CD247 Functions as a Prognostic Biomarker for Cutaneous Malignant Melanoma Based on the Analysis of Tumor-immune Microenvironment

Background: Cutaneous malignant melanoma (CMM) is among the most lethal cancers. The tumour microenvironment (TME) is closely linked with tumorigenesis, metastasis, and prognosis. Methods: We employed the ESTIMATE algorithm to calculate immune and stromal scores of malignant melanoma tissues from the Cancer Genome Atlas dataset, respectively, and determine core prognosis gene signature examined by COX proportional hazards model. Functional enrichment annotation, the Kyoto Encyclopedia of Genes and Genomes pathways, the Protein-Protein Interaction network, Weighted Gene Co-expression Network Analysis and overall survival analysis were used to formulate potential function of these genes that involved in immune-linked biological processes. CIBERSORT algorithm was used to estimate the abundances of immune cell types in CMM samples. Finally, the OncoLnc platform, the Gene Expression Proling Interactive Analysis resources, and the Human Protein Atlas database were applied to validate our results. Results (cid:0) 908 differential expressed genes and ten hub genes were screened, and GO annotation indicated that immune response and inammatory response were rmly involved in CMM tumorigenesis and progression. CD247, identied as the most signicant prognostic biomarker, highly expressed in tumor samples and possessed a better prognosis than low expressed samples. The correlation analysis of immune cells inltration unveiled that CD8 + T cell and Macrophages were intense signicant to CMM patients' prognosis. Survival analysis suggested that ten hub genes and inltrated immune cells are linked to the prognosis of CMM. ConnectiveMap analysis strongly indicated that L-securinine may be a promising candidate medicine for CMM patients. Conclusions: we deeply analyzed the immune-linked genes with the tumour microenvironment, and labeled CD247 as the most intriguing prognostic biomarker for CMM, which may bring better clinical outcomes for CMM patients.


Background
Cutaneous malignant melanoma (CMM), one of the most aggressive cancer with a reduced survival rate, features the primary cause of skin cancer-related death, and is notorious for its resist-therapy [1]. In the past several years, immunotherapeutic strategies have made signi cant progress and did improve patient survival time. However, owing to CMM's easy-to-recurrence and dreadful metastasis, amounts of patients adapting such treatments have no obvious durable response [2,3]. Besides, some adverse effects have emerged, such as autoimmunity, since the complexity of interaction between tumour cells and the tumour microenvironment (TME). TME comprises immune cells, endothelial cells, mesenchymal cells, in ammatory factors, and extracellular matrix (ECM) molecules [4]. The cells and particles in the TME are in a dynamic process and supported to be highly linked with various tumour behaviours, including proliferation, metastasis, and immune escape. Immune cells and stromal cells are the two main types of non-tumour components and considered to be of signi cant importance in the diagnosis and prognosis of tumours [5]. An increasing amount of evidence has elucidated the clinic pathological signi cance of TME in the prediction of treatment effects [6]. Therefore, understanding the molecular composition and function of TME has more effective regulation of cancer progression and immune response in CMM.
Extensive effort has been made on the mechanism of the occurrence and progression of CMM. However, the concrete pathogenesis of CMM still needs to be further formulated. Therefore, it is an urgent task to explore freshly molecular markers that represent a diagnostic and prognostic value for CMM. In this study, we comprehensively investigated the TME and immune cells associated with CMM to determine immune-related prognostic molecules for CMM. Finally, the HPA database, the OncoLnc platform, and the GEPIA tool were applied to validate the study outcomes. The whole experiment procedure was shown in gure 1.

| Raw data collection
The RNA-seq expression pro ling for CMM patients was acquired from the TCGA database (https://tcgadata.nci.nih.gov/tcga/gdc), including 32 normal samples and 436 tumour samples [7]. The work ow type was selected HTSeq-FPKM for further investigation, which including 472 les and 468 cases. Clinical data involving each patients' age, TNM staging, gender, tumour grade, and survival information were also downloaded from the TCGA website. The gene expression quanti cation was measured using the Illumina HiSeq2000 RNA Sequencing platform. Then the ESTIMATE algorithm (http://r-forge.r-project.org) was applied to calculate each samples' immune and stromal scores, respectively, which was performed by R software (4.0.2) with the help of relevant R packages: "estimate," "limma" and "utils." To validate our results [8], we explored the OncoLnc website (http://www.oncolnc.org/), GEPIA platform (http://gepia.cancer-pku.cn/), and the HPA (https://www.proteinatlas.org/) to verify the screened prognostic-genes.

