Untargeted metabolomics reveals gut microbiota metabolite alterations and their correlation with serum biomarkers in gastric cancer patients from high-altitude regions.

Introduction
Gastric cancer is a relatively common malignant tumor of the digestive tract. According to the 2022 GLOBOCAN statistics [1], gastric cancer ranks fifth in both incidence and mortality among all malignant tumors, with approximately 970,000 new cases and 660,000 deaths globally. In China, there are 359,000 new gastric cancer cases and nearly 260,000 deaths annually, accounting for 37.0% and 39.4% of the global totals, respectively. Qinghai Province, located on the Tibetan Plateau, the highest altitude region on Earth, reports gastric cancer as the leading malignancy in terms of incidence [2]. Even more concerning is that most gastric cancer patients are diagnosed at an advanced stage, presenting significant challenges for treatment and prognosis [3, 4]. Research on the diagnosis and treatment of high-altitude gastric cancer has thus become a key focus in the field of gastric cancer in China.
In recent years, significant progress has been made in the study of gastric cancer. Serum biomarkers play a crucial role in the diagnosis and treatment of gastric cancer. Traditional serum biomarkers for gastric cancer, such as CEA, CA19-9, CA24-2, and PGs [5], can be used for early screening and intervention in high-risk populations. Additionally, other biomarkers associated with gastric cancer have been discovered in recent years, including Establishment of a novel lysosomal signature for the diagnosis of gastric cancer with in-vitro and in-situ validation, HER2, MSI-H/dMMR, PD-L1, and CLDN18.2 [6, 7]. These biomarkers are of great value in the precise diagnosis, targeted therapy, and immunotherapy of gastric cancer, helping to identify patients suitable for targeted treatment. Some studies have shown that [8, 9], a metabolite produced by gut microbiota, also plays a significant role. Dysbiosis of the gut microbiota can lead to the release of toxic metabolites, which may exhibit tumor-promoting effects within the host. These metabolites could influence the levels of gastric cancer-related biomarkers in serum, but this hypothesis requires further research to confirm. In our country, many studies have shown differences in gut microbiota between healthy individuals and gastric cancer patients [10, 11]. However, research on the specific relationship between gut microbial metabolites and serum biomarkers of gastric cancer is still relatively limited. Therefore, this study uses non-targeted metabolomics to investigate the differences in metabolic products among gastric cancer patients in the Qinghai-Tibet Plateau region. By comparing gastric cancer patients with healthy controls, we analyze changes in metabolic products of gastric cancer patients and perform KEGG enrichment analysis to identify related pathways involved in these differences. This provides a reference for future diagnosis, prevention, and treatment of gastric cancer in plateau regions from the perspective of microbial metabolism.

Materials and methods

Study subjects
Patients newly diagnosed with gastric cancer and hospitalized between January 2024 and September 2024 at a tertiary general hospital in Qinghai Province were included in the study. The diagnosis strictly followed the 2022 version of the Gastric Cancer Diagnosis and Treatment Guidelines issued by the National Health Commission of China [12]. To achieve better statistical power in metabolomics analysis, at least 25 biological samples per group were required; thus, 30 gastric cancer patients and 30 healthy controls were enrolled, with informed consent obtained for fecal sample collection.

Inclusion criteria

Gastric cancer group
Residents aged 30–75 years, residing in high-altitude areas of Qinghai Province for over 10 years, who have been diagnosed with gastric adenocarcinoma for the first time through endoscopy and biopsy, without tumor rupture or bleeding, and excluding patients with pyloric obstruction, have not taken proton pump inhibitors, antibiotics, probiotics, antacids, or acid suppressants within one month before sampling, and have negative results from breath tests.

Healthy control group
Residents aged 30–75 years, residing in high-altitude areas of Qinghai Province for over 10 years, with normal gastric mucosa under endoscopy, no history of digestive system diseases or related symptoms; within 1 month before sampling, they have not taken proton pump inhibitors, antibiotics, probiotics, antacids, or acid suppressants, and the Helicobacter pylori breath test is negative.

