Gastric cancer (GC) is a significant worldwide health challenge, being the fifth most prevalent form of cancer and the third leading cause of cancer-related fatalities. The disease presents an urgent threat to both public health and global economies due to its high occurrence and death rates (De Martel et al., 2020; Rawla and Barsouk, 2019). The GC is commonly correlated with environmental variables such as infection by Helicobacter pylori, dietary choices, and lifestyle, but the diffuse type tends to have a more aggressive clinical progression and is typically associated with genetic factors (Smyth et al., 2020). The clinical manifestation of GC exhibits considerable heterogeneity, with symptoms that span from vague dyspepsia to more severe indications, hence adding complexity to the management and prognosis (Yang et al., 2021). Advancements in the understanding of the development of GC have resulted in better methods for diagnosis and treatment. However, there are still considerable obstacles to overcome.

ADME, which stands for absorption, distribution, metabolism, and excretion, encompasses the fundamental processes in pharmacokinetics (Furuta et al., 2005). It outlines the path a medication takes from its introduction into the body to its removal. The latest progress in the research of ADME related genes (ADMERGs) has emphasized their vital functions in the regulation of drug-metabolizing enzymes, transport proteins, and nuclear receptors, hence exerting a major impact on these pharmacokinetic processes (Aitken et al., 2008; Fletcher et al., 2010). ADMERGs have been found to have important roles in the genesis and progression of GC. Genetic variations in specific metabolic enzyme genes and transport protein genes can have an influence on the way chemotherapeutic medications are metabolized and distributed in the body. This, in turn, can have an impact on the effectiveness of treatment and the development of drug resistance. Furthermore, the presence of ADMERGs can impact the availability and toxicity of drugs, ultimately affecting how patients respond to treatment and their overall prognosis (Rodrigues, 2022). While research has revealed certain functions of ADMERGs in GC, the specific methods by which they operate are still not well understood. Hence, it is imperative to do additional study on the correlation between ADMERGs and the prognosis of GC to develop more effective tailored treatments.

In this paper, differentially expressed genes (DEGs) associated with ADME were found using the TCGA-GC dataset from the TCGA database, and candidate genes were obtained by taking intersections with ADEMRGs. Prognostic genes closely associated with GC outcome were identified by univariate Cox regression analysis and Least Absolute Shrinkage and Selection Operator (LASSO) regression. Subsequently, a predictive risk model was created using prognostic genes to classify patients based on risk scores. In addition, a nomogram was created to accurately predict the prognosis of GC patients. To investigate the biological mechanisms of these prognostic genes, enrichment analysis, immune infiltration, drug sensitivity analysis, ceRNA network construction, and analysis of prognostic gene expression patterns at the single-cell level were also performed. The findings will provide novel insights into the molecular mechanisms behind GC and propose prospective targets for therapeutic intervention, as well as serve as a helpful reference for future research on the etiology and treatment strategies of GC.

GC is a leading cause of cancer deaths worldwide, and ADMERGs are increasingly implicated in its progression and therapy. Studying ADMERGs effect on GC can help predict patient prognosis and therapy options. This study found differentially expressed ADMERGs in combination with the TCGA-GC dataset. Five prognostic genes (UGT1A1, ADH4, ADH1B, CYP19A1, and GPX3) were then identified using univariate Cox regression and LASSO regression. Subsequently, Risk scores for GC patients were then obtained based on the prognostic genes and combined with a clinically created nomogram. In recent years, numerous studies have attempted to construct prognostic models for gastric cancer (GC) based on multi-omics features, covering multiple research dimensions such as immune-related genes, m6A modifications, metabolic regulation, and ferroptosis mechanisms. For example, Ma et al. developed a model based on differentially expressed immune-related genes (DEIRGs) to predict overall survival in GC patients and assess immune infiltration levels, though it did not account for the influence of metabolic or drug response pathways (Ma et al., 2024). Similarly, Peng et al. established a prognostic model for the GC tumor microenvironment based on m6A regulator-related genes (Peng et al., 2024). Feng et al. analyzed six tumor-associated metabolic pathways. Wen et al. developed a gastric cancer (GC) model based on ferroptosis-related genes (comprising six genes), which demonstrated prognostic value for patient survival but lacked in-depth investigation into immune phenotypes. Additionally, some studies have constructed models based on different cancer stages. For instance, Liu et al. developed prognostic models for early- and late-stage cancers. Compared to these models, the prognostic model constructed in this study based on five ADME-related genes not only more accurately predicts patient survival by incorporating covariates such as age and clinical stage (T, N, M), thereby providing a basis for developing personalized treatment strategies for gastric cancer (GC) at different stages, but also explores differences in genomic variation, immune microenvironment, and drug sensitivity, investigates the molecular regulatory mechanisms of prognostic genes, and observes the distribution and expression patterns of these genes in single-cell datasets, offering new reference evidence and theoretical support for GC treatment.

