Fig 1.
Workflow diagrams are utilized to describe the entire research process.
Fig 2.
Identification of DEGs and WGCNA.
(A) Volcano plot illustrating differential expression of TRGs between HCC and normal tissues in the TCGA-LIHC. (B) The heat map suggests the top 60 DEGs. (C) Selection of the optimal soft-thresholding power for network construction. (D) Identification of gene co-expression modules through dynamic tree cutting. (E) Heatmap depicting the correlation between module eigengenes and clinical traits. (F, G) Scatter plots of module membership against gene significance for the blue module in HCC samples (F) and normal samples (G). (H) The Venn diagram suggests 141 candidate TRGs.
Fig 3.
PPI and functional enrichment analyzes of candidate TRGs.
(A) PPI network of the 141 candidate TRGs derived from the STRING database, with an interaction score threshold > 0.4. (B) The Venn diagram suggests the further screening of hub TRGs through four algorithms: Degree, EPC, MCC, and MNC. (C) Bubble plot denotes GO enrichment analysis. (D) Chord diagram visualizes associations in the GO enrichment analysis. (E) Bubble plot displays KEGG pathway enrichment results. (F) Chord diagram illustrates KEGG pathway associations.
Fig 4.
Prognostic signature construction and validation.
(A) Sankey diagram illustrating co-expression associations between TRGs and lncRNAs. (B) Further screening of prognostic TRLs via LASSO regression analysis. (C) Cross-validation of LASSO regression. (D) Four TRLs and their respective correlation coefficients. (E) Four TRLs were selected through multivariate COX regression analysis. (F) Summary plot of SHAP values illustrates the overall contribution of each feature to model predictions. (G) Beeswarm plot exhibits the distribution and direction of SHAP values for individual features across the cohort. (H, I) Waterfall plot (H) and force plot (I) visualize the additive contribution of each feature to a specific prediction, demonstrating how the base value is adjusted to the final output.
Table 1.
Homogeneity of the training and testing sets in terms of the baseline characteristics of clinical indicators.
Fig 5.
Clinical characteristics analysis.
(A) Circos plot of clinicopathological characteristics and comparative distribution between L-R and H-R groups. (B) Heat map visualizing the correlation between two risk groups and clinical characteristics. (C-J) Boxplot of 4-TRLs rooted in TRLs signature in HCC patients with different fusat (C), ages (D), genders (E), pathological grades (F), tumor stages (G), T stages (H), N stages (I) and M stages (J).
Fig 6.
Assessment of the prognostic signature.
(A-C) Distribution of risk scores grounded in the 4-TRLs signature, survival time, survival status, and expression heatmap of the four TRLs in the (A) training, (B) testing, and (C) all sets. (D-F) Kaplan-Meier survival curves comparing overall survival between H-R and L-R groups in the (D) training, (E) testing, and (F) all sets. (G-I) ROC curves evaluating the predictive accuracy of the risk model in the (G) training, (H) testing, and (I) all sets. (J-L) ROC curves comparing the prognostic performance of the 4-TRLs signature against conventional clinical parameters in the (J) training, (K) testing, and (L) all sets.
Fig 7.
Nomogram construction and validation.
(A) Forest plot from Cox regression analysis identifying independent prognostic factors. (B) C-index values comparing the predictive performance of the 4-TRLs signature and conventional clinicopathological features. (C) Nomogram integrating the 4-TRLs signature and clinical characteristics for predicting survival probability. (D) Calibration curves evaluating the agreement between predicted and observed survival outcomes. (E-G) Decision curve analysis assessing the clinical utility of the nomogram for predicting overall survival at (E) 1 year, (F) 3 years, and (G) 5 years.
Fig 8.
(A) Waterfall plot of somatic mutation characteristics in the H-R group. (B) Waterfall plot of somatic mutation characteristics in the L-R group. (C) Chromosomal locations of somatic mutations for the 50 most frequently altered genes in HCC. (D) Comprehensive annotation of mutated genes in HCC, comprising variant types and functional impacts. (E) Relationship of risk cohorts and CSC index. (F) Violin plot illustrating the differences in TMB between different risk groups. (G) Kaplan-Meier survival curves comparing overall survival between high and low TMB groups. (H) Kaplan-Meier analysis of overall survival across four subgroups stratified by combined TMB status and 4-TRLs risk signature. (I-K) Violin plots displaying differences in TIDE scores (I), dysfunction scores (J), and exclusion scores (K) between H-R and L-R groups.