Exclusion criteria
Subjects with other systemic diseases or mental disorders, those who did not provide informed consent or refused to cooperate, participants in other clinical drug trials, or those with improper sampling methods (e.g., collecting stool samples at night) were excluded. This study was approved by the ethics committee of the local hospital. All participants were fully informed of the sampling procedures and research plan, and signed an informed consent form.

General data collection and fecal sample collection and preprocessing

General data collection
Basic information of all participants, including age, sex, and pathological type, was collected.

Fecal sample collection and preprocessing
Fresh stool samples were provided by gastric cancer patients during routine stool examination upon admission and by healthy controls during health check-ups. Sterile swabs were used for sample collection, ensuring no contact with urine or wastewater. Samples were transferred to specialized sampling tubes and thoroughly mixed with gut sample preservation solution to obtain a fecal homogenate. Samples were then divided into three portions, sealed in sterile containers, and stored at − 80 °C for later analysis.

Experimental procedures

Metabolite extraction
Weighed 25 mg of the sample into an EP tube under low temperature, added homogenization beads, and 500 µL of extraction solution (methanol: acetonitrile: water = 2:2:1, V/V) containing isotopically labeled internal standards. Vortexed for 30 s, homogenized in a homogenizer (35 Hz, 4 min), and ultrasonicated in an ice-water bath for 5 min; repeated the process three times. Samples were left to stand at − 40 °C for 1 h, then centrifuged at 12,000 rpm (13,800×g, radius 8.6 cm) at 4 °C for 15 min. The supernatant was collected for analysis, with equal amounts pooled into QC samples.

Instrumental analysis
Polar metabolites were analyzed using a Vanquish UHPLC system (Thermo Fisher Scientific) with a Waters ACQUITY UPLC BEH Amide column (2.1 mm × 50 mm, 1.7 μm). The mobile phase consisted of 25 mmol/L ammonium acetate and 25 mmol/L ammonia in water (Phase A) and acetonitrile (Phase B). Sample plate temperature was set at 4 °C, injection volume was 2 µL. An Orbitrap Exploris 120 mass spectrometer was used for MS and MS/MS data acquisition under software control (Xcalibur v4.4, Thermo). Parameters: Sheath gas flow rate 50 Arb, Aux gas flow rate 15 Arb, Capillary temperature 320 °C, Full MS resolution 60,000, MS/MS resolution 15,000, Collision energy SNCE 20/30/40, Spray Voltage + 3.8 kV (positive) or − 3.4 kV (negative).

Data processing
Raw data were converted to mzXML format using ProteoWizard software and metabolites were identified using a customized R package with the BiotreeDB database (V3.0). Visualization was performed using another R package developed in-house.

Quality control
Real-time monitoring of instrument stability and signal consistency was crucial for ensuring data quality. Early detection and resolution of anomalies were prioritized.

Peak height variations in internal standards across QC samples
The retention time and response intensity of internal standards in QC samples (Figs. 1 and 2) demonstrated excellent stability, indicating reliable data acquisition by the instruments.

Presentation of QC samples in the 2D PCA score plot
Theoretically, QC samples should be identical. However, errors in extraction, detection, and analysis processes may result in differences between QC samples. Smaller differences indicate higher method stability and better data quality. As shown in Fig. 3, QC samples exhibit strong clustering, indicating excellent method stability.

Stability of internal standard response in QC samples
The internal standards are isotopically labeled metabolites introduced into the samples, with identical concentrations in QC samples. Smaller response differences in internal standards (median RSD ≤ 10%) reflect greater system stability and higher data quality. As shown in Table 1, the data quality of this experiment is excellent.

