Development and validation of a gene signature for gastric cancer prognosis based on propionate metabolism

Abstract

Background

Gastric cancer (GC), the fifth most common cancer worldwide, has high morbidity and mortality rates. Propionate metabolism, which plays a significant role in cancer progression, remains understudied in the context of GC development and progression. This study aimed to identify propanoate metabolism-related genes (PMRGs) with prognostic value, construct a predictive model for GC outcomes, and explore their associations with tumor microenvironment (TME) and therapeutic response.

Methods

The data of GC patients included RNA-sequencing expression profiles, the clinical data, and mutation data gained from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO). Data from GC patients in TCGA and GEO were used to develop and validate a prognostic model through univariate Cox and least absolute shrinkage and selection operator (LASSO) regression, by selecting key differentially expressed PMRGs (DE-PMRGs). Patients were divided into high- and low-risk groups by median risk scores. The prognostic model was evaluated by the Kaplan-Meier (K-M) curve and time-dependent receiver operating characteristic (ROC). A nomogram was created using PMRGs to predict 1-, 2-, and 3-year overall survival (OS), which was further validated with decision curve analysis (DCA) and calibration curves.

Results

There were 63 DE-PMRGs were identified, with eight key hub PMRGs selected by LASSO and Cox regression for a prognostic model that predicted better survival in the low-risk group. A nomogram was constructed to predict the 1, 2 and 3 years survival rates of GC patients. In the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of high-risk groups, DE-PMRGs are primarily enriched in pathways related to muscle cells and cardiac diseases. Additionally, five immune cell types showed disparities between high and low-risk groups. Immune infiltration analysis suggested a higher potential for immune therapy response in the low-risk cohorts, and drug sensitivity prediction in GC indicates a broader sensitivity to chemotherapy drugs in the high-risk groups.

Conclusions

We developed a prognostic model based on PMRGs to forecast clinical outcomes in GC patients independently. The model was refined to identify eight key genes, serving as a tool for prognosis and treatment planning in GC.

Highlight box.

Key findings

• Developed an 8-gene prognostic model based on propanoate metabolism-related genes (PMRGs) for gastric cancer (GC).

• Low-risk GC patients showed better survival and enhanced immunotherapy response.

• High-risk patients exhibited distinct tumor microenvironment and chemosensitivity patterns.

What is known and what is new?

• Propionate metabolism influences cancer progression, but its role in GC remains unclear.

• This study establishes the first PMRG-based prognostic model for GC, linking propionate metabolism to immune infiltration and therapy response.

What is the implication, and what should change now?

• This 8-gene signature may assist in risk stratification, immunotherapy selection, and personalized chemotherapy for GC patients. However, clinical validation of the model and elucidation of propionate metabolism’s mechanistic role in GC progression remain necessary.

Introduction

Gastric cancer (GC), ranking as the fifth most prevalent malignancy worldwide, is associated with substantial morbidity and mortality rates (1). Although multimodal therapies have advanced, early diagnosis continues to pose difficulties, leading to many patients being identified only at advanced stages when treatment choices are restricted (2,3). Various gene prediction models related to GC have emerged, including those associated with autophagy and senescence (4), copper metabolism mechanisms (5), and M2 macrophages (6). However, early GC is underdiagnosed, the detection rate of early-stage GC remains low, and the mortality rate of advanced GC remains high (7). This underscores the urgent need to identify novel biomarkers that are both specific and sensitive, and that provide prognostic value for the diagnosis and treatment of GC. Such advancements could significantly improve patient outcomes and survival rates.

