Identification and validation of RAS signaling-related genes for prognostic prediction and immunological characterization in gastric cancer

Introduction

Gastric cancer (GC) ranks as the fifth leading cancer in terms of incidence and the fourth leading contributor to cancer-associated fatalities globally, accounting for over one million new diagnoses and approximately 770,000 deaths reported annually (1,2). The incidence of GC exhibits marked geographic variation, and East Asia, Central Europe, and Eastern Europe exhibit the highest rates of occurrence, making up almost 87% of all newly reported cases globally, where dietary habits, including excessive tobacco smoking, and Helicobacter pylori (H. pylori) infection contribute significantly to disease burden (3,4). Despite significant improvements in surgical techniques, chemotherapy regimens, targeted therapy, and the recent introduction of immune checkpoint (ICP) inhibitors, the outlook for individuals diagnosed with advanced or metastatic GC is still unfavorable, with a 5-year survival rate below 30% in most regions (5-8). Therefore, these findings underscore the urgent need for precise molecular biomarkers to enable specific therapeutic strategies and to improve clinical outcomes by prolonging survival and lowering recurrence rates.

The RAS signaling cascade is crucial in regulating cell growth, differentiation, and survival (9). As a key molecular factor of oncogenic signaling, the RAS pathway transduces extracellular signals primarily through the RAF-MEK-ERK and PI3K-AKT-mTOR signaling pathways (10,11). The RAS protein signaling network is encoded by three ubiquitously expressed genes—HRAS, KRAS, and NRAS—among which KRAS represents the predominant oncogenic isoform (12). In GC, hyperactivation of the RAS signaling pathway has been linked to increased tumor aggressiveness and metastasis (13). Moreover, studies have shown that the RAS mutation influences the tumor microenvironment, modulates immune cell infiltration, immune escape, and affects the efficacy of ICP inhibitors (14,15).

Given the multifaceted role in tumor biology, RAS signaling has emerged as a promising target for molecular subtyping, prognostic biomarker development, and potential therapeutic intervention in GC. However, a comprehensive characterization of RAS-related gene expression patterns and their immunological and clinical significance in GC remains underexplored. We present this article in accordance with the MDAR and TRIPOD reporting checklists (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1131/rc).

Methods

Data collection

Transcriptomic profiles and corresponding mutation and clinical data of stomach adenocarcinoma (STAD) were retrieved from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/), encompassing 36 normal and 412 tumor tissue samples. Additionally, the microarray dataset GSE62254, containing 300 tumor samples, was accessed from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/), which was applied as an independent cohort for validation. The RAS signaling-associated signature genes were extracted based on a previously published literature (16). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Analysis of differentially expressed genes (DEGs)

Differential expression analysis for TCGA-STAD samples was carried out with the “edgeR” package, applying a cutoff of |log fold change (FC)| >0.585, false discovery rate (FDR) <0.05. Similarly, “edgeR” was utilized to evaluate transcriptional differences across high- and low-risk groups within the STAD cohort, with more stringent criteria set at |logFC| >1 and FDR <0.05.

Screening hub genes prognosis-related trait modeling

Clinical data were integrated with the expression profiles of DEGs. Univariate Cox regression was performed utilizing the “survival” package. The least absolute shrinkage and selection operator (LASSO) regression was then carried out and enhance model generalizability by applying the “glmnet” package, and the optimal penalty parameter (lambda) was selected through cross-validation to eliminate highly correlated genes and reduce model complexity. To construct a prognostic model, multivariate Cox regression was conducted on the genes identified by LASSO, employing the “survival” R package.

Using the prognostic model, individual risk scores were derived by multiplying the gene expression levels by their respective regression coefficients and summing them. The STAD cohort was then divided into high- and low-risk groups based on the median risk score as the cutoff. To compare the overall survival between the two risk groups, Kaplan-Meier survival analysis was carried out. Receiver operating characteristic (ROC) curves were created with the “timeROC” R package, and the area under the curve (AUC) was computed for 1-, 3-, and 5-year survival to assess the model’s prediction accuracy.

Additionally, the distribution of risk scores, survival curves, and expression heatmaps of prognostic genes were visualized to compare the high- and low-risk groups. Kaplan-Meier survival curves and ROC curves were generated for the validation cohort to determine the prognostic model. Heatmaps illustrating score distributions, survival statuses, and gene expression levels were generated for the high- and low-risk groups.

Independent prognostic analysis

Risk score distributions were visualized using violin plots to compare different clinical subgroups. Univariate and multivariate Cox regression analyses incorporating both clinical variables and risk scores were performed to identify independent prognostic factors, and results were displayed as forest plots. A nomogram predicting 1-, 3-, and 5-year overall survival was constructed using the rms package. Calibration curves and decision curve analysis (DCA) were applied to evaluate prediction accuracy and clinical utility.

Immune cell infiltration analysis

The gene set variation analysis (GSVA) algorithm was used to evaluate immune cell infiltration and immune-related functions between high- and low-risk groups. Expression levels of ICP genes were compared using boxplots. The ESTIMATE algorithm was applied to calculate the immune score, stromal score, tumor purity, and ESTIMATE score.

Single-sample gene set enrichment analysis (ssGSEA) and the tumor immune dysfunction and exclusion (TIDE) algorithm were used to assess immune cell abundance and immune evasion potential. Immune infiltration was also quantified using the IOBR package, integrating both CIBERSORT and ssGSEA outputs to evaluate 22 immune cell types. Correlation analyses were performed to explore relationships among differentially enriched immune cells.

Drug sensitivity prediction

We explored the CellMiner database (https://discover.nci.nih.gov/cellminer/) to screen for antitumor compounds which related to the expression of prognostic genes to discover potential targets for therapy and more efficient antitumor drugs. Furthermore, we also estimated the half-maximal inhibitory concentration (IC₅₀) to predict differential drug effects using the “pRRophetic” R package.

