The role of cuproptosis-related gene in the classification and prognosis of melanoma

Abstract

Background: Melanoma, as one of the most aggressive and malignant cancers, ranks first in the lethality rate of skin cancers. Cuproptosis has been shown to paly a role in tumorigenesis, However, the role of cuproptosis in melanoma metastasis are not clear. Studying the correlation beteen the molecular subtypes of cuproptosis-related genes (CRGs) and metastasis of melanoma may provide some guidance for the prognosis of melanoma.

Methods: We collected 1085 melanoma samples in The Cancer Genome Atlas(TCGA) and Gene Expression Omnibus(GEO) databases, constructed CRGs molecular subtypes and gene subtypes according to clinical characteristics, and investigated the role of CRGs in melanoma metastasis. We randomly divide the samples into train set and validation set according to the ratio of 1:1. A prognostic model was constructed using data from the train set and then validated on the validation set. We performed tumor microenvironment analysis and drug sensitivity analyses for high and low risk groups based on the outcome of the prognostic model risk score. Finally, we established a metastatic model of melanoma.

Results: According to the expression levels of 12 cuproptosis-related genes, we obtained three subtypes of A₁, B₁, and C₁. Among them, C₁ subtype had the best survival outcome. Based on the differentially expressed genes shared by A₁, B₁, and C₁ genotypes, we obtained the results of three gene subtypes of A₂, B₂, and C₂. Among them, the B₂ group had the best survival outcome. Then, we constructed a prognostic model consisting of 6 key variable genes, which could more accurately predict the 1-, 3-, and 5-year overall survival rates of melanoma patients. Besides, 98 drugs were screened out. Finally, we explored the role of cuproptosis-related genes in melanoma metastasis and established a metastasis model using seven key genes.

Conclusions: In conclusion, CRGs play a role in the metastasis and prognosis of melanoma, and also provide new insights into the underlying pathogenesis of melanoma.

Keywords: cuproptosis; machine learning; melanoma; metastasis model; prognostic model; subtype.

Figure 1

Article framework and workflow.

Figure 2

Classification of melanoma based on CRGs. (A) Molecular subtypes based on CRGs obtained under unsupervised consensus clustering. (B) The empirical cumulative distribution function (CDF) plot depicts the consistent distribution of different K values. (C) Relative increase in cluster stability by delta area fraction. (D) Comparison of the degree of immune cell infiltration of the three molecular subtypes*, P<0.05; **, P<0.01; ***, P<0.001. (E) Kaplan Meier analysis results of three molecular subtypes based on 12 CRGs. (F, G, H) pictures show the enriched pathways of differentially expressed genes obtained by comparing A₁, B₁, and C₁ molecular subtypes with each other using the GSVA method. (I) Heatmap of clinical information and gene expression profiles of the three molecular subtypes based on 12 CRGs.

Figure 3

Differentially expressed genes of three CRGs molecular subtypes. (A) VENN plot showing 71 intersecting differentially expressed genes across three molecular subtypes. (B) t-distributed Stochastic Neighbor Embedding (tSNE) analysis of three CRGs molecular subtypes. (C) Metascape enrichment analysis of DEGs with intersections of the three molecular subtypes. (D) GO enrichment analysis of DEGs with intersections of the three molecular subtypes.

Figure 4

Construction of the prognostic model. (A) Sankey diagram to describe the process of constructing a prognostic model based on CRGs-subtypes and gene subtypes. (B, C) Prognostic genes were screened using LASSO regression. (D, G) Kaplan Meier analysis of OS in melanoma patients in the training set; ROC curves for 6 key variable genes. (E, H) OS of melanoma patients in Kaplan Meier analysis validation set; ROC curves of 6 key variable genes. (F, I) Kaplan Meier analysis of OS in all melanoma patients; ROC curves of 6 key variable genes. (J) Nomograms predicting 1-, 3-, and 5-year OS probabilities in melanoma patients. (K) Calibration plots of the nomograms.

Figure 5

Risk curve and immune microenvironment analysis between high and low immune groups. (A) Risk curve in the training set. Take the median of the risk scores and use the median to divide the samples into high-risk and low-risk groups. (B) Risk curve in the validation set. (C) Risk curves of all samples. (D) Survival state diagram of the training set, red for dead and blue for survival. (E) The living state diagram of the validation set. (F) Survival state diagram of all samples. (G–I) Heat map showing the expression of 6 key variable genes in training set, validation set and all samples. (J) Correlation of 6 key variable genes with immune cells, red represents positive correlation and blue represents negative correlation. (K) Correlation of stroma score, immune score, and ESTIMATE with immune microenvironment.

Figure 6

Drug Sensitivity Analysis. (A–C) The sensitivity of the low-risk group to Sunitinib, VX.702, AZD6482 was higher than that of the high-risk group. The abscissa is the low-risk group and the high-risk group, and the ordinate is the value of the drug IC50. (D–F) The high-risk group had higher sensitivity to OSI.906, FH535, and Bryostatin.1 than the low-risk group. (G) Risk scores for A₁, B₁, and C₁ subtypes in CRGs molecular subtypes. (H) Risk scores for A₂, B₂, C₂ subtypes in genotyping. (I) Expression levels of CRGs in high and low risk groups.

Figure 7

Construction of metastasis model. (A) REFCV method to filter out key metastatic variables in train set. (B) REFCV method to filter out key metastatic variables in validation set. (C) Multi-model forest graph.

Figure 8

Interpretability of the metastasis model. (A) “SHAP” package to explain the importance of key variables to the model. (B) Contribution of each variable to the model. (C, D) Prediction of model.

Figure 9

Knockdown of fdx1 inhibits the proliferation of melanoma cells. (A) FDX1-targeting siRNAs to knockdown the expression levels of FDX1. (B) The proliferation of FDX1 knockdown cells at 12 h, 24 h, and 36 h. (C) Wound healing assay at 12 h, 24 h, 36 h, and 48 h.