Propionate, a short-chain fatty acid (SCFA) predominantly produced in the large intestine, exerts metabolic effects that extend beyond its site of origin, influencing the entire digestive tract, including the stomach (8). Chronic inflammation is a key risk factor for the development of GC (7), and SCFAs, including propionate, possess anti-inflammatory properties that may reduce cancer risk by mitigating chronic inflammation (9). Recent studies have shown that dysregulation of propionate metabolism leads to the accumulation of methylmalonic acid (MMA). This accumulation of MMA may affect the tumor microenvironment (TME) and the invasiveness of cancer cells by altering the extracellular matrix, activating inflammatory signaling pathways, inducing epithelial-mesenchymal transition (EMT), and modulating the functions of immune cells (10). Although this mechanism may also be applicable to GC, the exact role of propionate metabolism-related genes (PMRGs) in the pathogenesis of GC remains unclear. However, their dysregulation may affect the TME and disease progression. Thus, unraveling the complex interplay between propionate metabolism and GC could offer new insights for preventive and therapeutic approaches against this highly aggressive cancer.

In this study, we utilized genomic datasets to elucidate the role of PMRGs in GC. By employing univariate Cox regression and least absolute shrinkage and selection operator (LASSO) regression, we constructed a prognostic model based on PMRGs that can independently predict outcomes for GC patients. This model has the potential to serve as a predictive tool, offering new insights into the development of therapeutic strategies for GC and significantly impacting patient prognosis and treatment planning. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-390/rc).

Methods

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data collection

We obtained 443 RNA sequencing data, clinical profiles, and mutation data from GC patients through The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov). Additionally, we sourced the probe matrix file (GSE84437) and the platform file (GPL6947) from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo). Furthermore, we searched the GeneCards database (https://www.genecards.org) using “propionate metabolism” as a keyword and selected 237 propionate-related genes with a relevance score greater than 11 (available online: https://cdn.amegroups.cn/static/public/10.21037tcr-2025-390-1.xlsx).

Identification of differentially expressed propanoate metabolism-related genes

We employed the ‘limma’ R package to conduct differential expression analysis between normal and tumor samples, selecting differentially expressed genes (DEGs) using criteria of |log₂ fold change (FC)| >1 and an adjusted P value <0.05. Heatmaps and volcano plots were created to visualize the DEGs. Subsequently, the expression profiles of these DEGs were extracted and integrated with survival data. A univariate Cox proportional hazards model was then applied to ascertain genes significantly associated with patient prognosis. Using the “venn” package, we identified the intersection between DEGs and prognostic genes, yielding differentially expressed propionate metabolism-related genes (DE-PMRGs). A correlation circle plot was used to visualize the relationships among these genes.

Development and validation of a survival risk model for GC

To construct the prognostic model, we used the TCGA dataset as the training cohort and the GEO dataset for validation as the testing cohort. The model was developed by deriving a risk score equation: Risk score = Coef1 × ExpGene1 + Coef2 × ExpGene2 + ... + Coefn × ExpGenen, where Coef represents the regression coefficients of prognostic genes, and ExpGene denotes their expression levels. Patients were then stratified into high- and low-risk groups based on the median risk score.

Survival disparities between the high-risk and low-risk groups were assessed using Kaplan-Meier (K-M) curve analysis. The prognostic capability of the model was evaluated through time-dependent receiver operating characteristic (ROC) analysis. To further validate our findings, the risk score formula was applied to GC patients within the GEO database. K-M plots were generated using the “survminer” R package, and ROC curves were produced using the “survivalROC” R package.

Nomogram construction

The risk scores were compared with various clinical characteristics, including age, sex, grade, and TNM stage. A nomogram was then constructed incorporating these factors to predict survival rates. Calibration curves and decision curve analyses (DCAs) were developed to ensure the model’s predictive validity.

Functional analysis

In our study, we conducted gene set enrichment analysis (GSEA) to investigate the functional profiles of genes across different risk groups. Subsequently, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of differentially DE-PMRGs using the “clusterProfiler” R package, with criteria of count ≥1 and adjusted P value <0.05.

Immune infiltration analysis

In this study, utilizing the CIBERSORT algorithm, we assessed differential immune cell infiltration, evaluated correlations with immune cell types, and quantified their abundance. We then applied the Wilcoxon test to identify differences in the expression of immune checkpoint genes between high- and low-risk patient cohorts. We retrieved the files detailing immune cell infiltration from the Timer2.0 database (http://timer.comp-genomics.org/). Immune checkpoint genes were identified through a comprehensive literature review (11).