| Differential analysis of expressed genes
Initially, all samples were separated into high and low immune-score groups and high and low stromalscore groups based on the outcomes of the ESTIMATE analysis. Then, the R package "Limma" was used to identify the differential expression genes (DEGs) between high and low immune-score groups, as well as high and low stromal-score groups [9]. We set the screening conditions as the DEGs were: log 2 |fold change| > 1.5 and adjust p-values <0.05. Heatmaps of DEGs were drawn via the R package "pheatmap."
Then, a power function amn = cmn β (cmn = Pearson's correlation between gene m and n; amn = adjacency between gene m and n) was selected to erect weighted adjacency matrix. We transformed the adjacency to the topological overlap matrix (TOM), which tests the network connectivity of genes. Then, we developed average linkage hierarchical clustering in line with the TOM-based dissimilarity measure with a minimum size of 50 for the genes dendrogram. Therefore, genes that possessed similar expression pro les might be classi ed into the same module, which de ned as groups of genes with a high degree of correlation. Then a single-column vector called the module eigengenes were built to elucidate the potential relationship between the gene modules and clinicopathological parameters. The module eigengenes were representative of the gene expression pro les in each module, which was produced by preserving the rst principal component of a designated module.
Moreover, the module eigengenes represent a summary metric for the co-expression genes network for each module eigengene contains most of the variance in the raw data. The consistency between the module eigengenes expression and the expression of genes was labelled as the module membership. We calculated the dissimilarity of module eigengenes, selecting a cut line for module dendrogram to merge them. Functional analysis was performed based on the modules, which have a notable in uence on CMM.

| Functional Enrichment annotation and PPI network
Functional enrichment analysis was performed by using the R packages: "clusterPro ler," "enrichplot," "org.Hs.eg.db," " ggplot2" in R software to annotation, visualization and integrated outcomes of the DEGs, the corresponding biological processes (BP), cell components (CC), and molecular functions (MF) were determined via Gene Ontology (GO) [11], and the most enriched signalling pathways were identi ed through the Kyoto Encyclopedia of Genes and Genomes (KEGG) [12]. A false discovery rate < 0.05 was considered as the cut-off. The protein-protein interaction (PPI) network was constructed using the Search Tool for the Retrieval of Interacting Genes (STRING) database (https://string-db.org/cgi/input.pl), then, the top ten hub genes were identi ed using the CytoHubba, plug-in of Cytoscape (3.8.0) software. Module analysis for the detection of interaction networks was performed using the MCODE plug-in of Cytoscape platform [13].

| Clinical correlation analysis of CD247
For determining the core prognostic gene, the R package "VennDiagram" was utilized. We combined the DEGs obtained from PPI network and prognostic relevant genes acquired from the analysis of the COX regression model. Then, CD247 represented the most signi cant prognostic gene. Next, a series of R packages: "limma," "beeswarm," "ggpubr," "survival," and "survminer" were employed in R software to investigate the clinical correlation of CD247. More critical, the GEPIA tool (http://gepia.cancer-pku.cn/) was applied to verify the prognostic value of CD247 in other CMM cases.

| Immune cells in ltration analysis
To better reveal that the underlying relationship between in ltrating immune cells, including CD8 + T cell, CD4 + T cell, B cells, eosinophils, neutrophils, macrophages, and dendritic cell in CMM, and prognosis in CMM patients, R software was applied to analyze the clinical correlation of in ltrating immune cells with the help of R packages: "limma," "preprocesscore," "corrplot" and "vioplot." CIBERSORT algorithm, an analytical tool, to provide an estimation of the abundances of immune cell types in a mixed cell population performed by disease gene expression matrix [14].

| ConnectiveMap (CMap) and overall survival analysis
CMap (https://portals.broadinstitute.org/cmap/) is an open database that association with disease, genes, and drugs or small compounds based on the gene expression pro le. Therefore, CMap analysis was used to identify potential drugs or small compounds to mitigate cutaneous malignant melanoma. We regarded Mean < −0.4 and p-value < 0.05 as the screening standard. The survival analysis was shown using the Kaplan-Meier curve with a log-rank test, which is conducted by the R packages "survival" and "survminer." The survival curve illustrates the association between prognostic-genes and overall survival time.

