 in the validation set ( Figure 5(e)), both of which were considerably higher than the remaining clinicopathological variables, also showed that the model had great consistency. In the training * * **** ns * **** *** ns ns **** **** ns **** ns ns ns **** *** **  (g) diferences in TIDE scores among diferent cuproptosis clusters; and (h) diferences in immunotherapy response among diferent cuproptosis clusters. * P value <0.05, * * P value <0.01, * * * P value <0.001, * * * * P value <0.0001. and test sets, the OS of patients in the high-score grouping was considerably lower than that in the low-risk category (Figures 5(c) and 5(f), P value <0.001). Additionally, the prognostic risk model we created performs better than some current models when comparing the two indicators of AUC and the C index ( Figure 5(g)). Te results above show that the prognostic risk score model based on 13 CR-lncRNAs can precisely predict the prognosis of ccRCC patients.

Correlation of Prognostic Risk Scoring Models with Clinical
Pathological Features. Te association between risk scores and clinicopathological characteristics was also demonstrated by our investigation. As observed in Figure 6(a), grade and stage vary among various risk groups even if risk scores are really not related to age and gender. Furthermore, a greater risk score was signifcantly correlated with both a higher grade and stage (Figure 6(b)). Likewise, the outcomes of patients in the high-risk group were considerably worse than those in the low-risk group in all clinical subgroups, according to the fndings of the subsequent survival analysis (Figure 6(c)). Te risk score was also revealed to be an independent prognostic factor in ccRCC patients by univariate and multivariate Cox regression analysis (Figures 6(d) and 6(e)). In light of the aforementioned fndings, the prognostic risk score model, which is made up of 13 CR-lncRNAs, is a very promising biomarker that can not only accurately predict the prognosis of ccRCC patients but also assess their clinical progression.

Correlation of Prognostic Models with Tumor Immune Microenvironment and Immunotherapy Responses.
Te ssGSEA analysis suggests that the immune infltration features of the patients in the two risk groups varied. While patients in the low-risk group had higher rates of neutrophil, immature dendritic cell, and mast cell infltration, patients in the high-risk group had higher rates of activated CD4 T cell, activated CD8 T cell, and MDSC infltration (Figure 7(a)). It was further revealed by PCA analysis that the two patient groups represented various immune cell infltration microenvironments (Figure 7(b)). Te majority of immunological checkpoints were more strongly expressed in the high-risk group, whereas PD-L1 and PD-L2 expression were  Journal of Oncology more prominent in the low-risk group (Figure 7(c)). According to the current understanding, immunotherapy has a greater chance of helping tumor patients the more PD-L1 is expressed. Additionally, because the majority of immune checkpoints are signifcantly expressed in the highrisk group, it is more likely to produce immunosuppression, which will cause the cancer to advance in those people. We next used the TIDE online tool to predict immunotherapy responses for patients in the two groups once more. Te fndings demonstrated that the low-risk group had lower TIDE scores than the high-risk group (Figure 7(d)). Moreover, Figures 7(e) and 7(f ) show that better immunotherapy outcomes are signifcantly correlated with lower risk scores. Terefore, we can conclude that immunotherapy is more likely to be benefcial for patients in the low-risk group. Together, the prognostic risk score model may be helpful in identifying patients' TIME and forecasting their response to immunotherapy.

Construction of ceRNA Networks.
It is generally known that miRNA can infuence mRNA expression via binding to mRNA. As a ceRNA, lncRNA can also control the expression of mRNA by competitively binding to miRNA, infuencing the occurrence and progression of cancer. To learn more about the regulatory role of CR-lncRNA at the gene level, we frstly predicted the target miRNAs of the aforementioned prognostic CR-lncRNAs using the website miRcode, yielding 23 diferential miRNAs (Figure 8(a)). Te target miRNAs identifed above were then used to predict the target mRNAs via the miRDB, miRTarBase, and TargetScan websites, and a total of 174 diferentially overlapping  mRNAs were discovered (Figure 8(b)). We created a lncRNA-miRNA-mRNA ceRNA network based on the results above (Figure 8(c)).

