This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
This study aimed to examine the relationship between genetically predicted metabolite levels and gastric cancer (GC) risk using Mendelian randomization (MR), and to identify the metabolic pathways potentially involved.
Methods
We selected genetic instruments for metabolites from 64 genome-wide association studies covering 362,750 participants. A two-sample MR design was applied to evaluate the associations with GC using summary-level data from a combined analysis of the UK Biobank and FinnGen. The primary analysis relied on the inverse-variance weighted method, while the median-weighted and MR-Egger methods were used to account for potential violations of instrumental variable assumptions and provide the estimate even when a subset of instruments was invalid. The MR-Egger intercept test was performed to detect directional pleiotropy. Metabolites showing significant associations with GC were further examined using pathway enrichment analysis to identify relevant metabolic and lipid processes.
Results
MR analyses identified 25 and 17 metabolites that were positively and inversely associated with GC risk, respectively. Notably, hexanoylcarnitine and cis-4-decenoylcarnitine were strongly associated with increased risk, whereas pregnanediol disulfate, acetylcarnitine, prolyl-hydroxyproline, and X-18914 were associated with reduced risk, with no evidence of heterogeneity or directional pleiotropy. Enrichment analyses highlighted key metabolic pathways, including cysteine and methionine catabolism, beta-oxidation of pristanoyl-CoA (coenzyme A), oxidation of branched-chain fatty acids, and peroxisomal lipid metabolism.
Conclusion
This study identified a set of genetically predicted metabolites associated with GC risk, highlighting the potential utility of metabolite panels and lipid-based biomarkers for risk stratification and early detection. However, further standardization and extensive validation are necessary prior to clinical application.
According to GLOBOCAN 2022, gastric cancer (GC) is the fifth most commonly diagnosed cancer worldwide, with 968,350 new cases reported by 2022 [1]. Globally, there is a high incidence of GC in East Asian countries, especially where national cancer screening programs have been implemented [2]. In high-risk areas such as Japan and Korea, endoscopy has been included in the GC screening recommendations and has been shown to reduce the risk of GC by 61% and 47%, respectively [3]. However, endoscopy remains expensive and infeasible in many areas because of its invasiveness and complexity. Additionally, serious complications such as bleeding, infection, and patient discomfort are important factors to consider when implementing endoscopy for the early detection of GC [4]. Thus, non-invasive modalities such as Helicobacter pylori eradication, serum pepsinogen, and stool antigen tests have been introduced as potential screening strategies in low- and intermediate-risk areas [3]. However, these modalities have relatively low sensitivity and specificity for GC detection [3]. Therefore, identifying reliable biomarkers for more accurate screening and diagnostic tests is of great interest for the early detection of GC.
Metabolites, byproducts, or intermediates of metabolic processes are closely linked to the phenotype of a biological system and play key roles in regulating their functional expression [5]. Evidence from the European Prospective Investigation into Cancer and Nutrition study, which included data from 10 cohorts (238 cases and 626 matched controls), found positive associations between certain fatty acids, including oleic acid, di-homo-γ-linolenic acid, and α-linolenic acid with GC. In a prospective analysis of 400 participants, cross-validated findings reported inverse associations between α-linolenic acid, linoleic acid, palmitic acid, arachidonic acid, sn-1 lysophosphatidylcholine (LysoPC)(18:3), and sn-2 LysoPC(20:3) with GC [6]. Incorporating these metabolites into predictive models enhances the accuracy of gastric lesion progression and early GC risk assessment beyond models that only include age, sex, H. pylori infection, and histopathological findings [6]. A recent study of 1,800 participants also identified a metabolic signature of 26 metabolites related to H. pylori, which was linked to an increased risk of GC [7]. However, large-scale human studies remain limited and observational studies cannot fully control for confounding factors and metabolic changes, thereby restricting causal inference.
Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causal relationships between exposure and outcome, thereby minimizing confounding and avoiding reverse causation, which can affect observational studies [8]. By leveraging the random allocation of alleles at conception, MR can simulate certain aspects of a randomized controlled trial and reduce bias from unmeasured or residual confounders as well as underlying disease effects [9]. Three recent MR studies addressed the causal effects of circulating metabolites on GC risk using genome-wide association study (GWAS) data for 486 blood metabolites from the same source and GC outcome data from the UK Biobank [10–12]. In these studies, three metabolic pathways related to GC were identified: linoleic acid, valine, leucine, isoleucine, and histidine metabolism. More recently, large-scale summary statistics have become available, comprising 1,400 metabolites and metabolite ratios from the UK Biobank and GC outcome data from the FinnGen study, with histidine metabolism emerging as the most significant pathway associated with GC risk [13].
Building on this evidence and the hypothesis that additional metabolites may be involved in the metabolic pathways linked to GC risk, we expanded the data sources for both metabolites and GC by incorporating multiple GWAS datasets. Accordingly, we employed an MR approach to systematically examine metabolites associated with GC and explore the underlying metabolic pathways involved.
Methods
Data sources
Exposure data (metabolite GWAS summary statistics) were sourced from the MetaboAnalyst database (https://www.mgwas.ca/mGWAS/upload/MGBrowseView.xhtml). The initial dataset contained 337,881 independent variants associated with over 4,000 metabolite levels across 64 GWAS (r2<0.01 within 10,000 kb) excluding the UK Biobank [14]. After filtering for missing allele information and zero standard errors, 191,606 variants remained. A meta-analysis then yielded 175,130 significant and strong instruments (F-statistics>10 and P<5×10−8) across 2,286 metabolites [14]. Harmonizing these metabolite summary statistics with GC summary statistics (32,342,963 variants from FinnGen and the UK Biobank) resulted in 139,891 instrumental variables for the final MR association between 2,123 metabolites and GC (Figure 1).
As this study analyzed secondary data from previous published studies with anonymized data, ethical review and informed consent were not required.
Statistical analysis
The inverse-variance weighted (IVW) method assuming balanced pleiotropy was applied for the primary analysis [15]. In this method, we utilized single-nucleotide polymorphism (SNP)-specific Wald estimates derived by dividing the SNP-outcome association by the SNP-exposure association. These estimates were then pooled using a random-effects method, and heterogeneity was assessed using I2 (%) and Cochran’s Q test. Sensitivity analyses were conducted using the weighted median method [16] and MR-Egger regression [17]. Whereas the IVW method provides the lowest variance estimate under the assumption that all instrumental variables are valid, the weighted median method offers a balance between efficiency and robustness when some instrumental variables may be invalid. In addition, we applied MR-Egger regression, which can yield consistent estimates even in the presence of directional pleiotropy; however, this method is less efficient than IWV [15–17]. A series of complementary analytical strategies were applied to evaluate the potential influence of horizontal pleiotropy. The MR-Egger regression intercept was first estimated, with the statistical significance of the intercept differing from zero interpreted as evidence of directional pleiotropy [17]. All analyses were performed using the R statistical ver. 4.2.0 (The R Foundation) [18]. A false discovery rate correction was applied to control for multiple comparisons, with a significance threshold of Q<0.05.
To further explore the mechanisms linking metabolites and GC, we conducted enrichment analyses of GC-associated metabolic and lipid pathways using the web-based tool MetaboAnalyst ver. 6.0 (Wishart Research Group, University of Alberta) [19]. This platform integrates curated annotations from multiple databases (including Human Metabolome Database and Small Molecule Pathway Database) via the Relational Metabolic Pathway Database, ensuring broad coverage of metabolites and lipids. Its principle relies on over-representation analysis (hypergeometric testing) to determine whether significant metabolites are enriched in specific pathways more than expected by chance [19]. Compared to conventional single-database tools (e.g., Kyoto Encyclopedia of Genes and Genomes-only approaches), MetaboAnalyst provides a more comprehensive, cross-referenced metabolite library, greater flexibility in pathway definitions, and robust statistical frameworks, thereby increasing the reliability and interpretability of pathway-level insights into GC.