| Statistical methods
All data were demonstrated as mean ± SD. One-way analysis of variance was used to compare the stromal and immune scores in assigned groups by GraphPad Prism 8.0 software, and a two-tailed p < 0.05 was regarded as signi cant.

| Stromal and immune scores are highly linked with the TNM staging system and prognosis
The RNA-seq pro ling and clinicopathologic information about 468 patients with CMM were downloaded from the TCGA database. And then, we processed and analyzed the samples based on the work ow (see Fig. 1). Among all samples, the age of individuals was 62 8. 95.1% were white (n=445), 2.6% Asian (n=12), 2.3% were not reported (n = 10) and others were black or African-American. The proportion of patients with T1-T2 and T3-T4 was 37.6% (n = 176) and 62.4% (n = 292), respectively. Patients with N0-N1 and N2-N3 stand at 66.2% (n = 310) and 33.8% (n = 158), respectively. Patients with M0 and M1 reached 87.2% (n = 408) and 12.8% (n = 60), respectively (see additional le 3). Next, the ESTIMATE algorithm was applied to calculate stromal scores and immune scores of all samples, and the score was ranged from − 1806.847 to 1891.958 and − 1481.104 to 3769.121, respectively (see additional le 1). As ± for the N staging and M staging of CMM patients, both immune score and stomal score reveal no signi cant association between N/M staging and immune/stromal scores (Fig. 2e, 2f, and additional le 2). In terms of T staging, the order for stromal median score is: T4 < T3 < T2 < T1, and order for immune median score represent: T4 < T2 < T3 <T1. There is a signi cant difference between them (Fig. 2d). To unveil the underlying relationship between the stromal/immune/estimate scores and the overall survival (OS) of the CMM patients, we divided samples into high and low score groups according to the high/low stromal/ immune/estimate scores. Then, the Kaplan-Meier survival curve showed that the high score group of the immune scores has a higher survival rate than the low score group (Fig. 2a, p<0.001). Similar outcomes were observed in the high and low score groups of the immune/estimate scores (Fig. 2b, p = 0.076, Fig. 2c, p < 0.001).

| Analysis of Differential expression genes (DEGs) from CMM cases based on stromal and immune scores
In our current study, differential analysis of all RNA-seq data from 468 CMM cases was performed to understand better the relationship between the stromal/immune scores and the gene expression pro le of the samples (see additional le 4 and 5). Then, according to the distribution of the stromal/immune scores, a heatmap was drawn to reveal DEGs pro les of the samples, in which 1352 upregulated genes and 36 down-regulated genes were acquired based on the difference in stromal scores (log 2 FC > |1.5|, padj < 0.05). Similarly, 1042 upregulated genes and 94 down-regulated genes were obtained based on the differential analysis of immune scores (log 2 FC > |1.5|, p-adj < 0.05) (Fig. 3a). Notably, immune-related genes were explored via the intersection genes that were combined the upregulated or downregulated in both immune and stromal groups ( Fig. 3b. 900 upregulated genes and eight downregulated genes, see additional le 6).

| Weighted gene co-expression network analysis
The R package "WGCNA" was loaded in R software and set the power β to a soft-thresholding parameter β = 6 (scale-free R 2 = 0.87). MEDissThres was de ned as 0.25 to acquisition similar modules, and then 12 modules were generated. The black module included 221 genes, and the blue module included 1935 genes. While the brown module contained 2365 genes, and the green module contained 675 genes. Grey module represented 721 genes. The red module included 239 genes. The turquoise module included 1989 genes. Greenyellow module included 138 genes. The magenta module included 160 genes. The pink module included 220 genes. The purple module included 148 genes. Genes in the yellow module did not belong to any of the functional modules, and we neglected it. Besides, an intramodular analysis of Gene signi cance (GM) and module membership (MM) of the genes in the 11 modules involved was performed. Since GS and MM illustrated meaningful correlation, the result implied that the brown module was strongly related to CMM among the 11 modules. Moreover, genes belong to module brown had the highest positive correlation with CMM (see additional le 7). In contrast, genes in the red module had the highest negative ties with CMM (see additional le 8).