Predicting response to immunotherapy

To predict the response to immunotherapy in GC patients, we initially utilized the “ggpubr” package to investigate the correlation between tumor mutational burden (TMB) and risk scores in GC patients, visualized through boxplots and scatter plots. Subsequently, we collected immune prognostic scores (IPS) from The Cancer Immune Atlas (TCIA) database (https://tcia.at/home) and utilized the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm to stratify patients into high and low-risk groups for immunotherapy response.

Estimating drug sensitivity

To predict drug sensitivity in high- and low-risk GC patient groups, we employed the oncoPredict package, achieving significant results with a P value <0.001.

Statistical analysis

Statistical analyses were performed using R version 4.4.1 and Strawberry Perl 5.30.0.1. Survival differences among groups were examined via Kaplan-Meier analysis, while univariate and multivariate Cox regression analyses were employed to identify independent prognostic factors for GC. A P value <0.05 was considered statistically significant.

Results

Identification of differentially expressed propanoate metabolism-related genes

We compared gene expression levels between normal gastric and cancerous tissue samples from the TCGA database, identifying 63 DEGs associated with propionate metabolism-related. Among these, 43 genes were significantly upregulated and 20 downregulated in tumors. The findings are visualized through heatmap and volcano plots (Figure 1A,1B).

Figure 1.

Identification of PMRGs in GC. (A) Heatmap of differentially expressed PMRGs; (B) Volcano map of PMRGs with different expressions. Green indicates downregulation; red indicates upregulation (|logFC| >1.5, P<0.05); (C) the 43 PMRGs associated with GC prognosis; (D) the Venn diagram of 16 DE-PMRGs; (E) LASSO Cox regression analysis; (F) a signature was constructed using 8 identified genes. CI, confidence interval; DE-PMRGs, differentially expressed propionate metabolism-related genes; FC, fold change; FDR, false discovery rate; GC, gastric cancer; LASSO, least absolute shrinkage and selection operator; PMRGs, propanoate metabolism-related genes.

Constructing the survival risk and prognostic models of GC

Utilizing univariate Cox regression analysis, we selected 43 propionic acid metabolism-related genes significantly associated with GC prognosis (P<0.05) from 63 DEGs (Figure 1C). Subsequently, Venn diagrams identified 16 common genes (Figure 1D). Finally, LASSO Cox regression further refined 8 hub PMRGs (APOA1, GAMT, GCG, CD36, HBB, SERPINE1, BCHE, GNAS) for the prognostic model (Figure 1E,1F). The prognostic model was constructed based on the following risk score formula. Risk score = (0.0341099835455906) × APOA1 + (0.108688772593843) × GAMT + (0.11064420065487) × GCG + (0.0525716947531768) × CD36 + (0.026517124807422) × HBB + (0.178828166011863) × SERPINE1 + (0.0709157042300196) × BCHE + (0.0928920118000936) × GNAS.

GC Patients in the training and testing sets were stratified into high- and low-risk groups based on the median risk score. Survival status and risk scores are depicted in Figure 2A-2D with the expression of the 8 core PMRGs shown as a heatmap (Figure 2E,2F). Survival and risk curves reveal longer survival times for the low-risk group (Figure 2G). The area under the curves (AUCs) for predicting 1-, 3-, and 5-year survival rates were 0.654, 0.718, and 0.736 (Figure 2H). Concurrently, validation with the test group demonstrated that the clinical outcomes for the low-risk group were superior to those of the high-risk group (Figure 2I,2J). Univariate and multivariate Cox analyses of the training set confirmed age, stage, and risk score as independent prognostic factors for GC patients (Figure 3A,3B). Finally, clinical stratification analyses across various factors, including age, gender, and stage, showed consistently superior survival in the low-risk subgroup (Figure 3C-3H).

Figure 2.