Functional Enrichment Annotation.
We discovered that the aforementioned prognostic CR-lncRNAs were considerably enriched in cell proliferation, metastasis, stemness, and EMT as well as being signifcantly related with a variety of immune cells by enrichment analysis (Figures 9(a) and  9(b)). Te miRNA enrichment analysis revealed that the aforementioned miRNAs were signifcantly enriched in pathways involved in the development of cancer and immune infammation, including the p53 signaling pathway, the JAK-STAT signaling pathway, the expression of PD-L1, the PD-1 checkpoint pathway in cancer, the chemokine signaling pathway, and other pathways (Figure 9(c)). Te underlying biological pathways in patients in the high-risk and low-risk groups were then further investigated using GSEA analysis. According to the fndings, the IL-6/JAK/ STAT3 signaling, E2f targets, and epithelial-mesenchymal transition (EMT) pathways were considerably enriched in  the high-risk group. Patients in the low-risk group, on the contrary side, had signifcantly higher levels of pathways such oxidative ylation, protein metabolism, fat acid metabolism, and androgen response (Figure 9(d)).

Sensitivity Analysis of Antitumor Drugs.
With the use of the "pRRophetic" R package, we acquired 6 potentially sensitive medications to help further direct the individualized treatment of ccRCC patients. Results showed that in the high-risk group, acadesine (AICAR), all-trans retinoic acid (ATRA), palbociclib (PD-0332991), and cisplatin were more sensitive, whereas in the low-risk group, GSK1904529A and KIN001102 were more sensitive ( Figure 10).