| Functional Enrichment annotation and PPI network analysis of DEGs
We predicted the function of the intersection 908 differential expression genes (900 upregulated and eight down-regulated). Go annotation (GO), and KEGG pathway analysis was performed by R software, including biological process (BP), molecule function (MF), cellular component (CC), and enriched signalling pathways [15]. Sorting by adjusting the p-value, we select the top 10 terms of each section (see Table 1). The results unveiled that GO functions are mainly enriched in in ammatory and immune responses, T cell activation, regulation of lymphocyte activation, and regulation of T cell activation (Fig. 5a, 5c). In comparison, KEGG pathways are mainly enriched in Cytokine-cytokine receptor interaction and Hematopoietic cell lineage (Fig. 5b, Table 2). To analyze the association between genes with prognostic value, the STRING network tool was used to construct a PPI network of these genes ( Fig. 6a, 6b). Hub genes analysis of the PPI network was performed via the CytoHubba, a plug-in in Cytoscape, the core gene scores were calculated using the Maximal Clique Centrality (MCC) method [16]. Then, the top 10 key genes were shown as follows: CCL5, CXCL10, CD74, CXCL9, CD247, IL10, CXCL11, CXCR3, CD3E and SYK ( Fig. 6c, Table 2). The module contains 26 nodes and 67 edges. Cytoscape's MCODE was used to conduct a comprehensive modular analysis of the differential genes (Fig. 6d). Three modules based on scores were identi ed through MCODE analysis, and expanded nodes and edges were not displayed. (screen criteria: Degree cut-off: 2, Node score cut-off: 0.4, K-core: 2). The module one contains 8 points and 16 edges with a 4571 score, which including CCL5, CCL3, CCR5, CCL4, CXCR3, CXCL9, CXCL10, CXCL11. In module two, with ve nodes and ten edges marked 4500 scores, which containing CYBB, NCF1, RAC2, NCF2, NCF4. In module three, with six nodes and nine edges marked 4000 scores, which possessing CD3E, LCK, CD3G, ZAP70, CD3D, and CD28 (Fig. 6). Next, the above-talked genes identi ed by PPI key modules functional analysis were investigated using GO annotation and KEGG pathway in R software and visualized by ClueGo, a plug-in in Cytoscape (Kappa score: 0.4, Go tree interval: min-level:4, max-level:8, and p-value < 0.05). The outcomes demonstrated that these genes were largely involved in the immune and in ammatory response, regulation of lymphocyte activation, leukocyte proliferation and play a critical role in Cytokine-cytokine receptor interaction, Chemokine signalling pathway and cell adhesion molecules (CAMs) (Fig. 6e).  CXCL9 binding to a subset of G protein-coupled receptors and play a crucial role to induce chemotaxis, promote differentiation and multiplication of leukocytes, as well as cause tissue extravasation.

CD247
Cluster of differentiation 247 CD247 activated the T cell receptors (TCR) signalling cascade and promote assembling of the TCR/CD3 complex on the surface of T lymphocytes.
For a thoroughly investigated the relationship between identi ed prognostic-involved genes and overall survival in CMM patients, a univariate Cox model was employed to select prognostic-related genes highly correlated with overall survival (see Table 3, Fig. 7d). Next, we combined the prognostic-related genes with the immune-related DEGs from the PPI network, and the result showed that CD247 was the intersection gene. Therefore, CD247 was regarded as the critical prognostic-related gene for further study. To begin with, we compared the expression of CD247 in normal cases (n = 32) vs tumour cases (n = 436), and the result showed that CD247 highly expressed in tumour samples (p = 0.011) and high expression of CD247 cases have a higher survival rate than low group (p < 0.001). The expression of CD247 in both female patients and male patients have no signi cant difference (p = 0.12) (Fig. 7a). Then, the analysis of TNM staging was performed to support the prognostic value of CD247 (Fig. 7b), from the perspective of tumour in ltration depth (T staging), the expression of CD247 from T1 vs T2 and T1 vs T3 to T1 vs T4 demonstrated a sharp signi cant difference. At the same time, there was no meaningful relationship between CD247 and N/M staging. (Fig. 7b). Also, we mined the GEPIA online platform and collected 1019 CMM cases, which covered 461 tumour samples (n = 461) and 558 normal samples (n = 558). Then, we inspected the expression of CD247 in each sample between tumour and normal tissues as well as performed survival curve to convince prognostic characterize of CD247. As is showed in Fig. 8c, CD247 is higher expression in tumour samples compared with normal samples (p < 0.05). Survival analysis revealed that the CD247 higher expression group possesses a better prognosis than the lower group ( Fig. 7c).