Evaluation of the predictive values of the risk model. (A,B) Patient survival according to the risk score; (C,D) the risk score of each patient with GC; (E,F) heatmap of PMRGs expression in the high- and low-risk group; (G) OS of the propionate metabolism risk score in the TCGA-STAD cohort; (H) predicted 1-, 3-, and 5-year AUC values for the TCGA-STAD cohort; (I) OS of the propionate metabolism risk score in the validation cohort; (J) predictive ROC curves for the validation cohort. AUC, area under the curve; GC, gastric cancer; GEO, Gene Expression Omnibus; OS, overall survival; PMRGs, propanoate metabolism-related genes; ROC, receiver operating characteristic; STAD, stomach adenocarcinoma; TCGA, The Cancer Genome Atlas.

Figure 3.

Independent prognostic analysis and clinical relevance assessment. (A) Univariate Cox analysis; (B) multivariate Cox analysis; (C-H) K-M curves for OS stratified by gender, age, and stage in different risk groups. CI, confidence interval; K-M, Kaplan-Meier; OS, overall survival.

Development of a nomogram

A nomogram incorporating age, gender, grade, and tumor-node-metastasis (TNM) stage, accurately predicts 1-, 2-, and 3-year survival rates in GC patients (Figure 4A). The slope of the survival probability in the calibration curve is approximately 1, confirming the accuracy of its predictions (Figure 4B). DCA further validates that the nomogram is superior to other clinical predictors in terms of survival estimation (Figure 4C).

Figure 4.

Construction of nomogram. (A) Nomogram to predict 1-, 2-, and 3-year OS in GC patients; (B) the calibration curve; (C) the DCA curve. DCA, decision curve analysis; GC, gastric cancer; M, metastasis; N, node; OS, overall survival; T, tumor.

Functional analysis of signature

In GSEA annotated by the KEGG database in the training cohort, we identified significant enrichment of pathways related to molecular mechanisms of various cardiomyopathies, calcium signaling, axon guidance, basal cell carcinoma, JAK-STAT signaling, melanoma and others in the high-risk groups (Figure 5A,5B).

Figure 5.

Functional analysis. (A,B) GSEA of KEGG pathways for the high-risk group; (C,D) GO enrichment analysis; (E,F) KEGG enrichment analysis. BP, biological process; CC, cellular component; GSEA, gene set enrichment analysis; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function.

Our enrichment analysis has revealed that DE-PMRGs are predominantly involved in processes associated with the development and function of the muscular system in biological processes (BPs). Within the cellular component (CC), these genes localized to specific structures within muscle cells. In molecular function (MF) analysis revealed roles in processes related to the molecular functions of the muscular system (Figure 5C,5D). Furthermore, these DEGs are linked to 71 KEGG pathways pertinent to muscle cells and cardiac diseases, encompassing cytoskeleton in muscle cells, vascular smooth muscle contraction, and dilated cardiomyopathy, among others (Figure 5E,5F).

Immunological characteristics in the risk model

Five immune cell types were observed between high and low-risk groups with disparities. Elevated expressions of macrophages M2, mast cells resting and NK cells activated were noted in the high-risk cohort, whereas activated CD4 memory T cells and T cells follicular helper were more abundant in the low-risk group (Figure 6A,6B). Additionally, 28 immune checkpoint molecules exhibited differential expression, with HHLA2, LGALS9, and TNFRSF14 being more abundant in the low-risk group (Figure 6C). Risk scores positively correlated with macrophages M2, mast cells activated and monocyte, notably with a significant positive correlation between Mast cells activated and BCHE (Figure 6D).

Figure 6.

Analysis of immune-related comparing high- and low-risk groups. (A) Differential analysis of immune cells in patients across different risk groups; (B) proportions of 22 immune cell types in GC samples across high- and low-risk groups; (C) analysis of the expression levels of 28 immune checkpoints between the high-risk and low-risk groups; (D) correlation of different immune cells with risk scores and 8 central hub PMRGs. *, P<0.05; **, P<0.01; ***, P<0.001. GC, gastric cancer; NK, natural killer; PMRGs, propanoate metabolism-related genes.