Development of competing endogenous RNA (ceRNA) network

To explore the possible ceRNA regulatory network, we utilized the “multiMiR” R package to retrieve miRNAs that interacted with the identified signature genes from the mirdb database (https://mirdb.org/). Interaction data between miRNAs and lncRNAs were subsequently obtained from the ENCORI database (https://rnasysu.com/encori/); besides, lncRNAs with clipExpNum >10 were selected to ensure interaction reliability. The resulting mRNA-miRNA-lncRNA interaction network was illustrated with Cytoscape (version 3.10.2).

Analysis of consensus cluster

We conducted consensus clustering analysis with the “ConsensusClusterPlus” R package. Clustering was performed utilizing Euclidean distance and repeated over 1,000 iterations to ensure clustering stability and robustness. To assess prognostic variations among the identified subtypes, we then performed Kaplan-Meier survival analysis. To gain a deeper understanding of the immune landscape, we conducted ssGSEA using the “IOBR” package. The abundance of immune cell infiltration between tumor and normal samples was visualized through heatmaps and boxplots.

Cell lines

The human GC cell line AGS and the normal gastric epithelial cell line GSE1 were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). Both cell types were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and maintained at 37 ℃ in a humidified incubator containing 5% CO₂.

Quantitative real-time polymerase chain reaction (RT-qPCR)

Total RNA was extracted from the cultured cell lines using RNA Eazy Fast Cell Kit (TIANGEN Biotech. Co., Beijing, China) according to the manufacturer’s instructions. The purified RNA was subsequently reverse transcribed into complementary DNA (cDNA) using the FastKing RT Kit (TIANGEN Biotech. Co.). We performed real-time PCR using the SuperReal PreMix Plus (TIANGEN Biotech. Co.) and a quantitative real-time PCR system (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions. Each experiment was conducted with three independent biological replicates, and each sample was analyzed in technical triplicates. The relative expression levels in terms of FCs of target genes were calculated by the 2^−∆∆CT method. The sequences of all primers are shown in Table 1.

Statistical analysis

All analyses were conducted in R (v4.2.2). The Wilcoxon rank-sum test was used for two-group comparisons. Kaplan-Meier curves were compared with the log-rank test. Time-dependent ROC curves were constructed using the timeROC package. Statistical significance was defined as P<0.05. Significance levels are reported as: ****, P<0.00001; ***, P<0.0001; **, P<0.001; *, P<0.05; ns, not significant.

Results

Analysis of differential RAS-related genes

To investigate the involvement of the RAS signaling pathway in GC, we identified 7,585 DEGs from the TCGA-STAD dataset (Figure 1A), and the top 20 DEGs ranked by |logFC| were shown in Figure 1B. Then, we intersected DEGs with genes associated with the RAS signaling pathway, resulting in the identification of 95 differentially expressed RAS pathway-related genes (Figure 1C). To further explore the functional interactions among the 95 genes, a protein-protein interaction (PPI) network was developed, and consisted of 93 nodes and 955 edges (Figure 1D), suggesting extensive crosstalk among these genes within the cellular network. To elucidate the biological roles of the 95 RAS-related DEGs, we applied Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. A total of 853 GO terms were significantly enriched, including 749 biological processes (Figure 1E), 28 cellular components (Figure 1F), and 76 molecular functions (Figure 1G). Meanwhile, 93 significantly enriched pathways of KEGG enrichment analysis were yielded after excluding pathways categorized under human diseases (Figure 1H), further highlighting the diverse functional roles of the RAS pathway-related genes in GC progression.

Figure 1 Selection and primary analysis of candidate genes associated with RAS signaling-related genes. (A) The volcano plot displaying DEGs identified in the TCGA-STAD dataset. (B) Heatmap representation of the top 20 DEGs, ranked by (|logFC|). (C) UpSet plot illustrating the common genes across DEGs and RAS signaling-related genes. (D) PPI network of the 95 RAS signaling pathway-related DEGs. (E) Biological processes, (F) cellular components, and (G) molecular functions of GO according to RAS-related DEGs. (H) KEGG pathway enrichment results for 95 RAS-related DEGs. DEG, differentially expressed gene; EMT, epithelial-mesenchymal transition; FC, fold change; FDR, false discovery rate; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; NA, not applicable; PPI, protein-protein interaction; STAD, stomach adenocarcinoma; TCGA, The Cancer Genome Atlas.

Construction of the prognosis model based on RAS pathway-related genes

Univariate Cox regression identified 26 RAS-related genes significantly associated with overall survival (Figure 2A). Univariate Cox regression identified 26 RAS-related genes significantly associated with overall survival (Figure 2B,2C), and subsequent multivariate Cox regression identified six key genes (FOXO4, PLA1A, FGF1, FGF8, RIN1, and EGF) for inclusion in the prognostic model (Figure 2D). The risk score formula was: Risk Score = −0.339 × FOXO4 + 0.119 × PLA1A + 0.220 × FGF1 + (−0.150) × FGF8 + (−0.142) × RIN1 + 0.098 × EGF.

Figure 2 Identification of feature genes associated with GC prognosis. (A) Forest plot of the 95 RAS signaling-related DEGs. (B) Coefficient profiles of the signature genes generated across the log(λ) sequence. (C) LASSO coefficient spectrum of the 26 survival-related genes. (D) Forest plot of multivariate Cox regression analysis. (E) Expression levels of the six signature genes in TCGA-STAD samples. **, P<0.001; ***, P<0.0001; ns, not significant. CI, confidence interval; DEGs, differentially expressed genes; GC, gastric cancer; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; STAD, stomach adenocarcinoma; TCGA, The Cancer Genome Atlas.