Results
Metabolites positively associated with GC risk and underlying pathways
MR analysis revealed 183 genetically predicted metabolites associated with an increased risk of GC under the IVW method. Of these, 25 metabolites remained significant in the sensitivity analyses using median-weighted and Egger regression methods (Table 1, Supplement 1). Among them, hexanoylcarnitine showed robust associations (odds ratio [OR], 1.06; 95% confidence interval [CI], 1.04–1.07 for the IVW method; OR, 1.09; 95% CI, 1.07–1.11 for the weighted median method; OR, 1.04; 95% CI, 1.01–1.07 for the MR-Egger method) (Figures 2A, 3A). Similarly, cis-4-decenoylcarnitine was strongly associated (OR, 1.07; 95% CI, 1.05–1.09 for the IVW method; OR, 1.14; 95% CI, 1.11–1.18 for the weighted median method; OR, 1.09; 95% CI, 1.04–1.14 for the MR-Egger method) (Figures 2B, 3B). There was no evidence of heterogeneity (Cochrane’s Q test, all P>0.999) or directional pleiotropy (MR-Egger intercept, P=0.055, P=0.077, and P=0.427, respectively). Enrichment analysis emphasized the role of metabolic and lipid processes in GC (Figure 4), with cysteine and methionine catabolism being the most significant pathways (P=6.91×10−6).
Metabolites negatively associated with GC risk and underlying pathways
Using the IVW method, MR analysis identified 175 genetically predicted metabolites that were inversely associated with the risk under the IVW method. Of these, 17 metabolites remained significant in the sensitivity analyses using the median-weighted and Egger regression methods (Table 2, Supplement 2). Among them, pregnanediol disulfate (OR, 0.95; 95% CI, 0.92–0.98 for the IVW method; OR, 0.93; 95% CI, 0.89–0.97 for the weighted median method; OR, 0.89; 95% CI, 0.83–0.96 for the MR-Egger method), acetylcarnitine (OR, 0.48; 95% CI, 0.42–0.53 for the IVW method; OR, 0.46; 95% CI, 0.39–0.53 for the weighted median method; OR, 0.31; 95% CI, 0.17–0.58 for the MR-Egger method), prolyl-hydroxyproline (OR, 0.67; 95% CI, 0.53–0.84 for the IVW method; OR, 0.66; 95% CI, 0.51–0.87 for the weighted median method; OR, 0.66; 95% CI, 0.52–0.84 for the MR-Egger method), and X-18914 (OR, 0.92; 95% CI, 0.87–0.98 for the IVW method; OR, 0.92; 95% CI, 0.87–0.97 for the weighted median method; OR, 0.82; 95% CI, 0.72–0.94 for the MR-Egger method) consistently showed robust associations (Figure 2C–F, Figure 3D–F). Furthermore, there was no evidence of heterogeneity (Cochran’s Q test, all P>0.90) or directional pleiotropy (MR-Egger intercept, P=0.096, P=0.184, P=0.790, and P=0.064, respectively). Enrichment analysis also emphasized the role of metabolic and lipid processes in GC (Figure 5), with beta-oxidation of pristanoyl-CoA (coenzyme A) (P=6.2×10−5), oxidation of branched-chain fatty acids (9.01×10−5), and peroxisomal lipid metabolism (P=0.001) emerging as the top three significant pathways.
Discussion
This study found that specific genetically predicted metabolites were associated with GC risk. Among these, 25 metabolites, especially hexanoylcarnitine and cis-4-decenoylcarnitine, were associated with a higher risk, whereas 17 metabolites, especially pregnanediol disulfate, acetylcarnitine, prolyl-hydroxyproline, and X-18914, were associated with a lower risk. The key potentially relevant metabolic pathways include cysteine and methionine catabolism, beta-oxidation of pristanoyl-CoA, oxidation of branched-chain fatty acids, and peroxisomal lipid metabolism, suggesting that alterations in these processes may be associated with GC pathogenesis.