Discussion
Using the TCGA and GEO datasets, this study investigated the expression diferences of CRGS at the gene level between normal tissue and tumor samples, and further confrmed the ns **** **** * ns ** **** ns ns ns **** ns ns **** ns **** **** * ns ns ns **** *   Figure 7: Correlation of prognostic risk score models with immune infltration and response to immunotherapy. (a) Diferences in immune cell infltration between patients in high and low-risk groups; (b) PCA analysis reveals distinct immune microenvironments between diferent risk groups; (c) diferences in the expression of immune checkpoints between high and low-risk groups; (d) diferences in TIDE scores between high and low-risk groups; (e-f ) diferences in immunotherapy response between high and low-risk groups. * P value <0.05, * * P value <0.01, * * * P value <0.001, * * * * P value <0.0001. expression variations of CRGS at the protein level in the HPA datasets. In ccRCC, the majority of CRGS were lowly expressed, and a survival study afterward indicated that most CRGS were protective genes in ccRCC patients. In addition, the genetic mutation data analysis confrmed that the genetic mutation of the above genes in ccRCC is relatively rare. Te two subtypes of cuproptosis clusters were then established by consensus clustering-based on the expression of 19 CRGS, and further analysis proved that the subtype with high CRGS expression was substantially associated with higher survival. Tese fndings imply that cuproptosis might be a therapeutic target for people with ccRCC. It is interesting to note that there were signifcant diferences between the TIME of the two subtypes, with the subtype considerably downregulated in CRGS having a larger abundance of cytotoxic T lymphocytes (CTLs) infltration but also more MDSC infltration. It is well recognized that MDSC infuence immunosuppressive tolerance through a variety of methods as signifcant elements of the milieu that suppresses the immune response to cancer. Numerous studies have proven that MDSC, in particular, suppress the T-cell immunological response by creating a lot of reactive 14 Journal of Oncology oxygen species (ROS) [43][44][45]. Also, MDSCs have been linked to a number of tumor-related events, including angiogenesis, treatment resistance, and metastasis [46]. Tis may also explain why cluster B subtypes have lower survival rates and higher CTLs infltration. Notably, subsequent GSVA analysis also supported the fnding that patients with the cluster B subtype had substantial ROS PATHWAY enrichment. Te outcomes of the TIDE online tool also revealed that patients with the cluster B subtype responded to immunotherapy less favorably. Te statistics shown above clearly demonstrate that cuproptosis is highly related to the prognosis and immunotherapy of patients with ccRCC, opening up new research directions. LncRNA, which acts as master regulator of gene expression, has been linked to a number of cancers and can be used independently to predict a patient's prognosis and make a diagnosis of the disease [40][41][42]. By using WGCNA, we were able to recognize CR-lncRNA. Subsequently, prognostic characteristic genes were further screened using univariate Cox regression analysis, log-rank test, LASSO regression, and RF. Finally, using multiple stepwise Cox regression, an optimal prognostic risk score model made up of 13 CR-lncRNAs was constructed. Te model has strong predictive performance and consistency, as indicated by the ROC curve and the C index. Furthermore, it was discovered that the CR-lncRNA-based prognosis models developed using WGCNA and various machine learning algorithms were typically superior to some current models when compared to some lncRNA-based prognostic models developed in the TCGA-KIRC cohort in recent years.
Besides that, we investigated the relationship between predictive risk scores and clinicopathological characteristics and discovered that there was a substantial relationship between risk scores and clinicopathological variables in ccRCC. Furthermore, studies showed a signifcant positive correlation between the risk score and the tumor's aggressiveness, with the greater the risk score, the higher the tumor grade and stage. Subsequent analysis of immune checkpoint expression and immune infltration analysis confrmed that, except for PD-L1 and PD-L2, the remaining immune checkpoints were more highly expressed in the high-risk group, and the infltration abundance of MDSC was also higher. Tis demonstrates that patients with higher risk scores are more likely to produce an immunosuppressive microenvironment, enabling tumor cells to elude the immune system's surveillance and promoting the growth and development of malignancies. Furthermore, evidence that patients with greater risk scores had a worse response to immunotherapy came from the TIDE study. We conducted the GSEA analysis to investigate the mechanism underlying this diference. Pathways including EMT and IL6 Jak Stat3 Signaling were discovered to be considerably enriched in the high-risk group. Studies have already shown that activating the EMT pathway can promote tumor cell infltration, tumor migration, and metastasis. It can also cause the formation of an immunosuppressive microenvironment, which helps  Processing and Presentation  T cell_MCPCOUNTER  T cell follicular helper_CIBERSORT-ABS  T cell CD8+_CIBERSORT-ABS  T cell CD4+ memory_XCELL  T cell CD4+ T1_XCELL  Neutrophil_CIBERSORT  NK cell activated_CIBERSORT  Monocyte_XCELL  Mast cell activated_CIBERSORT  Macrophage M2_CIBERSORT  Granulocyte-monocyte  tumor cells to escape the immune system [47,48]. Meanwhile, the IL-6/JAK/ STAT3 Signaling pathway is overactivated in many forms of cancer, and it is implicated in driving cancer cell proliferation, invasion, and metastasis, as well as interacting with TIME to inhibit antitumor immune responses [49]. Moreover, it has been shown that the Stat3 transcription factor in the Stat3 signaling pathway can increase the expression of S100A8 and S100A9, preventing dendritic cell (DC) diferentiation and stimulating the accumulation of MDSC, which in turn mediates the  immunosuppressive efects [50]. According to the results above, there may be a connection between the activation of the aforementioned pathways and the diferences in TIME and immunotherapy responses amongst diferent risk subgroups.
We developed a lncRNA-miRNA-mRNA ceRNA network to more thoroughly elucidate the regulatory role of CR-lncRNA at the gene level. Following enrichment analysis, it was discovered that the aforementioned lncRNAs and miRNAs were strongly linked to tumor development, metastasis, prognosis, cell proliferation, and TIME. AICAR, ATRA, PD-0332991, Cisplatin, GSK1904529A, and KIN001-102 were among the six possible anticancer medications that were tested using drug sensitivity analysis. And research has shown that ATRA can enhance the survival of tumor-specifc CD8 T cells and upregulated MHC I expression in tumor cells to function as antitumor immunity [51][52][53]. Additionally, it can also promote MDSC diferentiation and maturation, which in turn lowers their population, triggering the immune system to inhibit tumor growth [54]. A highly selective CDK4/6 inhibitor known as PD-0332991 has been shown to have antiproliferative efects in a variety of malignancies, including renal cell carcinoma and liver cancer [55,56].
Tis study has some relative merits overall. First of, the CR-lncRNA-based prognostic model created by the WGCNA and several machine learning algorithms can successfully predict the prognosis of ccRCC patients. It ofers greater prediction performance and consistency when compared to several other lncRNA-based models already in use. Signifcant relationships between the model, TIME, and immunotherapy were also discovered in the fnal research. Tere are, however, some restrictions-based on bioinformatics analysis, and multicenter prospective studies are still required for validation in the latter phase, which is also the main objective of our future research work.

Conclusion
We explore the potential function of CRGS in ccRCC after a thorough investigation. Based on CR-lncRNA, a model for prognostic risk scoring was developed. Tis model can distinguish TIME, predict the efectiveness of immunotherapy, and provide great and independent prognostic performance in ccRCC patients, allowing for more personalized treatment. For upcoming ccRCC research, it ofers fresh perspectives and ideas.

Data Availability
Te datasets used to support the fndings of this study are publicly available in the GEO database (https://www.ncbi. nlm.nih.gov/geo/) and TCGA database (https://portal.gdc. cancer.gov/).

Conflicts of Interest
Te authors declare that they have no conficts of interest.