Fig 1.
Differential expression analysis and mutational landscape of PRG.
(A) Differential expression analysis of PRG between NC and STAD samples. Thresholds for significance were set at |fold change| ≥ 2 and adjusted p-value < 0.05. (B) CNV frequency analysis of DE-PRG signatures in STAD. (C) Waterfall plot illustrating the mutation burden landscape of DE-PRG signatures in STAD. (D) Prognostic significance and correlation analysis of DE-PRG signatures in STAD.
Fig 2.
Identification of PRG molecular subtypes and prognostic outcome analysis.
(A-C) Identification of distinct PRG molecular subtypes in STAD using unsupervised consensus clustering analysis. (D) PCA plot illustrating the distinct distribution patterns of the identified PRG subtypes. (E) Kaplan–Meier survival analysis based on the log-rank test to evaluate the prognostic outcomes of different PRG subtypes. (F) Differential expression analysis of DE-PRG signatures among the PRG molecular subtypes. (G-I) KEGG pathway enrichment analysis showing signaling differences among PRG molecular subtypes.
Fig 3.
Immune microenvironment landscape and immunotherapy response evaluation of PRG molecular subtypes.
(A–D) Immune infiltration status of PRG molecular subtypes assessed using the ESTIMATE algorithm. (E) Quantitative analysis of the infiltration levels of 23 immune cell types based on the ssGSEA algorithm. (F) TIDE score-based prediction of immune evasion potential. (G–I) IPS score analysis revealing the differential response of PRG molecular subtypes to CTLA-4 and PD-1 targeted therapies.
Fig 4.
Gene subtyping analysis based on DEGs among PRG molecular subtypes.
(A) Identification of DEGs among the PRG molecular subtypes. (B, C) GO and KEGG enrichment analyses of the identified DEGs. (D) PCA plot illustrating the distribution patterns of the three gene subtypes. (E) Kaplan–Meier survival curves of the gene subtypes based on log-rank analysis. (F) Expression profiles of DEGs across different clinicopathological features and molecular subgroups.
Fig 5.
Construction of the PRG score index using integrated machine learning algorithms.
(A) C-index calculated for both training and validation cohorts using combinations of 10 different machine learning algorithms. (B, C) Stratification of PRG score subgroups in the training and validation cohorts. (D, E) Kaplan–Meier survival analysis of PRG score subgroups in two independent cohorts. (F, G) Differential analysis of PRG scores across PRG score subgroups and gene subtypes. (H) Sankey diagram illustrating the potential relationships among PRG molecular subtypes, gene subtypes, PRG score subgroups, and clinical outcomes.
Fig 6.
Evaluation of the independent prognostic value of the PRG score index and construction of a nomogram model.
(A-D) Univariate and multivariate Cox regression analyses of clinicopathological variables and the PRG score in the training cohort and validation cohort. (E, F) Nomogram model construction and calibration curve analysis based on clinicopathological variables and the PRG score index in the training cohort. (G) Time-dependent ROC curve analysis in the training cohort. (H, I) Nomogram model construction and calibration curve analysis based on clinicopathological variables and the PRG score index in the validation cohort. (J) Time-dependent ROC curve analysis in the validation cohort.
Fig 7.
Immunotherapy response prediction and mutational burden landscape analysis of PRG score subgroups.
(A) Differential analysis of TMB scores between PRG scores subgroups. (B–D) IPS analysis of PRG score subgroups. (E, F) Waterfall plots illustrating the somatic mutation landscape of PRG score subgroups.
Fig 8.
Immune microenvironment landscape and drug sensitivity analysis.
(A–D) Evaluation of immune infiltration characteristics using the ESTIMATE algorithm. (E) Assessment of the infiltration proportions of 23 immune cell types based on the ssGSEA algorithm. (F) Drug sensitivity analysis of PRG score subgroups.
Fig 9.
Single-cell RNA sequencing analysis reveals the expression patterns of the prognostic signature across cellular subpopulations.
(A) Quality control of the single-cell RNA-seq dataset GSE163558. (B) Identification of the top 2,000 highly variable genes. (C) Batch effect correction and data normalization using the Harmony algorithm. (D, E) UMAP and t-SNE visualizations showing the distribution of 22 distinct cell subtypes. (F) Expression analysis of the prognostic signature across the 22 identified cell subtypes. (G, H) UMAP and t-SNE plots illustrating the distribution of 9 annotated cell clusters. (I) Quantitative analysis of the PRG signature expression across different cell clusters. (J) Expression profiling of the prognostic signature within the 9 annotated cell subpopulations.
Fig 10.
Comparative analysis of cell-cell communication networks between NC and GC samples.
(A) Bar plots showing the total number (left) and overall strength (right) of inferred intercellular interactions in NC and GC samples. (B) Differential intercellular communication network between GC and NC. (C) Circle plots depicting the number and strength of intercellular communications in NC (top row) and GC (bottom row). (D) Heatmaps displaying the number of interactions among different cell types in NC (left) and GC (right).
Fig 11.
Suppression of COL8A1 expression significantly reduces the proliferation and migration of STAD cells.
(A, C) WB analysis of COL8A1 expression in GES-1 and HGC-27 cell lines. (B, D) WB analysis of COL8A1 knockdown efficiency in siNC and siCOL8A1 groups. (E, F) Colony formation assays evaluating proliferative capacity. (G, H) Cell invasion assays assessing migratory ability. (I) CCK-8 assay measuring cell viability. *p < 0.05; **p < 0.01; ***p < 0.001; n = 3.