In addition, analysis of the expression levels of the six prognostic signature genes was conducted in the TCGA-STAD cohort, revealing that GC samples showed a significant upregulation of FGF8, PLA1A, and RIN1 related to normal tissues, while FOXO4 expression was markedly reduced in tumor samples (Figure 2E). Briefly, these genes have established roles in cancer biology. FOXO4, a Forkhead box O family transcription factor, regulates cell proliferation, apoptosis, and oxidative stress, often acting as a tumor suppressor downregulated in GC (17-20). PLA1A, a phospholipase enzyme, maintains membrane integrity and lipid metabolism, with overexpression linked to adverse outcomes in GC and other cancers (21-25). FGF1, a fibroblast growth factor, promotes proliferation and angiogenesis, is elevated in GC tissues and associated with tumor progression (26-29). FGF8, another FGF family member, drives androgen-dependent tumor growth and is upregulated in various malignancies (30). RIN1, a RAS-interacting protein, modulates signaling pathways with context-dependent roles as suppressor or promoter in GC (31-35). EGF activates PI3K pathways via EGFR, enhancing GC migration and invasiveness (36,37).

Evaluation of prognostic model performance

According to the prognostic model constructed from six signature genes, we stratified GC patients into high- and low-risk groups. The AUC for the overall survival of 1-, 3-, and 5-year was 0.690, 0.664, and 0.715 in the TCGA-STAD database, and was 0.614, 0.659, and 0.659 in the GSE62254 validation database, respectively, indicating moderate predictive accuracy (Figure 3A,3B). According to Kaplan-Meier analysis, individuals classified as high-risk showed significantly reduced overall survival relative to the low-risk group in the TCGA-STAD training dataset and the GSE62254 validation dataset (Figure 3C,3D). Furthermore, the distribution of risk scores, survival status, and the expression heatmap provided a comprehensive visualization of the model’s prognostic stratification (Figure 3E-3H). To address the modest predictive power noted, we compared our model’s performance with previously published gene signatures in GC, as shown in Figure S1, where our RAS signaling-related model demonstrates superior AUC values at 1-year (0.69), 3-year (0.664), and 5-year (0.715) time points compared to models from Chen et al. [Aging (Albany NY)], Han et al. (BMC Cancer), and others, highlighting its enhanced clinical relevance.

Figure 3 Assessment of prognostic performance. (A) ROC curves for overall survival prediction in the TCGA-STAD training dataset. (B) ROC curves for overall survival prediction in the GSE62254 validation dataset. (C) Kaplan-Meier survival curves across high- and low-risk groups. (D) Kaplan-Meier survival curves across high- and low-risk groups. (E) Risk score distribution and survival status plot in the TCGA-STAD training dataset. (F) Risk score distribution and survival status plot in the GSE62254 validation dataset. (G) Heatmap showing the expression patterns of the six prognostic signature genes across high- and low-risk groups in the TCGA-STAD cohort. (H) Heatmap displaying the expression patterns of the six prognostic signature genes across two risk groups in the GSE62254 validation cohort. AUC, area under the curve; ROC, receiver operating characteristic; STAD, stomach adenocarcinoma; TCGA, The Cancer Genome Atlas.

Independent prognostic value of the model

Kaplan-Meier analysis revealed that lower FGF1 expression was associated with improved overall survival (Figure 4A). Univariate Cox regression indicated that T stage, N stage, and risk score were significantly associated with survival (Figure 4B), while multivariate analysis identified age and risk score as independent prognostic factors (Figure 4C). A nomogram integrating these variables showed good calibration for 1-, 3-, and 5-year survival (Figure 4D). Excellent alignment between the predicted and observed survival results was observed through the calibration curves (Figure 4E). Moreover, DCA curves showed that the nomogram provided greater net benefit, indicating the potential clinical utility (Figure 4F).

Figure 4 Independent prognostic value of the model. (A) Kaplan-Meier survival curves for the prognostic gene FGF1. (B) Forest plot of integrating risk score and clinical parameters. (C) Forest plot of multivariate Cox regression analysis. (D) Nomogram incorporating risk score and clinical variables to predict 1-, 3-, and 5-year overall survival probabilities in GC patients. (E) Calibration curves for 1-, 3-, and 5-year survival. (F) DCA curves showing the net clinical benefit of the prognostic model. DCA, decision curve analysis; f-1, f-3, f-5 represent DCA curves for predicting 1-, 3-, and 5-year survival; GC, gastric cancer; M, metastasis; N, node; T, tumor.

Analysis of the immune landscape in GC

To investigate the immune characteristics associated with the prognostic model, we first evaluated the immune infiltration levels in both risk groups through ssGSEA, which highlighted notable distinctions between the two groups (Figure 5A). The levels of immune cell infiltration indicated that several immune cell types, including effector memory CD8⁺ T cells, gamma delta T cells, immature dendritic cells, and type 1 T helper cells, were significantly upregulated within the low-risk group compared to the high-risk group (Figure 5B). Violin plot, displayed using the ESTIMATE algorithm, demonstrated that significantly higher tumor purity was observed in the low-risk group, while the high-risk group showed substantial increases in immune score, stromal score, and ESTIMATE score (Figure 5C). To further assess the potential for immune evasion, we compared TIDE scores across both groups, and the findings demonstrated that the high-risk group had notably higher TIDE scores compared to the low-risk group (Figure 5D).

Figure 5 Assessment of immune cell infiltration within the two risk groups. (A) Heatmap of immune cell infiltration and immune-related functions within both risk groups. (B) Boxplot showing the differential abundance of 28 immune cell types among the two groups. (C) Violin plots comparing ImmuneScore, StromalScore, TumorPurity, and ESTIMATEScore between the two groups. (D) Violin plot of TIDE scores. *, P<0.05; ns, not significant. MDSC, myeloid-derived suppressor cell; TIDE, tumor immune dysfunction and exclusion.

We also applied the CIBERSORT algorithm to estimate the expression levels of 22 immune cells across both groups, which showed distinct immune infiltration patterns across both groups (Figure 6A). Box plot showed significant differences in six types of immune cells: T follicular helper cells, T regulatory cells (Tregs), resting natural killer (NK) cells, and activated mast cells were more prevalent in the low-risk group, while the high-risk group exhibited increased levels of M2 macrophages and resting mast cells (Figure 6B). Moreover, 27 ICP gene investigations showed that all of these were notably elevated within the high-risk group compared to the low-risk group (Figure 6C). Finally, correlation heatmaps of differentially immune cells were generated using ssGSEA and CIBERSORT, respectively, revealing distinct immune interaction patterns within both groups (Figures 6D,6E).