The response to immunotherapy

Recent studies have indicated that patients with higher TMB are more likely to benefit from immunotherapy (12). In this study, by analyzing tumor mutational characteristics and the relationship between TMB and survival, revealed that patients in the low-risk group exhibited relatively higher TMB (Figure 7A), which inversely correlated with their risk scores (Figure 7B). Specifically, patients with high TMB, particularly those in the low-risk subgroup, demonstrated better survival outcomes, whereas those with low TMB in the high-risk group had the poorest prognosis (Figure 7C). These findings suggest that patients with high TMB in the low-risk group may have the best response to immunotherapy. Further TIDE analysis indicated slightly higher TIDE scores in the high-risk group compared to the low-risk group (Figure 7D), implying a greater likelihood of immune escape among high-risk patients. Additionally, we assessed the potential efficacy of immunotherapy using the IPS, as well as the combination scores of IPS-CTLA-4, IPS-PD-1, and IPS-PD-1 combined with CTLA-4. The results indicated that the low-risk group had higher scores in all IPS metrics compared to the high-risk group (Figure 7E-7H), further confirming the greater likelihood of benefiting from immunotherapy in the low-risk cohort.

Figure 7.

Predicted response to immunotherapy. (A) Differences in TMB between the two risk groups; (B) Scatter plot of correlation between risk score and TMB; (C) K-M curves for patients with different TMB and risk scores; (D) TIDE scoring in the high-risk and low-risk groups. Comparison of IPS in the two groups with CTLA4 negative/PD-1 negative (E), CTLA4 negative/PD-1 positive (F), CTLA4 positive/PD-1 negative (G), CTLA4 positive/PD-1 positive (H). ***, P<0.001. H, high; IPS, immune prognostic scores; K-M, Kaplan-Meier; L, low; PD-1, programmed cell death protein 1; TIDE, Tumor Immune Dysfunction and Exclusion; TMB, tumor mutational burden.

The drug sensitivity in low‑ and high‑risk groups

Our analysis of drug sensitivity across risk groups revealed that 5-fluorouracil (5-FU), afatinib, cisplatin, and gefitinib exhibit high sensitivity in high-risk GC patients (Figure 8A-8D), while NU7441, dasatinib, BMS-754807, and AZD8055 showed better sensitivity in low-risk patients (Figure 8E-8H), supporting personalized therapy.

Figure 8.

Drug sensitivity across different risk groups. Sensitive drugs in the high-risk group: (A) 5-fluorouracil, (B) afatinib, (C) cisplatin, (D) gefitinib. Sensitive drugs in the low-risk group: (E) NU7441, (F) BMS-754807, (G) AZD8055, (H) dasatinib.

Discussion

GC is a highly heterogeneous and complex malignant tumor. The diversity of its metabolic characteristics, genetic variations, and immune microenvironment makes it difficult to achieve early diagnosis and treatment, thereby significantly affecting the prognosis of patients (13). Despite improved early detection through advanced endoscopy, many patients still present at late stages where surgery has limited benefit (14). Consequently, chemotherapy, molecularly targeted therapies, and immunotherapies have become key treatment modalities for GC. Although molecularly targeted therapies have provided new hope, their efficacy remains limited and variable (15). Therefore, the identification of specific and potent molecular targets is critical for improving therapeutic outcomes. With the advent and advancement of immunotherapeutic agents, immunotherapy is increasingly challenging conventional treatments, including chemotherapy and molecular targeting (16). By harnessing or augmenting the patient’s immune system to combat cancer cells, immunotherapy has demonstrated remarkable efficacy against certain malignancies (2), investigating the mechanisms and targets of immunotherapeutic genes is crucial for advancing GC treatment. This research could pave the way for new strategies and more personalized, effective treatments for patients.