To date, several omics studies have focused on the potential of DNA, RNA, and protein biomarkers to explore the characteristics of GC [20]. In our current review, the predictive performance of metabolites in comparisons with protein biomarkers was elucidated by Jung et al. [21] and Matsumoto et al. [22] only. These findings are in line with existing knowledge regarding the low sensitivity of carcinoembryonic antigen (9.5% vs. 95.9% for urine metabolic profiling and 10.5% vs. 69.0% for plasma metabolic profiling) and carbohydrate antigen 19-9 (12.7% vs. 95.9% for urine metabolic profiling and 10.5% and 2.86% vs. 69.0% for plasma metabolic profiling) biomarkers [21,22]. Therefore, it is suggested that the metabolomics approach has a better predictive performance than the protein biomarker approach.
Metabolic reprogramming, particularly through the Warburg effect, plays a critical role in GC by promoting glucose uptake, glycolysis, and lactate production, even under aerobic conditions [23]. This shift not only supports tumor growth, invasion, and drug resistance but also disrupts mitochondrial function and alters lipid and amino acid metabolism, contributing to poor prognosis [23]. Given that prompt prognostic surveillance may contribute to better clinical outcomes in patients with GC, and that prognosis prediction is affected by physicians’ judgment based on various clinical evaluations, such as TNM stages and histopathology, a recent systematic review of seven studies reported the predictive value of metabolites in identifying GC metastasis [24]. Accordingly, metabolites are classified into the main pathogenic pathways, including energy production, structural protein synthesis, pro-apoptotic effects, cell cycle arrest, antioxidant capacity, tumor angiogenesis, enhanced degradation of collagen extracellular matrix, mitochondrial enzyme impairment, T-cell dysfunction/inactivation, and cell viability, migration, and invasion [24].
In the present study, we determined the oxidation processes related to GC risk, notably beta-oxidation of pristanoyl-CoA, oxidation of branched-chain fatty acids, and the pathways associated with cysteine and methionine catabolism and peroxisomal lipid metabolism. These findings highlight the metabolic flexibility of GC cells, which reprogram the oxidative pathways to meet their increased energy and biosynthetic demands. Betaoxidation of pristanoyl-CoA, a peroxisomal process that breaks down branched-chain fatty acids like pristanic acid, provides an alternative energy source that supports tumor cell survival and growth, particularly under metabolic stress [25]. In addition, aberrant branched-chain fatty acid oxidation can alter cellular energy balance, redox homeostasis, and lipid signaling, thereby influencing tumor growth and progression [26,27]. Disrupted branched-chain fatty acid catabolism may also modulate oncogenic pathways such as PI3K/AKT/mTOR and AMPK/mTOR [26,27], both of which are critical in GC biology. This finding aligns with the inverse associations between total branched-chain amino acid levels and GC from the UK Biobank [28]. Besides, alterations in amino acid and lipid metabolism are central to GC biology, with methionine and cysteine catabolism influencing one-carbon metabolism, redox balance, and sulfur metabolite production, while peroxisomal lipid metabolism sustains tumor growth through fatty acid oxidation and ether lipid biosynthesis [25]. Methionine restriction suppresses proliferation, enhances antitumor immunity, and sensitizes tumors to therapy, whereas peroxisomal activity promotes survival by balancing oxidative stress, energy demands, and mitochondrial function primarily via PPAR signaling [25].