Statistical analysis
Statistical analysis and plotting were performed using SPSS 26.0. Basic characteristics between the two groups were compared using independent-samples t-tests. Multiple group comparisons were conducted using one-way ANOVA. For non-normal data or unequal variances (quartile method), the Kruskal–Wallis rank-sum test was used. The significance level was set at α = 0.05. Comparison of sample rates between groups was performed using Pearson’s χ² test. For non-normal data, Spearman correlation coefficients were used to analyze the correlation between gut microbiota abundance and differential metabolites. Correlation coefficients (r) and P values less than 0.05 were considered statistically significant. To account for potential non-linear relationships or plateau effects in metabolite responses, both parametric and non-parametric approaches were applied based on distribution characteristics and variance assumptions. The combined use of correlation, multivariate, and rank-based tests ensured robustness in identifying consistent trends.

Results

General characteristics of healthy and gastric cancer groups
This study included 30 patients in the gastric cancer group, with diagnoses confirmed by gastroscopy and histopathology. None had undergone surgical treatment, radiotherapy, or chemotherapy. Pathological diagnoses included 12 cases of antral adenocarcinoma, 6 cases of cardia adenocarcinoma, 6 cases of angular adenocarcinoma, 4 cases of in situ carcinoma of the antrum, and 2 cases of in situ carcinoma of the gastric angle.
The healthy group included 30 individuals with no digestive system symptoms, negative clinical tests, and normal gastric mucosa under gastroscopy.
The mean age of the gastric cancer group was 60.40 ± 9.902 years (range: 38–78), and that of the healthy group was 57.93 ± 11.447 years (range: 30–73), with no statistically significant difference (P = 0.386). The male-to-female ratio was 1.7:1 in the gastric cancer group and 1.1:1 in the healthy group, with no statistically significant difference (P = 0.432). See Table 2 for details.

Analysis of gut microbiota metabolites in gastric cancer and healthy groups
Metabolomics analysis identified 281 metabolites across samples from the gastric cancer and healthy groups. These metabolites were categorized into 8 main classes: Lipids and lipid-like molecules: 35.443%, Organic acids and derivatives: 29.114%, Organic oxygen compounds: 15.19%, Nucleosides, nucleotides, and analogues: 13.924%, Organoheterocyclic compounds: 2.532%, Amino acids and peptides: 1.266%, Benzenoids: 1.266%, Fatty acids: 1.266%. See Fig. 4 for the compositional pie chart of gut microbiota metabolites.

Quality control and data stability evaluation
To ensure data quality, TIC and EIC plots of internal standards across all QC samples were generated, demonstrating consistent retention times and peak intensities. RSD values for internal standards were all below 6.6%, and PCA showed tight clustering of QC samples (see Supplementary Biotree assay report). These results confirm the reliability and reproducibility of the LC-MS platform used.

Differential metabolite analysis between gastric cancer and healthy groups

Principal component analysis (PCA) of fecal metabolomics
Given the high-dimensional nature of metabolomics data, each substance in the sample represents a data dimension. Multivariate statistical analyses, starting with PCA, were conducted to explore the internal structure of the data. PCA is a statistical method that transforms potentially correlated variables into orthogonal, linearly uncorrelated variables (principal components). As an unsupervised model, PCA reveals internal data structures, reducing dimensionality to provide a 2D or 3D projection that highlights distribution trends and differences among samples. PCA does not explicitly enhance group distinctions due to its unsupervised nature, as it is affected by variables unrelated to grouping. For clearer group differentiation, the supervised classification model OPLS-DA (outlined in the next section) can be used. Using SIMCA software (V16.0.2, Sartorius Stedim Data Analytics AB, Umea, Sweden), data were log-transformed and centered (CTR) before automatic modeling. See Table 3 for PCA model parameters, and Figs. 5 and 6, and 7 for PCA scatter plots.

Fecal metabolomics analysis of gastric cancer and healthy groups using PLS-DA
Partial Least Squares Discriminant Analysis (PLS-DA) was used to compare the gastric cancer group and the healthy group. The analysis revealed a clear separation trend between the two groups. Data from the gastric cancer group were mainly clustered in the right quadrant, while data from the healthy group were predominantly in the left quadrant, indicating significant differences in metabolite levels between the groups. The model parameters, R²X = 0.238, R²Y = 0.953, and Q²=0.864, confirmed the robustness of the PLS-DA model. See Fig. 8.