Figure 6 Comprehensive analysis of immune cell infiltration and immune-related gene expression between high- and low-risk groups. (A) Heatmap showing the relative proportions of 22 immune cell types between the two groups. (B) Boxplots displaying differences in the proportions of immune cell types among the two groups. (C) Boxplot comparing the levels of expression for 27 ICP genes among the two groups. (D) Correlation heatmap of differentially abundant immune cells relying on the ssGSEA algorithm. (E) Correlation heatmap of differentially abundant immune cells based on the CIBERSORT algorithm. *, P<0.05; **, P<0.001; ***, P<0.0001; ns, not significant. ICP, immune checkpoint; NK, natural killer; ssGSEA, single-sample gene set enrichment analysis.

Enrichment analysis between high- and low-risk groups

Differential gene expression analysis was performed within both groups, which identified 1,368 genes with increased expression and 277 genes with decreased expression (Figure 7A). A heatmap illustrating the expression profiles of the leading 20 genes ranked by |logFC| is presented in Figure 7B. Subsequently, GO terms of the upregulated genes were mainly concentrated in epidermis development, skin development, and epidermal cell differentiation in biological processes, and cornified envelope, intermediate filament cytoskeleton, and intermediate filament in cellular components (Figure 7C). In contrast, downregulated genes were significantly enriched in biological processes, including muscle system process, regulation of membrane potential, and muscle contraction, and cellular component enrichment focused on synaptic membrane, collagen-containing extracellular matrix, and postsynaptic membrane (Figure 7D). Further, according to KEGG analysis, the estrogen signaling pathway, neuroactive ligand-receptor interaction, and Wnt signaling pathway were markedly associated with the upregulated gene set (Figure 7E); meanwhile, neuroactive ligand-receptor interaction, olfactory transduction, and cAMP signaling pathway were identified as the major pathways associated with the downregulated genes (Figure 7F). These results suggest that distinct biological functions and signaling pathways are involved in the tumor progression of both risk groups.

Figure 7 Enrichment analysis between both groups. (A) Volcano plot illustrating differential gene expression between both groups. (B) Heatmap showing the top 20 DEGs relying on (|logFC|) between both groups. (C) GO analysis identified biological functions associated with the upregulated DEGs. (D) GO analysis identified biological functions associated with the downregulated DEGs. (E) KEGG analysis was performed to identify signaling pathways enriched among the upregulated DEGs. (F) KEGG analysis was performed to identify signaling pathways enriched among the downregulated DEGs. DEGs, differentially expressed genes; FC, fold change; FDR, false discovery rate; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Analysis of drug sensitivity

To gain deeper insight into the underlying biological functions and therapeutic relevance of the characteristic genes, a ceRNA network including 6 key genes was constructed, reflecting the intricate regulatory interactions among the genes (Figure 8A). To explore possible therapeutic compounds aimed at modulating these key genes, we utilized the Drug-Gene Interaction Database (DGIdb) and visualized the predicted potential drugs, and several drugs were found to potentially interact with FGF1 and EGF, including pegbelfermin, aldafermin, and muparfostat related to FGF1, and nimotuzumab, EGF, and cetuximab associated with EGF (Figure 8B). Then, the association between TMB and the two risk groups was systematically examined, which indicated a significantly greater TMB in the low-risk group relative to the high-risk group, suggesting a stronger immunogenicity within low-risk patients (Figure 8C). We employed waterfall plots based on TMB data to compare the leading somatic mutations between the high- and low-risk groups, respectively (Figure 8D,8E). To assess potential therapeutic responses, we evaluated drug sensitivity between the two groups, and five drugs exhibited markedly different estimated IC₅₀ across the compared groups, including TGX221, rapamycin, pazopanib, GGP-60474, and A-77004 (Figure 8F). Additionally, the correlation plot between gene expression and drug response was predicted by the CellMiner database, and further supported the differential drug responsiveness of feature genes (Figure 9).

Figure 8 Drug sensitivity and mutational landscape analysis of the RAS-based prognostic model in GC. (A) ceRNA regulatory network of the characteristic genes. (B) Drug-gene interaction network predicted using the DGIdb database. (C) Violin plot visualization displaying TMB differences within both groups. (D) Waterfall plot illustrating the mutational landscape within the high-risk group. (E) Waterfall plot illustrating the mutational landscape within the low-risk group. (F) Violin plots contrasting the estimated IC50 values of five therapeutic agents within both groups. ceRNA, competing endogenous RNA; DGIdb, Drug-Gene Interaction Database; GC, gastric cancer; IC50, half-maximal inhibitory concentration; TMB, tumor mutation burden.

Figure 9 Correlation between key gene expression and drug sensitivity (IC50) in GC. Correlation analysis between gene expression and drug sensitivity visualized using CellMiner. GC, gastric cancer; IC50, half-maximal inhibitory concentration.

Analysis of consensus clustering

To further investigate the heterogeneity among STAD patients, consensus clustering was performed using the expression behavior of six characteristic genes. Findings indicated that individuals diagnosed with STAD patients were divided into two molecular subtypes (Figure 10A-10C). Besides, according to Kaplan-Meier analysis, the two subtypes exhibited notably different overall survival patterns, with cluster 1 showing a worse prognosis compared to cluster 2 (Figure 10D). A heatmap of immune cell infiltration illustrated distinct immune profiles between the two subtypes (Figure 10E). Box plots were utilized to visualize the enrichment levels of 28 distinct immune cell populations, and significant differences in the infiltration levels of 12 immune cell types were observed between the two subtypes, including activated B cells, activated CD4 T cells, CD56 dim NK cells, central memory CD8 T cells, effector memory CD4 T cells, eosinophils, mast cells, NK T cells, plasmacytoid dendritic cells, T follicular helper cells, type 17 T helper cells, and type 2 T helper cells (Figure 10F). The expression of ICP genes also showed a marked difference between cluster 1 and cluster 2, with lower expression observed in cluster 2 (Figure 10G). Using the ESTIMATE algorithm, cluster 2 was found to have a significantly higher tumor purity score, while cluster 1 showed notably elevated immune, stromal, and ESTIMATE scores, suggesting a tumor microenvironment with greater immune cell infiltration (Figure 10H).