| The relationship between immune cells in ltration abundance and prognosis of CMM
The genes expression matrix and clinical information were applied to assess the abundance of immune in ltrating cells in each case based on CIBERSORT algorithm and to unveil the relationship between immune cells in ltration and CMM prognosis (Fig. 8a, 8c). The vioplot showed the distribution of immune cells. Among them, the highest proportion of immune cells was Macrophage M0, followed by CD8 + T cells (Fig. 8b). Then, the prognostic correlation analysis was performed for immune cell in ltration, and CD247 was drawn (Fig. 8d). The results showed that all of which were functioned as a critical role in CMM microenvironment and equipped with prognostic values for CMM patients' survival time, among them, MacrophagesM0, MacrophagesM2, rested CD4 memory T cell and Eosinophils had a negative correlation with CD247 while CD8 + T cell, MacrophagesM1, activated CD4 memory T cell, regulatory T cell, helper follicular T cell, and activated NK cell has a positive correlation with CD247 (see Table 5).

| The survival curve of prognostic-related hub genes and ConnectiveMap (CMap) analysis
the Kaplan-Meier curve was drawn to visualize survival analysis, which is performed using R packages: "survival" and "survminer" in R software. A log-rank test was to verify the association between the ten key genes and the overall survival of CMM cases. As is demonstrated in Fig. 9. Exploring the CMap platform, we identi ed drugs or small compounds that may help ameliorate CMM. Mean < − 0.4 and p < 0.05 as the screening condition (see Table 4).  The OncoLnc platform (http://www.oncolnc.org/) was used to verify whether the ten prognostic related genes in the TCGA database also crucial for other CMM cases. We investigated 458 samples (n = 458) and applied the Kaplan-Meier curve to analyze survival time for CMM patients (Fig. 10). The results revealed that the higher expression of these genes may be potential indicator genes for a good prognosis and may offer new insight into the therapies for CMM. To further con rm the reliability of the prognosticresponse genes, IHC evidence from the HPA database was employed to display the expression of these genes (CD74, CD247, IL10, CXCL11, CXCR3, SYK) in normal tissues and tumour tissues from CMM at the protein level. The results showed that compared with normal tissues, these genes were signi cantly overexpressed in tumour tissues (Fig. 11).