Fecal metabolomics analysis using OPLS-DA
Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was used to further analyze the data. The OPLS-DA results also revealed a separation trend between the gastric cancer and healthy groups, confirming the metabolite differences. To ensure model reliability, permutation tests were performed. The regression lines of Q² and the intercepts with the y-axis were both below zero, indicating model reliability. See Figs. 9 and 10. Indicate the frequency of R²Y values. P < 0.05 indicates the optimal model.

Screening of differential metabolites
After PCA, PLS-DA, and OPLS-DA analyses, significant differences in fecal metabolites were identified between the gastric cancer and healthy groups. Fold-change analysis (FC analysis) and t-tests were performed for univariate statistical analysis, and a volcano plot was generated to illustrate the significant changes in metabolites. Some metabolites were significantly upregulated or downregulated in the gastric cancer group. See Fig. 11.

Each point represents a metabolite. The horizontal axis represents fold-change (log₂-transformed), and the vertical axis represents the t-test P-value (log₁₀-transformed negative values). Point size represents the VIP value from the OPLS-DA model. Red points indicate significantly upregulated metabolites, blue points indicate significantly downregulated metabolites, and gray points indicate non-significant metabolites.
The top 10 upregulated metabolites in the gastric cancer group included: Dethiobiotin, Glycylproline, Glycine, Hydroxyisocaproic acid, Tyramine, Methionine sulfoxide, 5-Aminopentanoic acid, Citrulline, Betonicine, Formiminoglutamic acid.
The top 10 downregulated metabolites included: Cytidine, 5’-Methylthioadenosine, Trehalose, Melibiose, Lotaustralin, Adenosine, Inosine, Ribothymidine, Raffinose, Galactinol, See Table 4; Fig. 12 for details.

Clinical phenotype analysis between gastric cancer and healthy groups
A comparison of clinical phenotypes between the gastric cancer group and the healthy group showed statistically significant differences in the following parameters: Total Protein (TP, g/L, t = − 4.403, P = 0.000), Albumin (ALB, g/L, t = − 2.629, P = 0.011), Globulin (GLO, g/L, t = − 3.613, P = 0.001), Total Cholesterol (TC, mmol/L, t = − 5.357, P = 0.000), High-Density Lipoprotein (HDL, mmol/L, t = − 3.252, P = 0.002), Low-Density Lipoprotein (LDL, mmol/L, t = − 5.143, P = 0.000), Alpha-Fetoprotein (AFP, ng/ml, z = 1.420, P = 0.035), Carbohydrate Antigen CA72-4 (U/ml, z = − 2.619, P = 0.009). For detailed comparisons, refer to Table 5.

Correlation analysis between clinical phenotypes and top 10 differential fecal metabolites in gastric cancer group
A correlation analysis was conducted between the top 10 upregulated and downregulated fecal metabolites and the clinical phenotypes with statistically significant differences between the gastric cancer group and the healthy group. Among the top 10 upregulated metabolites, Hydroxyisocaproic acid was positively correlated with albumin (r = 0.406, P = 0.026), Glycine was positively correlated with albumin (r = 0.407, P = 0.026), Tyramine was negatively correlated with alpha-fetoprotein (r = −0.380, P = 0.039), and Glycylproline was positively correlated with albumin (r = 0.452, P = 0.012). Among the top 10 downregulated metabolites, 5’-Methylthioadenosine was negatively correlated with high-density lipoprotein, low-density lipoprotein, and alpha-fetoprotein (r=−0.420, P = 0.021; r=−0.378, P = 0.040; r=−0.391, P = 0.033), Cytidine was negatively correlated with total protein and globulin (r = −0.476, P = 0.008; r = −0.486, P = 0.006), Adenosine was negatively correlated with globulin (r=−0.410, P = 0.024), and Galactinol was positively correlated with globulin (r = −0.370, P = 0.044). See Table 6.