UDP glucuronosyltransferase family 1 member A1(UGT1A1) helps detoxify and eliminate toxins from the body and outside sources by glucuronidation. Chemical degradation associated to UGT1A1 expression alterations causes digestive tract cancer. Several studies suggested that UGT1A1 glucuronidases cancer-causing compounds to make them water-soluble and excretable. Chemicals that cause stomach lining cancer are broken down by UGT1A1. And UGT1A1 expression or activity may alter stomach lining cancer-causing material elimination and GC risk. Chemical accumulation from UGT1A1 dysfunction promotes DNA damage and tumor risk, while genetic variations in UGT1A1 can affect GC risk via modifying the enzyme’s chemical elimination (Pereira et al., 2022). Additionally, the detoxification function of UGT1A1 has been confirmed to be closely associated with the risk of digestive system tumors. In this study, it showed significant correlations with PD-L1 and HER2, suggesting its potential important role in immune regulation (Tomono et al., 2023; Sathe et al., 2024). Alcohol dehydrogenase 4 (Class II, ADH4) and alcohol dehydrogenase 1B (Class I, ADH1B), as key enzymes in ethanol metabolism, play a central role in liver ethanol detoxification by catalyzing the oxidation of alcohol to acetaldehyde. Notably, the ADH4-mediated conversion of ethanol to acetaldehyde is an essential step in liver detoxification (PMID: 16801720). However, excessive alcohol consumption significantly increases the risk of gastric cancer (GC), which is closely related to ADH4 (Freedman et al., 2007). The acetaldehyde produced by ADH4 catalysis exhibits high reactivity. On one hand, it can bind to DNA to form adducts, directly inducing gene mutations and compromising the stability of genetic material (Wang et al., 2023). On the other hand, acetaldehyde generated by ADH4 metabolism in gastric tissue can also disrupt the gastric mucosal barrier, creating pathological conditions for carcinogenesis (Nieminen and Salaspuro, 2018). Additionally, polymorphisms in the ADH1B gene significantly influence alcohol metabolism efficiency. Individuals carrying specific variants exhibit reduced acetaldehyde clearance capacity, leading to increased acetaldehyde accumulation during heavy drinking and thereby further elevating the risk of gastric cancer (GC) (Aida et al., 2013; Cui et al., 2009). However, in this study, ADH4 did not show significant downregulation in tumor tissues, suggesting that its expression may be subject to more complex regulatory mechanisms - a phenomenon that has not been adequately discussed in existing research. The aromatase gene cytochrome P450 family 19 subfamily A member 1(CYP19A1) can convert androgens into estrogens, and reproductive tissue growth and function depend on estrogens (Mo et al., 2013). The hormone-sensitive pathways of estrogen signaling are connected to stomach cancer. The estrogen produced by CYP19A1 activates estrogen receptors in gastric tissue, altering cell growth and death. Our study further reveals its low expression in GC tissues, complementing previous literature reporting that its variants may disrupt hormone signaling pathways and thereby promote tumor growth (Frycz et al., 2017). At the level of genetic variation, genetic differences in the CYP19A1 gene can affect aromatase activity, which can affect estrogen levels and gastric epithelial cell hormonal control (Miyoshi et al., 2003). Glutathione peroxidase 3(GPX3), an antioxidant enzyme, reduces hydrogen peroxide and organic hydroperoxides to protect cells. GC tissues show GPX3 downregulation, which increases oxidative stress and tumor development (Chen et al., 2011). Hypermethylation of the GPX3 promoter in GC decreases the output of this protective enzyme, leading to loss of GPX3 activity and causing oxidative damage and malignancy in gastric epithelial cells. Also, an increase in reactive oxygen species (ROS) can damage DNA and lead to cancer, so GPX3 is needed to counteract ROS (Lan et al., 2017). Due to their roles in carcinogen detoxification (UGT1A1), alcohol metabolism and acetaldehyde production (ADH4, ADH1B), hormonal regulation (CYP19A1), and oxidative stress protection (GPX3), these five genes affect GC development. These gene abnormalities can increase GC risk by causing DNA damage, inflammation, metabolic abnormalities and aberrant cell proliferation. These key genes (UGT1A1, ADH4, ADH1B, CYP19A1, and GPX3) are crucial for the construction of our gastric cancer (GC) prognostic model. However, based on the risk scores derived from the prognostic genes and combined with clinical data, the nomogram analysis revealed that the current model’s AUC values ranged between 0.59 and 0.77, failing to exceed 0.8. Nevertheless, existing literature reports that the AUC values of GC-related models vary from 0.560 to 0.989 (Xu et al., 2024), indicating that our newly developed model possesses a certain ability to distinguish between favorable and poor prognoses in patients and holds some clinical guidance significance. However, its accuracy may still be relatively low. Further optimization by incorporating additional clinical information is necessary to enhance its predictive accuracy and generalizability, thereby facilitating more personalized treatment strategies.