Figure 10 Analysis of consensus clustering. (A) Consensus clustering heatmap when k=2, indicating stable subtype separation. (B) The CDF curve across various cluster numbers. (C) Relative change in area under the curve. (D) Survival curves for cluster 1 and cluster 2 molecular subtypes. (E) Heatmap illustrating the immune cell infiltration patterns across both subtypes. (F) Boxplots comparing the enrichment scores of 28 immune cell types between the two subtypes. (G) Boxplots showing the expression levels of ICP genes between the two subtypes. (H) Violin plots comparing ImmuneScore, StromalScore, TumorPurity, and ESTIMATEScore between the two subtypes. *, P<0.05; **, P<0.001; ***, P<0.0001; ****, P<0.00001; ns, not significant. CDF, cumulative distribution function; ICP, immune checkpoint; MDSC, myeloid-derived suppressor cell.

Validation of expression of feature genes by qRT-PCR

To validate the differential expression of the identified RAS signaling-related genes, qRT-PCR was performed on AGS and GSE1 cells. As depicted in Figure 11, the mRNA expression levels of FGF8, PLA1A, and RIN1 were significantly upregulated in GC samples compared to normal samples. Conversely, the mRNA expression of FOXO4 was significantly downregulated in GC samples. No significant differences in mRNA expression were observed for EGF and FGF1 between the normal and case groups. These findings suggest that FGF8, FOXO4, PLA1A, and RIN1 may serve as key RAS signaling-related genes involved in GC progression.

Figure 11 Relative mRNA expression of feature genes in GC and normal tissues. qRT-PCR analysis showing the relative mRNA expression levels of EGF, FGF1, FGF8, FOXO4, PLA1A, and RIN1 in normal and GC tissues. *, P<0.05; ns, not significant. GC, gastric cancer; qRT-PCR, quantitative real-time polymerase chain reaction.

Discussion

GC remains among the most common and deadly cancers globally, and approximately one million new diagnoses and 769,000 fatalities were reported in 2020, particularly in East Asia, where it continues to pose a significant public health burden (38,39). Thus, there is a critical demand for advanced therapeutic solutions to address the need for continuous understanding of GC pathogenesis, including ICP inhibitors, cellular immunotherapies, and cancer vaccines (40). Among the diverse signaling cascades implicated in GC pathogenesis, the RAS signaling pathway has attracted significant focus as a result of its crucial function in driving oncogenic transformation, tumor progression, and resistance to therapy (12). Given this, elucidating the prognosis and immunological implications of RAS pathway-related genes holds great promise for improving the clinical treatment of GC.

It is widely acknowledged that the RAS signaling pathway cascade serves as a key modulator in the control of cellular proliferation, differentiation, and survival, and aberrant activity of the RAS signaling pathway has been associated with the development of multiple malignancies, such as GC (41,42). In this study, we systematically identified 95 DEGs associated with the RAS pathway based on analysis of the TCGA-STAD dataset, suggesting that RAS-associated genes may be crucial participants in GC progression. Enrichment analysis highlighted the participation of these genes in several key molecular pathways, such as Ras, MAPK, and PI3K-Akt signaling, consistent with earlier findings that aberrant RAS activity promotes tumorigenesis through multiple mechanisms (43-45).

We developed a predictive model with strong prognostic value based on six RAS-associated key genes, including FOXO4, PLA1A, FGF1, FGF8, RIN1, and EGF, which demonstrated excellent predictive performance in both the TCGA and GSE62254 cohorts. Notably, a number of these genes have shown independent correlations with cancer progression and unfavorable clinical outcomes. As briefly introduced in the “Results” section, FOXO4 acts as a tumor suppressor in GC. Increasing evidence suggests that FOXO4 functions as a context-dependent tumor suppressor in various malignancies, including non-small cell lung cancer, breast cancer, gallbladder cancer, and GC (46-49). In GC, FOXO4 expression is often downregulated and has been correlated with poor clinical outcomes (50). Studies have shown that FOXO4 can antagonize RAS-induced oncogenic transformation, suggesting a functional interplay between FOXO4 activity and RAS pathway modulation (51). As noted in “Results”, PLA1A is overexpressed in GC and linked to adverse outcomes. Emerging studies have implicated PLA1A in various cancers, including melanoma, lung, prostate cancer, and GC, and overexpression of this gene is commonly correlated with adverse clinical outcomes, highlighting its potential as a novel therapeutic target (23-25,52). FGF1, as noted in “Results”, promotes proliferation and angiogenesis in GC. In GC, tumor tissues showed significantly elevated FGF1 expression compared to neighboring non-tumorous tissues, and another research revealed that the suppression of VEGFA and FGF1 expression by MiR-205-5p leads to reduced angiogenesis in GC (29,53). These results revealed that FGF1 may act as a promising biomarker and therapeutic target. As briefly introduced in the “Results” section, FGF8 drives tumor growth in various malignancies. In prostate cancer, the upregulation of FGF8, driven by androgen signaling, indicated a potential role in the development and maintenance of androgen-dependent tumors (54). RIN1, as highlighted in “Results”, modulates signaling with context-dependent roles in GC. Recent studies have elucidated its diverse roles across different cancer types, including breast, thyroid carcinoma, head and neck tumors, and colorectal cancer, highlighting the context-dependent functions as either a tumor suppressor or promoter (32-34). Further studies have also demonstrated that the expression levels of RIN1 modulate tumor activity in GC (35). EGF, as described in “Results”, activates pathways enhancing GC invasiveness. Some studies have revealed that evodiamine restricts GC cell migration and invasiveness by enhancing PTEN expression, thereby hindering the activation of the PI3K cascade initiated by EGF, highlighting the pivotal role of EGF in promoting tumor aggressiveness (55). These findings support the biological relevance of our model and imply a potential role for these genes in prognosis prediction and therapeutic intervention.

