Screening for EDGs
To identify the gene signature associated with metastasis in melanoma, mRNA expression levels were compared in T versus N and M versus N in the GSE15605 and GSE46517 datasets, respectively. |log2FC|> 1 and P
Identification of specific genes associated with metastasis in melanoma. (A,B) Differentially expressed genes (DEG) were screened by volcano plot comparing 16 normal (N), 46 primary melanoma (T), and 12 metastatic melanoma (M) skin tissues of GSE15605. (VS,D) DEGs were examined by volcano plot when comparing 7 N samples, 31 T samples, and 73 M samples in GSE46517. (E) Venn diagram for DEG overlap in 4 microarray datasets. |log2 FC|> 1 and adj. P
Functional Enrichment Analysis for DEGs
The selected 294 DEGs were subjected to GO and KEGG pathway analysis by R software. In the analysis of GO biological processes, DEGs were mainly focused on positive regulation of locomotion, mitotic cell cycle process, epithelial cell differentiation and the process based on actin filaments. In the KEGG pathway analysis, DEGs were dominantly enriched in cancer pathways, cancer microRNA, phospholipase D signaling pathway, and purine metabolism (Fig. 2).
DEG GO and KEGG pathway enrichment analysis. (A) Top 10 biological processes enriched for DEG. (B) Top ten enriched KEGG pathways for DEGs.
PPI Network Analysis and Genetic Screening Hub
The STRING database was used to draw the DEG PPI network diagram. DEGs with an interaction score ≥ 0.7 were eligible to build the relational network. We ranked the top 20 genes of the entire network based on the CytoHubba models of the Cytoscape plugin: degree, MCC, MNC, EPC, and bottleneck (Fig. 3a–e). A Venn analysis was performed to obtain the intersection of these genes. Notably, the top 20 genes in the topological analysis algorithms included four hub genes: CDK1, FOXM1, KIF11, and RFC4, which may be involved in the development of metastatic melanoma (Fig. 3f).
Identification of hub genes. (A–E) Hub genes were identified using five methods of topological analysis (A) Diploma, (B) Bottleneck, (VS) MCC, (D) multinational, and (E) EPC with cytoHubba (F) A Venn diagram showed that four hub genes were identified.
Validation of hub gene expression
To further verify previously defined hub genes, we obtained expression data from 472 TCGA-SKCM samples, including 104 primary melanoma and 368 metastatic tissue samples from the UALCAN database. The results showed that three hub genes (CDK1, KIF11 and RFC4) were significantly upregulated in metastatic melanoma tissues compared to primary melanoma tissues (Fig. 4a–d). Significant differences are displayed as follows: *PPPP> 0.05). To explore the protein expression level of four hub genes in SKCM, we performed immunohistochemical analysis of protein expression using the HPA database. As shown in Fig. 4th–h, we found that except for moderate RFC4 staining in primary and metastasized melanoma tissues, the other three hub genes showed moderate expression in primary tumor tissues, but showed strong expression in metastatic melanoma tissue. .
Validation of hub gene expression in primary and metastatic melanoma samples. (AD) The boxplot of mRNA expression level of hub genes was obtained by UALCAN platform. *PPPE–H) Immunohistochemical staining of CDK1, RFC4, FOXM1 and KIF11 in primary melanoma and metastatic melanoma tissues using the HPA database in SKCM.
Prognostic value of hub genes in SKCM using GEPIA database and UALCAN database
To assess the effect of hub gene expression on the prognosis of SKCM, we performed survival analysis to identify the association of hub genes with overall survival (OS) and disease-free survival (DFS) in the GEPIA database. As shown in Fig. 5, the up-regulated expression of two hub genes (CDK1: OS P= 0.037; FOXM1: operating system P= 9.7e−06) were positively correlated with a poor prognosis. Expression of KIF11 and RFC4 had no correlation with OS and DFS. To confirm the results, the relationships between four hub genes and OS were investigated using the UALCAN database. The results suggest that melanoma patients with high CDK and FOXM1 expression had shorter OS (CDK1: OS P= 0.047; FOXM1: operating system P= 0.00043) (Fig. S1).
Association of hub gene expression levels and SKCM prognosis based on TCGA SKCM datasets by GEPIA. P
To further verify the prognostic value of hub genes in melanoma patients, Cox regression analysis was performed by univariate and multivariate on the TCGA and GSE46517 dataset. The results indicated that FOXM1 was significantly associated with OS and was a risk factor with HR > 1 in the TCGA-SKCM cohort (univariate Cox: HR=1.31, 95% CI 1.15–1.5, PP 1 and the Pwas close to 0.05 (HR = 1.23, 95% CI 0.98–1.55, P= 0.053), suggesting that an elevated CDK1 level was correlated with a poor prognosis of melanoma. Two other hub genes were not correlated with overall survival in melanoma patients. Due to limited sample size, the difference was not significant in GSE46517 (Fig. S2).
Immune Infiltration Analysis
We analyzed the correlations between the expression of CDK1 and FOXM1 and the levels of immune infiltration of six immune cells (CD8+ T cells, CD4+ T cells, B cells, neutrophils, macrophages and myeloid dendritic cells) in the microenvironment of SKCM metastases using TIMER. PP= 1.94e−02) (Fig. 6a, S3). CDK1 expression was negatively correlated with macrophage cell infiltration levels (Rho = -0.164, P= 2.02e−03) and positively correlated with the degree of neutrophil cell infiltration (Rho = 0.269, P= 2.72e-07) (Fig. 6b). CDK1 expression was not associated with levels of immune infiltration of CD8+ T cells, CD4+ T cells, B cells, or dendritic cells (Fig. S3).
Hub gene analysis. (A) Correlation analyzes of FOXM1 expression with the degree of myeloid dendritic cell infiltration in SKCM metastases via the TIMER database. (B) Correlation analyzes of CDK1 expression and immune infiltrates (neutrophils and macrophages) in SKCM metastases via the TIMER database. (VS) Drug-hub gene interaction network. The blue nodes represented the drug. The line represents the interaction relationship between CDK1 and the drug.
Analysis of drug-gene interaction networks
The DGIdb database was used to predict potential drug targets that interacted with the CDK1 and FOXM1 genes. We identified 10 drugs that interacted with CDK1; however, neither drug interacted with FOXM1, as shown in Figure 6c. These results could reveal therapeutic targets related to metastatic melanoma.
