Abstract
The aim of this study is to find those N7-methylguanosine (m7G) methylation-related regulator genes (m7GMRRGs) which were associated with melanoma prognosis and use them to develop a prognostic prediction model. Clinical information was retrieved online from The Cancer Gene Atlas (TCGA) and the Gene Expression Omnibus (GEO). R software was used to extract m(7)GMRRGs by differential expression analysis. To create a prognostic risk model, univariate and multivariate Cox regression analyses were employed for the evaluation of the prognostic significance of m(7)G methylation modifiers. Internal validation using cohort from TCGA (training set) and external validation using cohort from GEO (validation set) of the model were carried out. The model's predictive performance was confirmed by using the Kaplan-Meier, univariate, and multivariate Cox regression, and receiver operating characteristic curve (ROC) by constructing column line plots incorporating clinical factor characteristics. Immune infiltration analyses were performed to assess the immune function of m(7)GMRRGs. Drug sensitivity analysis was conducted to study chemotherapeutic drug treatment cues. Prognostic models using four m(7)GMRRGs (EIF4E3, LARP1, NCBP3, and IFIT5) showed good prognostic power in training and validation sets. The area under the curve (AUC) at 1, 3, and 5 years for GEO-melanoma were 0.689, 0.704, and 0.726, respectively. The prediction model could distinctly classify patients with melanoma into different risk subgroups (P < 0.001 for TCGA-melanoma and P < 0.05 for GEO-melanoma). Clinical characteristics were taken into account in Cox regression and AUC analysis, which highlighted that the risk score served as an independent risk factor determining the prognosis of patients with melanoma. Immuno-infiltration analysis showed that m(7)GMRRGs could potentially regulate CD8(+) T cells as well as regulatory T cells (Treg cells). Results of our study indicate a association between m(7)GMRRGs and melanoma prognosis, and the prognostic prediction model using m(7)GMRRGs may predict the prognosis of patients with melanoma well. Nevertheless, these results may provide a clue for potential better options of melanoma treatment but need further validation in futural studies.