Development and validation of a ferroptosis-related lncRNAs signature to predict prognosis and microenvironment for melanoma

Abstract

Ferroptosis plays an important role in cancer. However, studies about ferroptosis-related lncRNAs (FRLs) in skin cutaneous melanoma (SKCM) are scarce. Moreover, the relationship between prognostic FRLs and tumor microenvironment (TME) in melanoma remains unclear. This study investigates the potential prognostic value of FRLs and their association with TME in SKCM. The RNA-sequencing data of SKCM were downloaded from The Cancer Genome Atlas (TCGA) database. Melanoma patients were randomly divided into training and testing groups in a 1:1 ratio. A signature composed of 19 FRLs was developed by the least absolute shrinkage and selection operator (LASSO) regression analysis to divide patients into a low-risk group with a better prognosis and a high-risk group with a poor prognosis. Multivariate Cox regression analysis suggested that the risk score was an independent prognostic factor. The Area Under Curve (AUC) value of the risk score reached 0.768 in the training group and 0.770 in the testing group. Subsequent analysis proved that immune-related signaling pathways were significantly enriched in the low-risk group. The tumor immune cell infiltration analysis demonstrated that melanoma in the high-risk group tended to be immunologically "cold". We identified a novel FRLs signature which could accurately predict the prognosis of patients with melanoma.

Keywords: Ferroptosis; Melanoma; Prognosis; Tumor microenvironment; lncRNA.

Fig. 1

Consistent cluster analysis of SKCM. a The consistent cluster CDF when k ranges from 2 to 9. b The change in the area under the CDF curve when k is between 2 to 9. c The overlap of consistent clustering matrix in clusters for k = 2. d The distribution of each sample in the range of k from 2 to 9. e Significant difference in K-M survival curves between cluster 1 and cluster 2 (p < 0.05). f The clinicopathological differences between cluster 1 and cluster 2

Fig. 2

Gene mutations and the tumor microenvironment in the two clusters. a–b The comparison of immune checkpoint genes expression in the two clusters. c–e The comparison of commonly mutated gene expression in the two clusters. f–h Differences in the immune microenvironment scores between Cluster 1 and Cluster 2

Fig. 3

Construction and validation of the FRLs signature. a, b The LASSO regression was conducted with the minimum criteria. c–m The Kaplan–Meier survival curves for FRLs in signature

Fig. 4

Assessment of the prognostic prediction ability of the FRLs signature. a, f The Kaplan–Meier survival analysis for high- and low-risk groups in training and testing groups. b, g The ROC curve was used to evaluate the predictive efficiency of the prognostic signature. c, h The DCA curve demonstrated that the risk score performed better than traditional clinicopathological factors in predicting the prognosis of SKCM. d, e, i, j The distribution plots of the risk score and survival status in the training and testing groups. k, n The heat map of FRLs for the low-and high-risk groups in the training and testing groups. l, o The univariate Cox regression of prognostic factors in the training and testing groups. m, p The multivariate Cox regression of prognostic factors in the training and testing groups

Fig. 5

Pathway enrichment analyses. Immune-related signaling pathways and processes were enriched in the low-risk group (a–j). All p < 0.05, FDR < 0.25

Fig. 6

Analysis of differences in immune infiltration and gene expression in two subgroups. a Heatmap for immune responses based on TIMER, CIBERSORT, CIBERSORT-ABS, QUANTISEQ, MCPCOUNTER, XCELL, and EPIC algorithms among the high- and low-risk groups in the TCGA cohort. b ssGSEA for the association between immune cell subpopulations and related functions. c Expression of immune checkpoints among high- and low-risk groups in TCGA cohort

Fig. 7

The Relationships between the Risk Score and Tumor-Infiltrating Immune Cells (a–h)

Fig. 8

The differential expression of FRLs was detected by qPCR. a, b Compared with the HaCaT cell line, the FRLs of LINC00861 and LINC01094 were significantly lower in the A2058 and A375 cell lines. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001