Identification of Transcriptional Heterogeneity and Construction of a Prognostic Model for Melanoma Based on Single-Cell and Bulk Transcriptome Analysis

Abstract

Melanoma is one of the most aggressive and heterogeneous life-threatening cancers. However, the heterogeneity of melanoma and its impact on clinical outcomes are largely unknown. In the present study, intra-tumoral heterogeneity of melanoma cell subpopulations was explored using public single-cell RNA sequencing data. Marker genes, transcription factor regulatory networks, and gene set enrichment analysis were further analyzed. Marker genes of each malignant cluster were screened to create a prognostic risk score, and a nomogram tool was further generated to predict the prognosis of melanoma patients. It was found that malignant cells were divided into six clusters by different marker genes and biological characteristics in which the cell cycling subset was significantly correlated with unfavorable clinical outcomes, and the Wnt signaling pathway-enriched subset may be correlated with the resistance to immunotherapy. Based on the malignant marker genes, melanoma patients in TCGA datasets were divided into three groups which had different survival rates and immune infiltration states. Five malignant cell markers (PSME2, ARID5A, SERPINE2, GPC3, and S100A11) were selected to generate a prognostic risk score. The risk score was associated with overall survival independent of routine clinicopathologic characteristics. The nomogram tool showed good performance with an area under the curve value of 0.802.