Table 1.
Baseline Characteristics Comparison of HCC Patients from the GSE14520 and TCGA-LIHC Databases.
Fig 1.
The workflow of this study involves the use of TCGA-LIHC (The Cancer Genome Atlas Liver Hepatocellular Carcinoma dataset), GSE14520 (GEO dataset), FerrDb (Ferroptosis Database), DEG (Differentially Expressed Genes), and LASSO (Least Absolute Shrinkage and Selection Operator).
Fig 2.
Differential gene analysis results (A): Volcano plot showing the expression of differential genes in normal and tumor samples from the TCGA dataset.
(B) Venn diagram of upregulated differentially expressed genes and ferroptosis-related genes (FRGs). (C) Venn diagram of univariate Cox genes in TCGA_train and TCGA_test datasets. (D) Lasso regression coefficient plot of the 11 FRGs. (E) Tuning parameters in the Lasso model. (F) Heatmap of the expression of five gene features in normal and tumor tissues from the TCGA dataset.
Fig 3.
Model construction and validation (A) Construction of the model based on multiple machine learning algorithms.
(B) C-index of the StepCox[forward] model across different datasets. (C) AUC values of the StepCox[forward] model for 1-year, 2-year, and 3-year outcomes in different datasets. (D-F) ROC curves for 1-year, 2-year, and 3-year outcomes in various datasets.
Fig 4.
Validation of the prognostic model (A-D) Kaplan-Meier survival curves for the overall survival of two risk subgroups in TCGA_train, TCGA_test, TCGA, and GSE14520.
Dashed line represents the Median Overall Survival (OS). (E) Meta-analysis of univariate Cox regression. High: High risk group based on the ferroptosis-related gene signature (nFRGs). Low: Low risk group based on the ferroptosis-related gene signature (nFRGs).
Fig 5.
Performance of the prognostic model risk subgroups in the TCGA and GSE14520 datasets.
(A-D) The PCA 3D analysis, distribution of risk scores, correlation between survival time and survival status for each patient, and heatmap of the prognostic model genes in the risk subgroups for TCGA. (E-H) The PCA 3D analysis, distribution of risk scores, correlation between survival time and survival status for each patient, and heatmap of the prognostic model genes in the risk subgroups for GSE14520.
Table 2.
Comparison between our ferroptosis-related gene prognostic model and previously published models.
Fig 6.
Comparison with previously published ferroptosis-related gene models.
(A) C-index for each model in the different datasets. (B) HR heatmap for each model in the different datasets, with different colors representing different hazard ratios (HR), where darker colors indicate higher HR. Statistical significance is indicated by symbols: * for p < 0.05 and ** for p < 0.01, indicating statistical significance.
Fig 7.
Construction and validation of the nomogram (A-C) Forest plots of univariate and multivariate Cox analyses for risk score subgroups and clinical features in the TCGA dataset, along with the construction of the risk score nomogram.
(D-F) Forest plots of univariate and multivariate Cox analyses for risk score subgroups and clinical features in the GSE14520 dataset, along with the construction of the risk score nomogram.
Fig 8.
Functional enrichment analysis of DEGs between risk score subgroups (A, B) shows GO enrichment analysis of DEGs between high-risk and low-risk groups.
(C-D) KEGG enrichment analysis of DEGs between high-risk and low-risk groups. (E) Gene set C7 enrichment analysis (GSEA). (F) Gene set Hallmark enrichment analysis (GSEA).
Fig 9.
PPI network and nFRGs Pearson correlation analysis (A) PPI network diagram of nFRGs.
(B) Heatmap of nFRGs correlation.
Fig 10.
Tumor mutation analysis (A) The Kaplan-Meier survival curve shows the combined impact of TMB and risk score on survival prognosis.
(B) A waterfall plot showing detailed mutation information for the top 20 genes in the TCGA dataset. (C, D) Waterfall plots showing detailed mutation information for the top 20 genes in the high-risk (C) and low-risk (D) groups. (E, F) Mutation information in the high-risk (E) and low-risk (F) groups categorized by different mutation types. (G, H) Correlation heatmaps of the top 20 mutated genes in the high-risk (G) and low-risk (H) groups.
Fig 11.
Immune microenvironment (A) Comparison of immune cell infiltration between the high-risk and low-risk groups using the ssGSEA algorithm.
(B) Comparison of immune function between the high-risk and low-risk groups using the ssGSEA algorithm. (C) Comparison of immune cell infiltration between the high-risk and low-risk groups using the CIBERSORT algorithm.
Fig 12.
Correlation between risk score and immune cells (A-N) The relationship between risk score and immune cell infiltration abundance.
Fig 13.
The relationship between nFGRs and their risk subgroups in the immune environment and immune therapy response.
(A) Correlation analysis of nFGRs and immune infiltrating cells; (B) Comparison of immune checkpoint expression between the high-risk and low-risk groups; (C) Differences in StromalScore between the high-risk and low-risk groups; (D) Differences in TIDE scores between the high-risk and low-risk groups. (E)Differences in IPS between the high-risk and low-risk groups.
Fig 14.
Comparison of drug sensitivity differences between the high-risk and low-risk groups.
Table 3.
Candidate drug predicted using DSigDB.