WGCNA and single-cell analysis reveal ferroptosis-related gene signatures for hepatocellular carcinoma prognosis and therapy

Abstract

Background: Hepatocellular carcinoma (HCC) is a globally serious malignant tumor with high incidence and mortality. Ferroptosis, a newly discovered form of regulated cell death, is significant in tumor initiation and growth. Herein, we performed bioinformatics analysis in order to investigate the expression heterogeneity of ferroptosis-related genes as well as its correlation with HCC clinical outcomes.

Methods: The gene expression data of HCC were downloaded from the TCGA and GEO databases. Weighted Gene Co-expression Network Analysis (WGCNA) was deployed to build gene co-expression networks and explore ferroptosis-related gene modules. We used single-cell RNA sequencing data to analyze the expression of these genes in different cell types. Survival analysis and functional enrichment analysis were utilized to study the biological function and clinical significance of these genes in HCC.

Results: Several ferroptosis-related gene modules were identified by WGCNA, one of which was significantly linked with the clinical characteristics of HCC. Analysis of single-cell sequencing data revealed distinct expression of these core genes in different cell types. Survival analysis revealed that certain ferroptosis-related gene expressions were strongly correlated with patient survival outcomes. Functional enrichment analysis indicated that these genes mainly participate in oxidative stress response, iron metabolism, and apoptosis.

Conclusions: This study reveals the expression heterogeneity of ferroptosis-related genes in HCC and may provide new molecular targets for HCC prognosis and therapy.

Keywords: Bioinformatics; Ferroptosis; Hepatocellular carcinoma.

Fig. 1

Lysosome-associated cell death in liver hepatocellular carcinoma (HCC) through heatmaps and a Venn diagram. A: Heatmap of differential expression displays gene expression differences between tumor and normal tissues. Red indicates higher expression in tumors, blue indicates higher expression in normal tissues. B: Extended Heatmap shows a broader range of gene expression data. Similar color coding for expression levels, highlighting more genes. C: Venn diagram compares gene expression data from TCGA and GTEx

Fig. 2

Survival outcomes and clustering of gene expression data. A: Hazard ratio plot displays hazard ratios for various genes. Indicates genes significantly associated with survival, with some having higher risk (hazard ratio > 1). B: Consensus Matrix shows clustering of samples into two groups. Two distinct clusters are identified, suggesting different gene expression profiles. C: Kaplan-Meier plot comparing survival between the two clusters. Cluster A has better survival than Cluster B, with a significant p-value (p = 0.007). D: t-SNE Plot visualizes the separation of samples into two clusters. Clear distinction between Cluster A (blue) and Cluster B (red)

Fig. 3

Analyses of gene expression and immune characteristics in HCC. A: Heatmap shows gene expression differences between two clusters (C1 and C2). Red indicates higher expression, blue indicates lower expression. B: Boxplots of gene expression compares gene expression levels between clusters C1 and C2. Significant differences in expression for several genes. C, D, E: Boxplots of Scores: Compare ImmuneScore, StromalScore, and ESTIMATEScore between clusters. Cluster C2 shows higher scores in all categories, indicating differences in tumor microenvironment

Fig. 4

Risk assessment and gene expression in HCC. A: Model Performance displays performance metrics for different models. Various models are compared based on their accuracy and reliability. B: Boxplot showing risk scores for two clusters (A and B). Cluster A has significantly higher risk scores than Cluster B (p < 2.22e-16). C: Sankey diagram Illustrates the relationship between clusters, risk levels, and survival outcomes. Cluster A is associated with higher risk and worse survival. D: Heatmap shows gene expression differences between clusters

Fig. 5

Risk prediction and survival in HCC. A–C: ROC Curves show model performance over 1, 3, and 5 years. AUC values indicate good predictive accuracy for survival. D–F: Kaplan-Meier plots for low and high-risk groups. High-risk groups have significantly worse survival (p < 0.001, p = 0.004). G–I: Decision curves evaluate clinical usefulness of risk models. Models provide net benefit across various risk thresholds

Figure 6

Analyses of immune cell infiltration and gene expression in HCC. A: Correlation Matrix shows correlations between different immune cell types. Red indicates positive correlation, blue indicates negative. B: Violin plots compares immune cell infiltration between low and high-risk groups. C: Boxplots displays scores for immune-related functions. D: Gene expression boxplots compares expression of key genes between risk groups

Fig. 7

Gene expression and correlations in HCC. A: Heatmap displays gene expression across samples. B: Expression levels compares gene expression in normal vs. tumor tissues. C: Correlation circle Illustrates correlations between genes. D: Network diagram visualizes gene correlations

Fig. 8

The proportions of various immune cell types in HCC. There are variations in immune cell compositions associated with different expression levels of these genes, indicating potential roles in immune response modulation in HCC

Fig. 9

Analyses of gene performance in predicting outcomes in HCC. A–C: AUC over time show area under the curve (AUC) for different genes over 1 to 5 years. D: Circos plot displays genomic locations of key genes. E: ROC curves compares ROC curves for various genes. FTL shows the highest AUC, indicating strong predictive ability

Fig. 10

Analyses of gene performance in predicting outcomes in HCC. A: Mutation overview shows types and frequencies of mutations. Missense mutations are most common, with CHGA and ATP6V0D2 frequently mutated. B: Mutation types displays mutation type distribution. C > T and T > G mutations are prevalent. C: Heatmap shows mutation frequency in specific genes. CHGA and ATP6V0D2 have the highest mutation frequencies. D: Sample distribution illustrates mutation occurrence across samples

Fig. 11

Analyses of copy number variations (CNVs) in HCC. A: CNV and mRNA correlation shows correlation between CNV and mRNA expression. Significant correlations for certain genes like AP4B1 and FTL. B: Survival differences displays survival differences between CNV groups. No significant survival impact across CNV groups. C: Heterozygous CNV shows heterozygous amplifications and deletions. Frequent in genes like MT3 and CHGA. D: Homozygous CNV shows homozygous amplifications and deletions. Less frequent compared to heterozygous CNVs

Fig. 12

Analyses of gene correlations, methylation differences, and survival metrics in HCC. A: CNV and mRNA correlation between CNV and mRNA expression. Significant correlations for genes like CHGA and ATP6V0D2. B: Methylation differences in tumor vs. normal tissues. C: Survival analysis impact of genes on different survival metrics. A: CNV and mRNA correlation between CNV and mRNA expression. Significant correlations for genes like CHGA and ATP6V0D2. B: Methylation differences in tumor vs. normal tissues. C: Survival analysis impact of genes on different survival metrics

Fig. 13

Immune cell types in HCC using data from the GSE98638 dataset. A and B: UMAP plots showing cell type distribution. Different immune cell types like CD4Tconv, CD8T, and Treg are identified. C: Bar plot of cell type proportions across patients. Variability in immune cell composition among patients. D: Cell Type Distribution. Pie chart of overall cell type proportions. CD4Tconv cells are the most prevalent

Fig. 14

UMAP plots illustrating the expression levels of six genes. Gene expression varies across cells, with FTL being prominently expressed in HCC. FTL shows strong expression across many cells, while others like CHGA and MT3 have more limited expression