Discussion
Cutaneous malignant melanoma (CMM) is a skin-related devastating cancer characterized by complicated internetwork, and many genetic factors participate in its progression [1,5]. For years, the traditional therapeutic methods, such as immunotherapies, chemotherapies, and targeted-therapies, have improved patients' survival to some extent [17]. However, early-stage diagnostic markers and effective therapies for CMM patients are unavailable at present. Indeed, many patients in clinical suffered drug resistance and allergic by adapting these therapies and result in poor prognosis [3]. Therefore, more patient-friendly therapeutic and valuable prognostic molecules must be established. In recent years, owing to bioinformatics methods evolving that ensure researchers have access to the public database to explore and offer a new angle to carcinoma-involved molecules.
In this study, we divided all the cases downloaded from the TCGA platform into two different groups marked high and low immune/stromal scores, respectively. Then, 908 DEGs (900 upregulation and eight down-regulation) were screened from the immune score and stromal score based on the ESTIMATE algorithm [8]. C-X-C motif ligand 9,10,11 (CXCL9, CXCL10, CXCL11) is a member of the CXC family chemokine. They are principally expressed by immune cells such as macrophages and T cells, and they play a signi cant role in immune cell in ltration into the tumour bed. They are supposed to act as a tumour-suppressive molecule [18,19]. CXCR3, the receptor for CXCL9 and CXCL10, is highly expressed on CD8 + and CD4 + immune-in ltrating lymphocytes, it is highly likely that local production of CXCL9/10 can modulate T-cell recruitment and activation in human cancers [18][19][20]. These chemokines are increased following expose to anti-PD-1 and that they are crucial for therapeutic activity and correlate with patient prognosis [21]. The expression of CXCL11 correlated with improved responses to atezolizumab and was the most signi cantly upregulated gene in macrophages responding to immune checkpoint blockade and its function highly relative to CXCL9 and CXCL10 [22,23]. C-C motif ligand 5, CCL5, is a small protein that belongs to a large family of cytokines and displays chemotaxis activity for it is involved in promoting the migration of several leukocytes into in ammation sites. CCL5 is secreted by a wide variety of cells, including T cells, NKs, and some tumour cells. Previous data showed that targeting CCL5 was su cient to inhibit the in ltration of NK cells signi cantly and subsequently enhance the tumour growth [24,25].
Macrophage migration inhibitory factor (MIF) is an in ammatory cytokine and serves as a regulator of the innate immune system. Studies have shown that MIF induces an immune-suppressive environment that supports melanoma progression and metastasis. CD74 is the central receptor for MIF and the invariant chain of the MHC class II, which plays an essential role in antigen presentation. MIF and CD74 are attractive targets for immunotherapy. MIF binding to CD74 activates the PI3K/AKT and MAPK signalling pathways, and both these pathways have been related to monocyte immune-suppression and macrophage M2-like polarization to regulate the immune response against metastatic melanoma [26][27][28].
Spleen tyrosine kinase (SYK) functions as a non-receptor kinase, mediating signal transduction of cellular transmembrane receptors and act as immunoreceptors and integrins. SYK has been demonstrated to be a critical regulator of the target of rapamycin (mTOR) activity in B-cell lymphoma [29,30]. IL-10 is a potent anti-in ammatory molecule produced by innate and adaptive immune cells, including T cells, NK cells, as well as tumour cells. IL-10 is an immune-regulatory cytokine that may exert immune-stimulatory effects on CD8 + T cells, depending on their state of activation [21]. CD247 plays a crucial role in triggering the TCR signalling pathway, and high expression levels of CD247 was associated with poor overall survival in lower-grade glioma [2]. Nevertheless, the correlation of CD247 and CMM patients' overall survival has been no reported in recent years.
Then, we used a univariate Cox model to determine prognostic-related genes highly correlated with overall survival (OS). We combined the identi ed prognostic-related genes with the immune-related DEGs from the PPI network, and CD247 was the intersection gene for further explored, pair-difference analysis between tumour cases and normal cases and overall survival curve was performed for CD247. The results showed that CD247 highly expressed in tumour samples and possessed a better prognosis than lower expressed samples. In these ndings, WGCNA analysis showed that 11 modules were identi ed.
Then, the relationship between each module traits was studied. Speci cally, genes belong to module brown had the highest positive correlation with CMM patients' prognosis, and CD247 included in this module, which suggested CD247 higher expressed samples have a better prognosis.
Moreover, CMM patients with high expression of the ten genes (CCL5, CXCL10, CD74, CXCL9, CD247, IL10, CXCL11, CXCR3, CD3E, and SYK) were associated with better OS. IHC results from the HPA veri ed that IL10, CXCR3, SYK, CD247, CD74, CXCL11 were signi cantly overexpressed in CMM tissues. The results suggested that CD247 was an essential biomarker in CMM tumorigenesis, progress, and prognosis. Also, the expressions of CD247 was positively correlated with six immune cells in ltration Even though the paramount role of TME in CMM has been reported, many engaged in the immune in ltrating CD8 + T cell. Notably, our study embraces several merits. First, we examined the correlation between immune in ltrating macrophages (M0, M1, M2) and CD247 expression, which is indicated that macrophages, an indispensable part of the microenvironment (TME), was involved in CMM tumorigenesis and progression. Second, the RNA-seq pro ling matrix was collected from the TCGA platform, which contained 468 CMM cases. We explored the OncoLnc website, the HPA resource, and the GEPIA database to con rm the results. Finally, the TME and immune-related molecules interact in a changing procession, which subjected to numerous factors, such as immune cells and genetic aspects.
We thoroughly investigated the underlying association of TME and immune-linked genes from the element of CMM microenvironment, stromal/immune condition, CD247, and other prognostic-relevant genes and immune cells in ltrating.

Conclusion
In summary, our study mainly illustrated the interaction of TME and immune-linked genes. We nally labelled CD247 as a promising prognostic biomarker for cutaneous cancer.  Figure 1 The work ow of data processing. WGCNA: weighted gene expression network analysis. PPI: proteinprotein interaction.      Survival analysis of prognostic-related differential genes. The Kaplan-Meier survival curves reveal the correlations between the expression levels of differential genes and the CMM patients' overall survival time.

Figure 10
Survival curve of prognostic genes based on samples from the OncoLnc platform. The results of OncoLnc data are likely to TCGA outcomes. High gene expression is associated with good prognostic for CMM patients (p < 0.05 in Log-rank test).