Additionally, the immune landscape analysis revealed distinct immunological profiles across both risk groups. Elevated levels of effector memory CD8 T cells, gamma delta T cells, immature dendritic cells, and type 1 T helper cells were observed within the low-risk group, indicating a more active anti-tumor immunological reaction, while elevated TIDE values and enhanced expression of ICPs were within the high-risk group, suggestive of increased immune evasion potential (56-60). These findings highlight the prognostic implications of tumor immunogenicity in GC.

Furthermore, the drug sensitivity analysis provides additional insights into the potential therapeutic effects of GC. We observed that several small-molecule inhibitors, including TGX221, rapamycin, pazopanib, CGP-60474, and A-770041, had significantly decreased IC₅₀ levels, which were observed within the low-risk group, implying potential therapeutic benefit. The integration of TMB analysis further confirmed that low-risk patients harbored significantly higher mutation rates, and high TMB means significant clinical benefits, which may render them more responsive to immunotherapy or targeted treatments (61). Ultimately, the findings of this research highlight the importance of RAS pathway-associated genes in the prognosis and immune response of GC. The six signature genes serve as a promising biomarker for risk stratification and offer crucial insights for tailoring immunotherapy and targeted therapy strategies. Prospective studies should focus on experimental validation and clinical translation of these findings.

While our study provides novel insights into RAS signaling-related genes in GC through a multifaceted prognostic model integrating immune landscape analysis, TMB, drug sensitivity predictions, ceRNA networks, and consensus clustering—distinguishing it from existing transcriptomic models that often focus on general DEGs or singular aspects like mitochondrial or cuproptosis-related signatures—several limitations must be acknowledged. To partially address the speculative nature of the ceRNA and drug-target networks, we incorporated qRT-PCR experiments to validate the differential expression of the six signature genes in GC cell lines versus normal gastric epithelial cells, confirming bioinformatics findings for most genes. However, this represents preliminary validation, and further functional experiments (e.g., knockdown/overexpression assays or in vivo models) are required to elucidate the mechanistic roles in GC pathogenesis. Additionally, the retrospective nature of our analysis, relying on public databases like TCGA and GEO, may introduce selection bias and limit generalizability to diverse populations. The model’s modest predictive power (AUC 0.61–0.71) raises questions about clinical utility, particularly without large-scale prospective validation, and potential overfitting during LASSO-based development warrants caution. The absence of in vitro or prospective clinical confirmation further constrains the translational applicability of our findings. Future studies should prioritize multicenter prospective cohorts, advanced functional validations, and external datasets to overcome these limitations and enhance the model’s robustness for precision oncology in GC.