Previous MR analyses explored the associations between 486 genetically predicted circulating metabolites and GC using summary statistics from the UK Biobank (1,029 cases and 474,841 controls) or the FinnGen cohort (1,307 GC cases and 287,137 controls) [10–12]. These analyses highlighted three metabolic pathways, linoleic acid metabolism, valine/leucine/isoleucine metabolism, and histidine metabolism, and suggested potential protective roles for specific metabolites, such as 3-methyl-2-oxovalerate, piperine, and the dipeptide Phe-Phe, although the findings were inconsistent across datasets, with additional inverse associations reported for tryptophan, non-adecanoate, and erythritol. To overcome the limitations of earlier analyses that relied on single GWAS sources [29], our study integrated summary statistics from 64 metabolite GWAS and employed GC data obtained via meta-analysis of the UK Biobank and FinnGen cohorts, thereby increasing statistical power and generalizability. However, the significant metabolites and pathways identified in our study differed from those reported in previous studies, thereby offering novel insights into the metabolic basis of GC risk.
Although the increased sample size enhanced the statistical power, heterogeneity in metabolite measurements across different metabolomic platforms may introduce variability and complicate the harmonization of exposure data. In addition, discrepancies in summary statistics between FinnGen and UK Biobank, arising from differences in population structure, case definitions, and analytical pipelines, could contribute to inconsistencies in MR estimates. Another potential limitation of this study was the absence of bidirectional MR analyses. While metabolites can indeed be influenced by disease states, such as GC, our analyses were restricted to testing the causal direction from metabolites to GC risk. This was primarily due to data availability; we obtained only SNP-level associations for significant variants from the metabolite GWAS meta-analysis (64 studies) rather than complete summary statistics for each metabolite, which are required to examine the effect of GC on metabolite levels. Therefore, we cannot completely exclude the possibility of reverse causality. Future studies with publicly available full GWAS summary statistics for metabolites are essential to validate the causal relationships between metabolites and GC. Furthermore, an important limitation of our study is that both the exposure and outcome datasets were primarily derived from individuals of European ancestry. This may restrict the generalizability of our findings to populations with different genetic backgrounds and etiologies of GC. In particular, gastric adenocarcinoma is more prevalent in East Asian populations, such as Koreans and Japanese, where H. pylori infection, dietary patterns, and genetic susceptibility factors contribute more prominently to disease risk than in European populations [30]. Given the epidemiological and etiological differences in GC between European and Asian populations, replication in Asian cohorts is a critical future step to validate the causal metabolites identified in this study and better understand population-specific disease mechanisms.
In summary, this study found that genetically predicted metabolites are associated with GC risk and may serve as potential biomarkers for early detection, with 25 metabolites being linked to an increased risk and 17 to a reduced risk. These findings underscore the possible role of the disruption of amino acid metabolism and other key metabolic pathways in GC pathogenesis.
Notes
Conflict of interest
No potential conflict of interest relevant to this article was reported.
Funding
Tho Thi Anh Tran was funded by the Master, PhD Scholarship Programme of Vingroup Innovation Foundation (VINIF), code VINIF.2023.TS.119. This research was also funded by Vietnam National University Ho Chi Minh City (VNU-HCM) under the grant number C2024-44-34.
Data availability
All data for this study are available upon reasonable request to the corresponding author.
Author contribution
Conceptualization: TH, VMT, TTAT. Data curation: TH. Formal analysis: TH. Methodology: TH, VMT, TTAT. Writing–original draft: TH. Writing–review & editing: VMT, TTAT. Final approval of the manuscript: all authors.
Flowchart for identification of instrumental variables.
Figure 2
Funnel plots for Mendelian randomization (MR) analysis of hexanoylcarnitine (A), cis-4-decenoylcarnitine (B), pregnanediol disulfate (C), acetylcarnitine (D), prolyl-hydroxyproline (E), and X-18914 metabolites and gastric cancer (F). The horizontal axis represents the estimated effect of each single-nucleotide polymorphism on the exposure variable and the vertical axis represents the precision or uncertainty of these estimates. IVW, inverse-variance weighted; SE, standard error.