Similar to the findings of previous studies, the findings of this research demonstrate the complex correlation between the prognostic gene profile linked to ADME activities and the immune microenvironment in GC (Avolio et al., 2025; Li et al., 2024). The results emphasize notable disparities in the prevalence of immunological and stromal cells among the two indicated risk groups, with specific focus on immune cells such as Effector memory CD8 T cells, Activated B cells, and Central memory T cells. The presence of these differences highlights the potential impact of the immune landscape on the prognosis of patients with GC and the processes that drive tumor growth. The presence of a positive association between the risk score and the quantity of cell types such mast cells, plasmacytoid dendritic cells, and effector memory CD4 T cells indicates that the high-risk group may have a milieu that suppresses the immune system to a greater extent (Lazăr et al., 2018). This could exacerbate the unfavorable outcome reported in these individuals, as such an environment may enhance tumor evasion from immune monitoring and promote cancer development and spread (Yang et al., 2021). The significant variations in the levels of expression of 24 immunological checkpoints, such as the increased expression of BTLA, CD200, and CD40LG in the high-risk group, provide additional evidence that immune evasion mechanisms are more active in patients with higher risk scores. Therefore, utilizing personalized patient risk scores to tailor more effective immunotherapies may represent a promising direction, and future studies could incorporate clinical data from immunotherapy treatments to validate the predictive capability of TIDE scores in real-world therapeutic applications.