Fig 9.
Tumor immune characteristics and drug sensitivity profiling.
(A, B) Differences in immune cell infiltration between risk groups assessed by adopting the CIBERSORT algorithm, presented as a waterfall plot (A) and violin plots (B). (C) Correlation heatmap between the 4-TRLs signature and 22 immune cell types. (D) Differential immune function activity between risk groups analyzed via single-sample gene set enrichment analysis. (E) Correlation matrix illustrating associations between the 4-TRLs signature and 29 immune checkpoint genes. (F) Distribution of immune subtypes in H-R and L-R groups defined by the 4-TRLs signature. (G-J) Violin plots illustrating differential sensitivity (IC₅₀) to 8 clinically relevant anticancer agents between risk groups.
Fig 10.
Consensus clustering matrix at k = 4. (B) Relative changes in the area under CDF curves for k = 2–9. (C) Consensus clustering CDF plot for cluster numbers k = 2–9. (D) Sample clustering tracking graph for k = 2–9. (E) OS of different clusters. (F) Heat map of immune cell infiltration differences among different clusters. (G) Correlation graph among immune cells. (H) Differences in the abundance of 22 immune cells between the four clusters. (I-J) Immune checkpoint genes differences analysis of HCC patients with different clusters.
Fig 11.
Single-Cell RNA Sequencing Data Analysis.
(A) Violin plots display gene counts and sequencing depth across cells. (B) Pearson correlation analysis demonstrates negative correlation between sequencing depth and mitochondrial gene expression, and positive association with gene counts. (C) Volcano plot underlines highly variable genes across the cell population. (D) UMAP nonlinear dimensionality reduction identifies 27 distinct cell clusters. (E) Heatmap suggests the top 30 marker genes for the first three clusters. (F) The 27 cell clusters are annotated into 8 major cell types. (G) Violin plots illustrate expression patterns of the 4-TRLs across different cell populations. (H-K) Distribution of individual TRLs: (H) CYTOR in T cells, tissue stem cells, and B cells, (I) AC026356.1 in endothelial cells and macrophages, (J) AL512598.1 and (K) MIR210HG primarily in hepatocytes. (L) Dot plot summarizes expression levels of the 4-TRLs across major cell types, illustrating both expression intensity and prevalence.
Fig 12.
Pilot experimental validation of the 4-TRLs signature and internal assessment of lncRNA AC026356.1.
(A-D) RT-qPCR analysis of AC026356.1 (A), AL512598.1 (B), CYTOR (C), and MIR210HG (D) expression levels in 15 paired HCC tumor and adjacent normal specimens, compared using Student's t-test. (E) Differential expression of 4-TRLs between tumors and matched benign tissues in TCGA-LIHC cohort. (F) K-M survival curves stratified by AC026356.1 expression levels. (G) The risk score distribution plot stratified by AC026356.1 expression levels. (H) Survival time and status distribution plot rooted in AC026356.1 expression levels. (I) ROC analysis of AC026356.1's diagnostic accuracy. (J) Time-dependent ROC curves for AC026356.1 high/low expression groups. (K) Heat map illustrating the correlation between AC026356.1 differential expression and clinical features. (L-M) Forest plots of univariate (L) and multivariate (M) Cox regression analyzes.
Fig 13.
Prediction of a telomerase-associated lncRNA-miRNA-mRNA regulatory axis.
(A) lncRNA-miRNA interaction network predicted by utilizing StarBase database. (B) Predicted binding sites between the lncRNA AC026356.1 and target miRNA. (C) Box plot illustrating differential expression of AC026356.1 between tumor and normal tissues in the TCGA-LIHC cohort. (D-E) Comparative expression analysis of AC026356.1 in the TCGA-LIHC dataset: (D) unpaired tumor versus normal tissues; (E) paired tumor and adjacent non-tumorous tissues. (F) Scatter plot illustrating the correlation between AC026356.1 and hsa-mir-126-5p expression. (G) Venn diagram illustrating the overlap of predicted target mRNAs from four databases (mirDIP, miRWalk, StarBase, and TargetScan). (H) Venn diagram identifying four telomerase-associated target mRNAs. (I) Box plots displaying differential expression levels of the four telomerase-related target mRNAs in the TCGA-LIHC dataset. (J-M) Correlation analyzes between AC026356.1 expression and target mRNAs: (J) YAP1, (K) SKAP2, (L) NFIB, and (M) PRKAA2.