Conclusions

In the current investigation, we systematically identified and confirmed RAS signaling-related genes, and six key prognostic genes (FOXO4, PLA1A, FGF1, FGF8, RIN1, and EGF) were ultimately selected to construct a robust prognostic model, which provided novel insights into the predictive value and immune implications of genes associated with the RAS signaling pathway in GC, offering potential diagnostic markers and therapeutic interventions for precision medicine in oncology. Despite the comprehensive analysis and validation carried out in this investigation, a few limitations need to be considered, including the lack of prospective multicenter validation and experimental verification, and the underlying mechanisms and therapeutic targets require further investigation.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the MDAR and TRIPOD reporting checklists. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1131/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1131/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1131/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1131/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Inoue M. Epidemiology of Gastric Cancer-Changing Trends and Global Disparities. Cancers (Basel) 2024;16:2948. [Crossref] [PubMed]
3. Röcken C. Predictive biomarkers in gastric cancer. J Cancer Res Clin Oncol 2023;149:467-81. [Crossref] [PubMed]
4. Smyth EC, Nilsson M, Grabsch HI, et al. Gastric cancer. Lancet 2020;396:635-48. [Crossref] [PubMed]
5. Mazurek M, Szewc M, Sitarz MZ, et al. Gastric Cancer: An Up-to-Date Review with New Insights into Early-Onset Gastric Cancer. Cancers (Basel) 2024;16:3163. [Crossref] [PubMed]
6. Guan WL, He Y, Xu RH. Gastric cancer treatment: recent progress and future perspectives. J Hematol Oncol 2023;16:57. [Crossref] [PubMed]
7. Burz C, Pop V, Silaghi C, et al. Prognosis and Treatment of Gastric Cancer: A 2024 Update. Cancers (Basel) 2024;16:1708. [Crossref] [PubMed]
8. Ji X, Wang G, Pan D, et al. Efficacy and safety of pembrolizumab in advanced gastric and gastroesophageal junction cancer: a systematic review and meta-analysis. BMC Gastroenterol 2025;25:173. [Crossref] [PubMed]
9. Simanshu DK, Nissley DV, McCormick F. RAS Proteins and Their Regulators in Human Disease. Cell 2017;170:17-33. [Crossref] [PubMed]
10. Bahar ME, Kim HJ, Kim DR. Targeting the RAS/RAF/MAPK pathway for cancer therapy: from mechanism to clinical studies. Signal Transduct Target Ther 2023;8:455. [Crossref] [PubMed]
11. Norton C, Shaw MS, Rubnitz Z, et al. KRAS Mutation Status and Treatment Outcomes in Patients With Metastatic Pancreatic Adenocarcinoma. JAMA Netw Open 2025;8:e2453588. [Crossref] [PubMed]
12. Yang X, Wu H. RAS signaling in carcinogenesis, cancer therapy and resistance mechanisms. J Hematol Oncol 2024;17:108. [Crossref] [PubMed]
13. Won Y, Choi E. Mouse models of Kras activation in gastric cancer. Exp Mol Med 2022;54:1793-8. [Crossref] [PubMed]
14. Liu Y, Xie B, Chen Q. RAS signaling and immune cells: a sinister crosstalk in the tumor microenvironment. J Transl Med 2023;21:595. [Crossref] [PubMed]
15. Bezu L, Kepp O, Kroemer G. Immunogenic chemotherapy sensitizes RAS-mutated colorectal cancers to immune checkpoint inhibitors. Oncoimmunology 2023;12:2272352. [Crossref] [PubMed]
16. Luo CJ, Abudukeremu Y, Rao ML, et al. A Novel Prognostic Risk-Scoring Model Based on RAS Gene-Associated Cluster in Pediatric Acute Myeloid Leukemia. Cancer Med 2025;14:e70716. [Crossref] [PubMed]
17. Kong YX, Li ZS, Liu YB, et al. FOXO4-DRI induces keloid senescent fibroblast apoptosis by promoting nuclear exclusion of upregulated p53-serine 15 phosphorylation. Commun Biol 2025;8:299. [Crossref] [PubMed]
18. Intuyod K, Chomwong S, Thongpon P, et al. Expression of FOXO4 Inhibits Cholangiocarcinoma Cell Proliferation In Vitro via Induction of G(0)/G(1) Arrest. Anticancer Res 2020;40:6899-905. [Crossref] [PubMed]
19. Liu W, Li Y, Luo B. Current perspective on the regulation of FOXO4 and its role in disease progression. Cell Mol Life Sci 2020;77:651-63. [Crossref] [PubMed]
20. Orea-Soufi A, Paik J, Bragança J, et al. FOXO transcription factors as therapeutic targets in human diseases. Trends Pharmacol Sci 2022;43:1070-84. [Crossref] [PubMed]
21. Wepy JA, Galligan JJ, Kingsley PJ, et al. Lysophospholipases cooperate to mediate lipid homeostasis and lysophospholipid signaling. J Lipid Res 2019;60:360-74. [Crossref] [PubMed]
22. Yang G, Liu S, Maghsoudloo M, et al. PLA1A expression as a diagnostic marker of BRAF-mutant metastasis in melanoma cancer. Sci Rep 2021;11:6056. [Crossref] [PubMed]
23. Zhou Y, Chang M, Wang N, et al. Phosphatidylserine-Specific Phospholipase A1 Limits Aggressiveness of Lung Adenocarcinoma by Lysophosphatidylserine and Protein Kinase A-Dependent Pathway. Am J Pathol 2022;192:970-83. [Crossref] [PubMed]
24. Paulo P, Ribeiro FR, Santos J, et al. Molecular subtyping of primary prostate cancer reveals specific and shared target genes of different ETS rearrangements. Neoplasia 2012;14:600-11. [Crossref] [PubMed]
25. Sullivan KM, Li H, Yang A, et al. Tumor and Peritoneum-Associated Macrophage Gene Signature as a Novel Molecular Biomarker in Gastric Cancer. Int J Mol Sci 2024;25:4117. [Crossref] [PubMed]
26. Zou Y, Hu J, Huang W, et al. Non-Mitogenic Fibroblast Growth Factor 1 Enhanced Angiogenesis Following Ischemic Stroke by Regulating the Sphingosine-1-Phosphate 1 Pathway. Front Pharmacol 2020;11:59. [Crossref] [PubMed]
27. Castillo-Castrejon M, Sankofi BM, Murguia SJ, et al. FGF1 supports glycolytic metabolism through the estrogen receptor in endocrine-resistant and obesity-associated breast cancer. Breast Cancer Res 2023;25:99. [Crossref] [PubMed]
28. Liu N, Zhang J, Sun S, et al. Expression and clinical significance of fibroblast growth factor 1 in gastric adenocarcinoma. Onco Targets Ther 2015;8:615-21. [Crossref] [PubMed]
29. Zhang J, Zhang J, Pang X, et al. MiR-205-5p suppresses angiogenesis in gastric cancer by downregulating the expression of VEGFA and FGF1. Exp Cell Res 2021;404:112579. [Crossref] [PubMed]