Gut microbiota ferment dietary fibers to produce SCFAs, primarily including acetate, propionate, and butyrate. These SCFAs interact with G protein-coupled receptors (GPR41, GPR43, and GPR109A) expressed on intestinal epithelial and immune cells, playing a key role in inhibiting inflammation and carcinogenesis (17). While butyrate, serving as an energy source for colonocytes, has been extensively studied for its anti-inflammatory and anti-carcinogenic properties (18-20), the health-promoting characteristics of propionate have received less attention (7). Nonetheless, current research indicates that the role of propionate metabolism in cancer is dualistic. On one hand, propionate can enhance epithelial cell characteristics and inhibit the EMT, thereby reducing tumor invasiveness and metastatic potential (21). On the other hand, dysregulation of propionate metabolism can lead to the accumulation of MMA, a metabolite that exhibits oncogenic properties in tumor cells. MMA promotes tumor invasiveness and metastasis by activating the transforming growth factor (TGF)-β signaling pathway and inducing EMT. Additionally, MMA can drive tumor progression by increasing reactive oxygen species (ROS) levels and inducing inflammatory responses (22). However, research on propionate metabolism-related biomarkers in GC is still in its early stages. This study aims to address this gap by investigating PMRGs as potential prognostic biomarkers for GC. We have developed a risk model based on eight key PMRGs to explore the potential impact of propionate on GC prognosis and treatment.

This study aims to investigate the impact of PMRGs on the prognosis and treatment outcomes of GC patients. To achieve this, we utilized data from TCGA to construct a prognostic risk scoring system based on PMRGs. Building on this foundation, we developed and refined a prognostic model using LASSO Cox regression analysis. Ultimately, eight key PMRGs—APOA1, GAMT, GCG, CD36, HBB, SERPINE1, BCHE, and GNAS—were identified as core components of the prognostic model.

Apolipoproteins are the primary protein components of high-density lipoprotein (HDL), with apolipoprotein A1 (ApoA1) being one of the key apolipoproteins that play a crucial role in regulating cholesterol transport and metabolism (23). Reprogramming of cholesterol metabolism is closely associated with various diseases, including cancer, and has been shown to influence the proliferation, migration, and invasive capabilities of tumor cells (24). In addition, ApoA1-mediated lipid metabolism can reshape and reduce 7-ketocholesterol levels, thereby directly inhibiting tumor necrosis factor signaling in tumor-associated macrophages (TAMs) through mitochondrial translation inhibition (25). Guanidinoacetate N-methyltransferase (GAMT) is an enzyme that plays a crucial role in the creatine biosynthesis pathway (26). Previous study has shown that creatine can inhibit tumor growth (27). Glucagon (GCG) inhibits tumor angiogenesis, thereby enhancing the suppression of tumor growth by 5-FU. Additionally, GCG targets vasculogenic mimicry, further blocking the tumor’s access to its blood supply. This dual disruption of the tumor vascular system significantly curtails tumor growth and development (28). Cluster of differentiation 36 (CD36) is a transmembrane protein involved in fatty acid uptake and metabolism, playing a significant role in tumorigenesis and metastasis driven by high-fat diets (HFDs) (29). Further research indicates that CD36 promotes cancer development by maintaining lipid homeostasis (30). Hemoglobin beta-chain (HBB), primarily expressed in red blood cells for oxygen transport (31), has been found to have consistently elevated expression levels in circulating tumor cells (CTCs) from breast cancer, prostate cancer, and lung cancer (32). The SERPINE1 gene encodes plasminogen activator inhibitor-1 (PAI-1) , a protein that is upregulated in both inflammation and cancer (33). Studies have shown that SERPINE1 is a target gene of miR-192, which influences tumor behavior by regulating the expression of PAI-1. The EMT is a critical process in tumor metastasis. During EMT, the expression of miR-192 is downregulated, resulting in increased expression of its target gene, SERPINE1, thereby promoting the progression of EMT (34,35). Butyrylcholinesterase (BCHE) is a plasma enzyme that hydrolyzes ghrelin and bioactive esters (36), and its levels are found to be reduced in the serum of cancer patients (37). The Gαs protein, encoded by the GNAS gene, is one of the most frequently mutated G proteins in tumors (38). Research has shown that mutations in this gene are associated with the development of cancers, including appendiceal cancer, pituitary adenomas, endometrial adenocarcinoma, and gastric adenocarcinoma (39).