Discussion
Given the moderate sample size and discovery-focused design, the present findings should be interpreted as hypothesis-generating rather than definitive. In recent years, with the increasing depth of research into metabolomics by scholars, it has been discovered that changes in gut microbiota and their metabolites are significantly influenced by geographical regions [13, 14]. Gastric cancer is one of the most prevalent malignant tumors in the Qinghai Plateau region [2]. Despite the regional disease burden, there is a critical gap in metabolomics-based profiling of gastric cancer patients living under high-altitude conditions, where hypoxia and environmental stress may influence tumor biology and gut microbial function.
Current research on gastric cancer in high-altitude regions primarily focuses on factors such as immunity and genetics, while studies on the non-targeted metabolomics of gut microbiota in gastric cancer patients from high-altitude regions remain relatively scarce. Therefore, this study analyzes the changes in gut microbiota metabolites between healthy individuals and gastric cancer patients from high-altitude areas, exploring the differential metabolites of gut microbiota and their correlation with clinical phenotypes. From the perspective of non-targeted metabolomics of gut microbiota, this study aims to uncover the factors contributing to gastric cancer in high-altitude regions, providing new insights for the prevention and precision treatment of gastric cancer in such areas.
Metabolomic analysis of samples from the gastric cancer group and the healthy group revealed that the identified metabolites belonged to eight categories: lipids and lipid-like molecules, organic acids and derivatives, organic oxygen compounds, nucleosides/nucleotides and analogs, heterocyclic compounds, amino acids and peptides, benzene derivatives, and fatty acids. These findings are consistent with the results of non-targeted metabolomics studies on gastric cancer patients by Jiang Xiaoman et al. [14]. Multivariate statistical models, including PCA, PLS-DA, and OPLS-DA, demonstrated clear separation between groups, validating the robustness of the data and highlighting disease-associated metabolic shifts. Comparing the expression levels of metabolites between the gastric cancer group and the healthy group, 27 metabolites showed significant differences. Among these, the top 10 upregulated metabolites in the gastric cancer group were Dethiobiotin, Glycylproline, Glycine, Hydroxyisocaproic Acid, Tyramine, Methionine Sulfoxide, 5-Aminopentanoic Acid, Citrulline, Betonicine, and Formiminoglutamic Acid. The top 10 downregulated metabolites were Cytidine, 5’-Methylthioadenosine, Trehalose, Melibiose, Lotaustralin, Adenosine, Inosine, Ribothymidine, Raffinose, and Galactinol.
These findings differ from studies on non-high-altitude gastric cancer by Schmidt [15] and Xu Chuxuan [16], which reported significant downregulation of citrulline and tyramine in serum metabolomics. Jing [17] reported that the levels of 13 metabolites, including isoleucine, lactic acid, glutamic acid, glutathione, trimethylamine oxide, 4-hydroxyphenylacetic acid, tyrosine, phenylacetylglutamine, hypoxanthine, citrulline, valine, acetoacetic acid, and methylamine, vary with the progression of gastric cancer. The observed differences in this study may partly reflect environmental adaptations in high-altitude populations and support the need to explore altitude-specific cancer mechanisms.
To mitigate false positives in our high-dimensional dataset, all reported correlations underwent false discovery rate (FDR) correction. Nonetheless, given the exploratory scope of this study, Pearson correlation analysis was employed, and future work will incorporate multivariate regression models and targeted metabolomics for validation. Hydroxyisocaproic acid, a leucine catabolism product, was significantly positively correlated with albumin, potentially reflecting adaptations in amino acid metabolism under chronic hypoxia typical of high-altitude environments. Similarly, glycine and glycylproline showed positive associations with albumin, which may indicate altered protein turnover or muscle protein mobilization in gastric cancer patients residing at high altitude. These findings align with prior observations that leucine-derived metabolites can stabilize albumin structure and support nitrogen balance [18].