Furthermore, the contrasting expression of m6A-RGs in the two risk groups indicates that RNA methylation may have a crucial impact on regulating the immune response and affecting the clinical outcomes of patients with GC. The excessive production of m6A-RGs in the high-risk group may likely cause the disruption of immune-related pathways, leading to the observed changes in immune cell infiltration and checkpoint expression (Zhang et al., 2016). The medication sensitivity assessments offer useful insights into prospective therapy methods for various risk groups. Our research suggested that patients classified as low-risk are more prone to positive responses to medications such as QS11 and ABT.888. These treatments specifically target hypoxia pathways and inflammatory signaling, respectively. On the other hand, the high-risk group was more sensitive to drugs such as DMOG and VX.702. The Spearman association study provides more evidence, indicating a positive correlation between the risk score and sensitivity to BIBW2992, which is an EGFR inhibitor. Conversely, there is a negative correlation between the risk score and the drugs DMOG, Imatinib, AP.24534, and VX.702. The findings indicate that the prognostic gene signature not only has the ability to predict patient outcomes, but also has the potential to inform treatment decisions in clinical practice. Patients categorized as high-risk may get advantages from treatments that specifically target immunological checkpoints or EGFR signaling, and individuals in the high-risk group may also respond better to medicines that target hypoxia and inflammatory pathways. ESCC is characterized by mutations in the p53 gene, which enhance the production of the AGAP1 protein. This protein, in turn, increases the synthesis of exosomes, leading to the accelerated growth and spread of cancer cells (Feng et al., 2023). According to a study, QS11 has the ability to hinder the activity of AGAP1, which in turn reduces the growth and spread of ESCC cells. Furthermore, research has identified somatic (P53) mutations in GC as well (Han et al., 2024). The study found that low-risk patients showed considerable enrichment in pathways such as DNA replication and nucleotide excision repair. It is hypothesized that patients with a low risk profile may have a higher likelihood of seeing positive outcomes from gene- and protein-level therapy.

The integration of differential expression analysis, mRNA-miRNA-lncRNA interaction prediction, and scRNA-seq provides a comprehensive view of the molecular and cellular heterogeneity in GC (Li et al., 2025; Wan et al., 2015). Significant differences in miRNAs (e.g., mir590, mir7152, mir4420, mir5000) were identified between the high-risk and low-risk gastric cancer (GC) groups. These miRNAs exhibited significantly upregulated expression in the high-risk group, suggesting their potential role in suppressing oncogenic pathways to inhibit GC tumor growth, which aligns with existing literature. Similarly, the lncRNA (MIR100HG) was also markedly upregulated in the high-risk group. Previous studies have demonstrated that MIR100HG serves as a reliable prognostic biomarker associated with GC cell proliferation, migration, and invasion. The differentially expressed miRNAs and lncRNAs uncovered in this study may play pivotal roles in GC progression. They hold promise as prognostic markers and could provide novel therapeutic avenues for GC treatment.

ScRNA-seq data from GC tissues showed significant cell diversity using Seurat R package. ScRNA-seq data also showed significant differences in the expression levels of key prognostic genes such ADH1B, ADH4, CYP19A1, and GPX3 across cell types. The lower expression of these genes in tumor tissues suggests a role in GC onset and progression. For instance, decreased ADH1B and ADH4 expression in tumor cells can lead to acetaldehyde accumulation, which can promote cancer, DNA damage, and tumor growth. Decreased GPX3 may increase oxidative stress, which promotes gastric epithelial cell malignancy (Ren et al., 2021).

GC development’s molecular mechanisms can be understood by studying miRNA-lncRNA interactions and single-cell data. The identified miRNAs, lncRNAs, and target genes may be GC biomarkers. Additionally, they could be targeted for new therapeutic methods. Customized treatment techniques that account for tumors' diverse cell and molecular profiles are crucial because GC tissues contain distinct types of cells. ADME-related prognostic genes vary in expression in different cell types, suggesting that they can affect the metabolic milieu of the tumor microenvironment (TME) and targeted therapeutic efficacy (Gu et al., 2015).