Univariate and multivariate Cox regression analyses demonstrated the independence of the risk score. Consequently, we developed a nomogram that integrates clinical features to offer a customized scoring system. This system has been validated for its superior performance in predicting the prognosis of GC patients, providing an innovative method for forecasting survival outcomes. In order to broaden the clinical application of the risk score, we constructed a nomogram chart integrated with the clinical characteristics. DCA and collaborative curves demonstrated the efficacy of nomogram chart in predicting GC prognosis, providing a simple and feasible tool for clinical doctors to evaluate the prognosis of GC patients. At the same time, based on the risk score, we conducted drug sensitivity analysis, and found that high and low-risk groups have different treatment responses to different drugs, paving the way for personalized immunotherapy for GC patients.

To investigate the potential molecular functions of PMRGS, we conducted GSEA analysis on different risk cohorts. Surprisingly, a large number of cardiomyopathy-related pathways were enriched in the high-risk group. A recent study revealed a mechanism by which patients with heart disease have an increased risk of cancer: small extracellular vesicles or vesicles (sEVs) secreted post-myocardial infarction carry various tumorigenic factors which accelerate the growth of cancer cells when (40).

The TME plays a crucial role in shaping malignant phenotypes, including tumor growth, invasion, metastasis, drug resistance, and immune evasion (41). Our research has discovered a significant elevation in the expression levels of NK cells activated, M2 macrophages, and quiescent mast cells in high-risk GC patients. Macrophages that infiltrate the TME are referred to as TAMs, which exhibit two distinct polarization states: M1 polarization, which has anti-tumor effects, and M2 polarization, which promotes tumor development (42). In GC, TME often promotes M2 polarization of TAMs while inhibiting M1 polarization. This shift is a key factor contributing to the development of immune tolerance (43). Further research indicates that SCFAs exacerbate this process. SCFAs not only inhibit M2 macrophage polarization but also trigger autophagy in cancer cells and promote additional M2 polarization of macrophages, thereby accelerating tumor progression (44). The gut microbiota and their metabolites, including propionate, play an important role in regulating immune function. They affect not only the fine-tuning of T cell effectors, regulation, and memory phenotypes but also directly influence the efficacy of chemotherapy, radiotherapy, and immunotherapy (45). Interleukin-17A (IL-17A) is a crucial pro-inflammatory cytokine that plays a vital role in the fibrotic process of various organs. In the context of GC, IL-17A produced by mast cells not only triggers changes in peritoneal cells but also directly leads to stromal fibrosis in GC patients (46). TME analysis is crucial for optimizing immunotherapy in GC patients. Additionally, we assessed the drug sensitivity of patients in both high and low-risk groups to enhance clinical efficacy. The results indicate that high-risk individuals are sensitive to a broader range of chemotherapeutic drugs compared to the low-risk group. In conclusion, the propionate prognostic model can comprehensively evaluate propionate metabolism patterns. In clinical practice, the risk score can be utilized to predict the efficacy and prognosis of immunotherapy.

Despite our research highlighting the potential of a propanoate metabolism-related prognostic model for predicting GC prognosis and identifying patients sensitive to chemotherapy and immunotherapy, there are limitations. First, our analysis relies heavily on publicly available databases and lacks independent clinical validation. Additionally, the current findings are primarily based on bioinformatics analysis, necessitating experimental methods to verify the specific roles of the eight genes involved in PMRGs in GC. To enhance the reliability of our study, future work should incorporate more experimental data. Concurrently, larger-scale prospective clinical trials are needed to confirm the effectiveness and applicability of our risk prognostic model in real clinical settings.

Conclusions

We developed a prognostic model based on PMRGs, which can independently predict clinical outcomes in GC patients. This model serves as a prognostic tool, offering new insights for advancing therapeutic strategies in GC, thereby influencing patient prognosis and treatment planning. We refined the model using LASSO and Cox proportional hazards regression analysis. Ultimately, eight key genes (APOA1, GAMT, GCG, CD36, HBB, SERPINE1, BCHE, GNAS ) were identified as core components of the prognostic model.

Supplementary

The article’s supplementary files as

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.