Among downregulated compounds, 5′-methylthioadenosine (5′-MTA) was negatively correlated with HDL, LDL, and AFP, which may reflect disrupted methionine salvage pathways and cancer-associated inflammatory modulation, supporting the emerging evidence that the tumor microbiome plays a key role in regulating immune responses and influencing the effectiveness of immunotherapy [19, 20]. Cytidine’s inverse association with total protein and globulin suggests perturbations in nucleoside metabolism and liver synthetic function. The study by Kejin Li and Paerhati Shayimu [21, 22] found that multivariate analysis identified the ratio of albumin to globulin, PA, ALB, and TRF as independent risk factors. Serum nutritional biomarkers combined with surgical factors reliably predict the risk of anastomotic leakage after rectal cancer surgery, highlighting its importance in preoperative assessment although direct mechanistic links require further investigation. Adenosine’s negative correlation with globulin further implicates nucleoside signaling in immune regulation. In combination with machine learning-based gastric cancer immunogenomic studies and post-gastrectomy osteoporosis risk prediction models, these findings underscore the potential of metabolomics data in precision oncology and long-term risk prediction [23, 24]. Galactinol displayed a positive correlation with globulin, possibly relating to altered glycosylation patterns of serum proteins and galectin-mediated immune interactions [25, 26]. Future studies will leverage high-resolution LC-MS/MS and proteomic integration to elucidate the underlying molecular mechanisms.
Although several top metabolites were consistently detected and annotated with high confidence, the absence of orthogonal validation or external replication limits the immediate clinical translatability of these findings. The study has several limitations that should be considered when interpreting these findings. The sample size, while sufficient for exploratory analysis, remains modest, limiting the generalizability of our findings. The heterogeneity of gastric cancer, including variations in molecular subtypes and disease staging, was not fully accounted for. Additionally, while we focused on a high-altitude cohort, environmental factors such as hypoxia and high-altitude acclimatization may have influenced the metabolic profiles observed, necessitating more targeted studies to explore these effects. Furthermore, our current approach did not include pathway enrichment analysis or mechanistic exploration of altitude-related metabolic perturbations. Future studies will incorporate these aspects by increasing sample size, stratifying by molecular subtype, and performing pathway enrichment analyses using KEGG and MetaboAnalyst.
To minimize potential bias from plateau effects or nonlinear response behaviors, we used both parametric and non-parametric statistical methods depending on data distribution. Spearman correlation was applied for non-normally distributed clinical–metabolite relationships, allowing detection of non-linear monotonic trends. Multiple validation layers, including consistent QC clustering, multivariate separation, and clinical correlation, support the robustness of our findings. In conclusion, this study sheds light on the unique metabolic alterations in gastric cancer patients from high-altitude regions, offering insights into the potential role of gut microbiota metabolites in disease pathogenesis. By linking metabolic changes with clinical biomarkers, we highlight the potential of metabolomics to identify novel diagnostic and therapeutic targets. Future work will further validate these findings through multi-omics approaches, pathway analysis, and functional assays to deepen our understanding of the mechanisms driving gastric cancer in high-altitude environments.
Although this study is based on an untargeted metabolomics approach, we ensured rigorous metabolite identification through MS1 and MS2 spectral matching against the BiotreeDB (v3.0) database. Several of the most significantly altered metabolites, including cytidine, glycine, trehalose, hypoxanthine, and 5′-methylthioadenosine, were identified with high confidence under the Metabolomics Standards Initiative (MSI) criteria (Levels 1 or 2). Their consistent detection across quality control samples, along with reproducible retention times and MS/MS fragmentation patterns, lends additional credibility to these findings. While targeted validation using authentic standards was not feasible within the scope of this study, it represents a key focus for future research to strengthen the clinical applicability of these biomarkers.