Regarding the limitations of current GC diagnostic methods (such as gastroscopy and tumor marker detection) in terms of sensitivity and specificity, this study identified five ADME-related prognostic genes (UGT1A1, ADH4, ADH1B, CYP19A1, and GPX3) that exhibit unique expression patterns in tumor tissues (UGT1A1, ADH1B, CYP19A1, and GPX3 show significant downregulation). These findings hold potential for developing new tools and strategies to improve GC diagnosis and treatment. First, a multiplex gene quantification assay based on minimally invasive samples (peripheral blood, gastric juice, exfoliated cells) could be developed for early screening of high-risk populations (e.g., individuals with family history or chronic H. pylori infection). This tool could compensate for the limitations of gastroscopy, reduce missed diagnoses, and improve early detection rates. Second, detecting ADME gene expression (e.g., downregulation of UGT1A1, ADH1B, GPX3) in gastroscopically suspicious lesions could enhance diagnostic accuracy through combined assessment, guide further examinations, and reduce false-negative results. Additionally, an ADME risk score could effectively stratify GC patients into high-risk and low-risk groups. Significant differences were observed between the two groups regarding immune microenvironment, drug sensitivity, and prognosis. The high-risk group (with poorer prognosis) requires aggressive treatment strategies, including neoadjuvant chemotherapy combined with targeted therapy (e.g., EGFR inhibitors), radical surgery, and adjuvant therapy, to reduce recurrence risk. In contrast, the low-risk group (with better prognosis) should avoid overtreatment, with surgical intervention or low-toxicity adjuvant chemotherapy being preferred options, along with intensified postoperative surveillance. These findings provide valuable guidance for developing precision medicine strategies.

Due to the relatively limited sample size in the training, validation and scRNA-seq datasets, particularly the scarcity of clinically normal tissue samples—our study may be susceptible to batch effects or population bias. This constraint likely hindered the model’s ability to fully capture the comprehensive features and distribution patterns of the data, partially compromising the stability and reliability of the results (AUC >0.6 but <0.8) and limiting the generalizability of findings to broader populations. Similarly, the inconsistent results between RT-qPCR and single-cell data are likely attributable to insufficient sample size and sample heterogeneity. Furthermore, insufficient sample resources led to an extremely small clinical validation cohort (n = 5 for tumor/adjacent tissues), and the regulatory mechanisms of the ceRNA network relied solely on bioinformatic predictions without experimental validation, weakening the robustness of conclusions. Moving forward, we will further investigate the differential activity of these genes in specific cell populations. This will involve experimental validation of the ceRNA network’s regulatory mechanisms through approaches such as RNA interference (RNAi), overexpression assays, constructing animal models, and gene knockout experiments. These studies aim to elucidate the precise role of the ceRNA network in gastric cancer (GC) initiation and progression, as well as its potential crosstalk with other regulatory pathways. Finally, we will attempt to obtain additional single-cell sample data for analysis and collect extra clinical samples to validate our research findings via qPCR. Meanwhile, we will specifically measure the relationship between TIDE scores in the high-risk group and actual immune therapy responses to further verify the predictive value of TIDE scores. These efforts are expected to provide more reliable and actionable insights for improving GC treatment strategies.

This work concludes by offering a thorough investigation of the correlation between ADME-related prognostic genes and the immunological microenvironment in GC. The recognition of unique immunological profiles and medication sensitivities linked to various risk groups emphasizes the possibility of tailored treatment approaches in GC. Additional investigation is necessary to examine the fundamental mechanisms that cause these connections and to confirm these discoveries in medical environments, with the ultimate objective of enhancing patient outcomes through customized therapy methods. Our research helps us grasp GC’s complex molecular and cellular networks. The found miRNAs, lncRNAs, and prognostic genes offer promising potential for further research and therapeutic applications, such as targeted medicines and personalized GC treatment.

This study effectively discovered predictive genes associated to ADME in GC by thorough bioinformatics analysis of data from the TCGA and GEO databases. The results provide useful knowledge on the genetic basis of GC and emphasize the potential of these genes as biomarkers for the detection, prediction of outcome, and targets for treatment. Through the integration of differential expression analysis, miRNA-lncRNA interaction prediction, and single-cell RNA sequencing data, we have discovered noteworthy connections between these prognostic genes, the tumor microenvironment, and cellular heterogeneity in GC.

The identified genes and their regulatory networks serve as a new reference framework for future study in GC, namely in comprehending the disease’s course and in formulating tailored treatment methods. In order to further understand how these prognostic genes, contribute to the development of GC, it is necessary to conduct additional experimental research to uncover the exact mechanisms involved. Moreover, the practical implementation of these discoveries, which involves creating specific treatments, shows potential for enhancing the identification and management of patients with GC.