30. Gnanapragasam VJ, Robson CN, Neal DE, et al. Regulation of FGF8 expression by the androgen receptor in human prostate cancer. Oncogene 2002;21:5069-80. [Crossref] [PubMed]
31. Zhang W, Veisaga ML, Barbieri MA. Role of RIN1 on telomerase activity driven by EGF-Ras mediated signaling in breast cancer. Exp Cell Res 2020;396:112318. [Crossref] [PubMed]
32. Picello L, Dalle Nogare M, Puggina D, et al. Exploring the tumor suppressor role of RIN1 in familial thyroid carcinoma. Endocr Relat Cancer 2025;32:e240344. [Crossref] [PubMed]
33. Osei Saahene R, Barnes P, Yeboah FA, et al. Expression of RAS and RAB interactor 1 (RIN1) in head and neck tumors at selected hospital in Ghana. PLoS One 2024;19:e0301295. [Crossref] [PubMed]
34. Wang Y, Zhang J, Zheng CC, et al. C20orf24 promotes colorectal cancer progression by recruiting Rin1 to activate Rab5-mediated mitogen-activated protein kinase/extracellular signal-regulated kinase signalling. Clin Transl Med 2022;12:e796. [Crossref] [PubMed]
35. Zhou B, Tie P, Li D, et al. Unveiling the Power of Mitochondrial Fission and Fusion: A Five-Gene Signature for Personalized Prognosis in Gastric Cancer. Curr Med Chem 2024; Epub ahead of print. [Crossref]
36. Ye J, Li J, Zhao P. The Silkworm Carboxypeptidase Inhibitor Prevents Gastric Cancer Cells' Proliferation through the EGF/EGFR Signaling Pathway. Int J Mol Sci 2023;24:1078. [Crossref] [PubMed]
37. Yang R, Wen J, Yang T, et al. Evodiamine Inhibits Gastric Cancer Cell Proliferation via PTEN-Mediated EGF/PI3K Signaling Pathway. Evid Based Complement Alternat Med 2021;2021:5570831. [Crossref] [PubMed]
38. Xia JY, Aadam AA. Advances in screening and detection of gastric cancer. J Surg Oncol 2022;125:1104-9. [Crossref] [PubMed]
39. Mousavi SE, Ilaghi M, Elahi Vahed I, et al. Epidemiology and socioeconomic correlates of gastric cancer in Asia: results from the GLOBOCAN 2020 data and projections from 2020 to 2040. Sci Rep 2025;15:6529. [Crossref] [PubMed]
40. Yang WJ, Zhao HP, Yu Y, et al. Updates on global epidemiology, risk and prognostic factors of gastric cancer. World J Gastroenterol 2023;29:2452-68. [Crossref] [PubMed]
41. Jiang J, Jiang L, Maldonato BJ, et al. Translational and Therapeutic Evaluation of RAS-GTP Inhibition by RMC-6236 in RAS-Driven Cancers. Cancer Discov 2024;14:994-1017. [Crossref] [PubMed]
42. Xie M, Zhang L, Han L, et al. The ASH1L-AS1-ASH1L axis controls NME1-mediated activation of the RAS signaling in gastric cancer. Oncogene 2023;42:3435-45. [Crossref] [PubMed]
43. Yang Y, Yuan L, Du Y, et al. Network pharmacology and in vitro experiments to investigate the anti-gastric cancer effects of paeoniflorin through the RAS/MAPK signaling pathway. Discov Oncol 2024;15:659. [Crossref] [PubMed]
44. Liu H, Wang H, Dong A, et al. The Inhibition of Gastric Cancer Cells' Progression by 23,24-Dihydrocucurbitacin E through Disruption of the Ras/Raf/ERK/MMP9 Signaling Pathway. Molecules 2022;27:2697. [Crossref] [PubMed]
45. Morgos DT, Stefani C, Miricescu D, et al. Targeting PI3K/AKT/mTOR and MAPK Signaling Pathways in Gastric Cancer. Int J Mol Sci 2024;25:1848. [Crossref] [PubMed]
46. Meng J, Li Y, Wan C, et al. Targeting senescence-like fibroblasts radiosensitizes non-small cell lung cancer and reduces radiation-induced pulmonary fibrosis. JCI Insight 2021;6:e146334. [Crossref] [PubMed]
47. Zhu Y, Zhang C, Yin Q, et al. FOXO4 suppresses cisplatin resistance of triple-negative breast cancer by inhibiting autophagy. Am J Med Sci 2025;369:252-63. [Crossref] [PubMed]
48. Zhao C, Zhang J, Yang ZY, et al. Ponicidin inhibited gallbladder cancer proliferation and metastasis by decreasing MAGEB2 expression through FOXO4. Phytomedicine 2023;114:154785. [Crossref] [PubMed]
49. Wang XH, Jiang ZH, Yang HM, et al. Hypoxia-induced FOXO4/LDHA axis modulates gastric cancer cell glycolysis and progression. Clin Transl Med 2021;11:e279. [Crossref] [PubMed]
50. Qi J, Cui D, Wu QN, et al. Targeting Wnt/β-Catenin Signaling by TET1/FOXO4 Inhibits Metastatic Spreading and Self-Renewal of Cancer Stem Cells in Gastric Cancer. Cancers (Basel) 2022;14:3232. [Crossref] [PubMed]
51. Zhang F, Virshup DM, Cheong JK. Oncogenic RAS-induced CK1α drives nuclear FOXO proteolysis. Oncogene 2018;37:363-76. [Crossref] [PubMed]
52. Imani S, Roozitalab G, Emadi M, et al. The evolution of BRAF-targeted therapies in melanoma: overcoming hurdles and unleashing novel strategies. Front Oncol 2024;14:1504142. [Crossref] [PubMed]
53. Rejali L, Piroozkhah M, Jahanbin M, et al. Comprehensive analysis identifies endocrine fibroblast growth factors as promising prognostic markers for colorectal carcinoma. Sci Rep 2024;14:28754. [Crossref] [PubMed]
54. Cruz-Hernández CD, Smith B, Billet S, et al. A Cyanobacteria-Derived RNA Aptamer Resensitizes Prostate Cancer to Hormone Therapy. Cancer Res 2025;85:2714-25. [Crossref] [PubMed]
55. Zhou Z, Zhou Y, Zhang Z, et al. Progress on the effects and underlying mechanisms of evodiamine in digestive system diseases, and its toxicity: A systematic review and meta-analysis. Phytomedicine 2024;132:155851. [Crossref] [PubMed]
56. Gebhardt T, Park SL, Parish IA. Stem-like exhausted and memory CD8(+) T cells in cancer. Nat Rev Cancer 2023;23:780-98. [Crossref] [PubMed]
57. Saura-Esteller J, de Jong M, King LA, et al. Gamma Delta T-Cell Based Cancer Immunotherapy: Past-Present-Future. Front Immunol 2022;13:915837. [Crossref] [PubMed]
58. Leblanc-Hotte A, Audiger C, Chabot-Roy G, et al. Immature and mature bone marrow-derived dendritic cells exhibit distinct intracellular mechanical properties. Sci Rep 2023;13:1967. [Crossref] [PubMed]
59. Hou Y, Sun L, LaFleur MW, et al. Neuropeptide signalling orchestrates T cell differentiation. Nature 2024;635:444-52. [Crossref] [PubMed]
60. Huang X, Liu Y, Qian C, et al. CHSY3 promotes proliferation and migration in gastric cancer and is associated with immune infiltration. J Transl Med 2023;21:474. [Crossref] [PubMed]
61. Wang Z, Huang D, Li S, et al. Role of tumor mutational burden in patients with urothelial carcinoma treated with immune checkpoint inhibitors: a systematic review and meta-analysis. Front Immunol 2025;16:1592761. [Crossref] [PubMed]