Our aim was to construct a signature that accurately predicted the prognostic and immune response of melanoma. First, the weighted co-expression network analysis (WGCNA) algorithm was used to identify the hub genes related to clinical phenotypes of melanoma in the cancer genome atlas (TCGA) database. Nest, the least absolute shrinkage and selection operator (LASSO) analysis was used to dimensionality reduction of these hub genes and constructed a prognostic signature to predict the prognosis and immunosuppressive response of melanoma. Through in-depth analysis, we constructed a 5-mRNA prognostic signature and verified its prognostic value in internal (TCGA-SKCM, = 452) and external independent datasets (GSE53118, = 79). Based on this signature, the tumor immune microenvironment (TME) of melanoma was characterized, and the result was found that patients in the high-risk group had lower CD8 T cell infiltration and immune checkpoint expression (PD-1, PD-L1, CTLA4), as well as higher M0/M2 macrophage infiltration. Our results also found the risk score based on a 5-mRNA signature was significantly associated with tumor mutational burden (TMB) and tumor stem cell markers (CD20, CD38, ABCB5, CD44, etc.). Lastly, we built a nomogram for clinician prediction for the prognosis of patients with melanoma. Our findings indicated that the 5-mRNA signature has an important predictive value for the overall survival of melanoma. By analyzing the tumor immune microenvironment and tumor stem cell marker between different groups, a new method is provided for the stratified diagnosis and treatment of melanoma.