Conclusion
This study reveals significant alterations in faecal metabolites derived from gut microbiota in gastric cancer patients from high-altitude regions compared to healthy controls. Using untargeted metabolomics, we identified key metabolite shifts—particularly in lipid- and organic acid-related pathways—and observed meaningful correlations with serum albumin, lipid profiles, and tumor markers such as alpha-fetoprotein. Importantly, this study integrates high-altitude environmental context with gut microbial metabolism and systemic biomarkers, offering novel insights into gastric cancer pathophysiology in hypoxic settings. The results suggest a potential role for metabolites such as glycylproline, hydroxyisocaproic acid, and 5′-methylthioadenosine as indicators of disease status or progression. However, given the exploratory nature of the study, limitations related to sample size, high-altitude confounders, and molecular subtyping should be addressed in future studies. Planned efforts include pathway enrichment analysis, expanded multi-omics integration, and targeted validation of candidate biomarkers using independent cohorts. These steps will be critical for translating preliminary metabolomic signals into clinically actionable insights for gastric cancer detection and management in high-altitude populations.

Limitations
This study has several limitations. First, the relatively small sample size (30 patients per group) may limit statistical power and reduce the ability to account for the inherent heterogeneity of gastric cancer, including factors such as Lauren classification, disease stage, and molecular subtypes (e.g., EBV status, MSI). No power analysis was performed prior to study initiation, and while statistical adjustments for multiple comparisons were conducted (e.g., FDR), the risk of false positives remains. Second, although participants were matched based on age, sex, and Helicobacter pylori status, other high-altitude-related physiological confounders—such as hemoglobin levels and oxygen saturation—were not controlled for, potentially affecting metabolic profiles.
Third, gastric mucosal flora in healthy controls was not assessed, and colonization by other potentially pathogenic bacteria (e.g., Streptococcus spp.) was not excluded. Fourth, the LC-MS approach employed only HILIC columns, limiting the detection of lipophilic metabolites such as bile acids and sphingolipids. Additionally, the resolution of the Orbitrap Exploris 120 (60,000) may be insufficient to distinguish key isomers (e.g., d-/l-lactic acid), and the current annotation relied on BiotreeDB v3.0, which may not fully cover colony-specific metabolites like polyamines and tryptophan derivatives. While MS/MS-based spectral matching and retention time validation were performed and several metabolites were annotated with MSI Level 1 or 2 confidence, no orthogonal (targeted) validation using authentic standards was conducted in this study. Fifth, the study did not incorporate pathway enrichment analysis or a mechanistic exploration of how high-altitude hypoxia might influence gut microbiota metabolism. Finally, the presentation of some figures lacked consistent referencing and explanatory context, and sections of the discussion were overly detailed, which may hinder clarity for readers.

Future studies
To overcome the above limitations, future studies will expand the sample size and diversity through multi-center recruitment and include detailed stratification by Lauren classification, tumor stage, and molecular subtypes (e.g., EBV, MSI, genome stability). Power analyses will be conducted to ensure adequate statistical validity, and matched physiological parameters related to high-altitude acclimatization—such as hemoglobin levels, oxygen saturation, and red blood cell indices—will be incorporated. Control group design will be strengthened by including gastric mucosal microbiota profiling and broader pathogen screening beyond H. pylori. Methodologically, future metabolomics efforts will integrate RP-LC/MS to enhance detection of lipophilic metabolites and employ higher-resolution instruments or orthogonal techniques (e.g., ion mobility) to improve isomer discrimination. Expanded databases and reference libraries will be used to better annotate microbial-specific metabolites.
To deepen mechanistic understanding, planned studies will perform pathway enrichment analyses (e.g., via KEGG and MetaboAnalyst) and investigate metabolic reprogramming under hypoxic or high-altitude conditions. Targeted LC-MS/MS validation of top-ranked metabolites identified in this study (e.g., cytidine, glycine, trehalose) using authentic standards will be conducted to confirm their diagnostic and biological relevance. Targeted metabolomics, proteomics, and integrative multi-omics (e.g., metagenomics, transcriptomics) will be applied to explore host–microbiota interactions and validate biomarker candidates functionally using cellular or animal models. Finally, figure formatting, statistical annotations, and discussion clarity will be improved in subsequent publications to enhance reader comprehension. This comprehensive approach will support the development of predictive biomarkers and therapeutic targets tailored to gastric cancer in high-altitude populations.

Supplementary Information