Figure 3
Scatter plots for Mendelian randomization (MR) analysis of hexanoylcarnitine (A), cis-4-decenoylcarnitine (B), pregnanediol disulfate (C), acetylcarnitine (D), prolyl-hydroxyproline (E), and X-18914 metabolites and gastric cancer (F). The horizontal axis denotes the impact of instrumental variables on metabolites, and the vertical axis represents the effect of instrumental variables on gastric cancer. Each black dot represents an individual single-nucleotide polymorphism (SNP), and the vertical and horizontal lines represent the corresponding 95% confidence intervals. The slope of the line represents the estimated causal effect of the various MR methods. IVW, inverse-variance weighted.
Figure 4
Enriched pathways of metabolites positively associated with gastric cancer. IMD, inherited metabolic disorder.
Figure 5
Enriched pathways of metabolites negatively associated with gastric cancer. CoA, coenzyme A; ESR, estrogen receptor.
Table 1
Genetically predicted metabolites positively associated with gastric cancer risk
1. Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63.
2. Wong MC, Huang J, Chan PS, Choi P, Lao XQ, Chan SM, et al. Global incidence and mortality of gastric cancer, 1980–2018. JAMA Netw Open 2021;4:e2118457.
3. Januszewicz W, Turkot MH, Malfertheiner P, Regula J. A global perspective on gastric cancer screening: which concepts are feasible, and when? Cancers (Basel) 2023;15:664.
4. Waddingham W, Kamran U, Kumar B, Trudgill NJ, Tsiamoulos ZP, Banks M. Complications of diagnostic upper Gastrointestinal endoscopy: common and rare: recognition, assessment and management. BMJ Open Gastroenterol 2022;9:e000688.
5. Qiu S, Cai Y, Yao H, Lin C, Xie Y, Tang S, et al. Small molecule metabolites: discovery of biomarkers and therapeutic targets. Signal Transduct Target Ther 2023;8:132.
6. Huang S, Guo Y, Li ZW, Shui G, Tian H, Li BW, et al. Identification and validation of plasma metabolomic signatures in precancerous gastric lesions that progress to cancer. JAMA Netw Open 2021;4:e2114186.
7. Yu Y, Zheng Z, Gao X, Gu Y, Zhang M, Hu B, et al. Plasma metabolomic signatures of H. pylori infection, alcohol drinking, smoking, and risk of gastric cancer. Mol Carcinog 2025;64:463-74.
8. Hu X, Cai M, Xiao J, Wan X, Wang Z, Zhao H, et al. Benchmarking Mendelian randomization methods for causal inference using genome-wide association study summary statistics. Am J Hum Genet 2024;111:1717-35.
9. Sobczyk MK, Zheng J, Davey Smith G, Gaunt TR. Systematic comparison of Mendelian randomisation studies and randomised controlled trials using electronic databases. BMJ Open 2023;13:e072087.
10. Chen S, Zhang B, Wang S, Yang M, Shen Q, Zhang R, et al. Blood metabolites and gastric cancer risk: a bidirectional 2-sample Mendelian randomization study. Medicine (Baltimore) 2025;104:e43339.
11. Zhang C, Huang DL, Zhou K, Cai JT, Liu D, Tan MH, et al. Human blood metabolites and gastric cancer: a Mendelian randomization analysis. BMC Gastroenterol 2024;24:478.
12. Qian L, Song J, Zhang X, Qiao Y, Tan Z, Li S, et al. Elucidating the causal relationship between 486 genetically predicted blood metabolites and the risk of gastric cancer: a comprehensive Mendelian randomization analysis. Front Oncol 2024;14:1418283.
13. Yu H, Liu Y, Xie P, Feng W, Xu D. Genetic evidence of circulating metabolites linking to gastric cancers: a two-sample Mendelian randomization study. Ann Med Surg 2025;87:571-6.
14. Burgess S, Davey Smith G, Davies NM, Dudbridge F, Gill D, Glymour MM, et al. Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome Open Res 2019;4:186.
15. Bowden J, Del Greco MF, Minelli C, Davey Smith G, Sheehan N, Thompson J. A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization. Stat Med 2017;36:1783-802.
16. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016;40:304-14.
17. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 2015;44:512-25.
18. R Foundation for Statistical Computing. A language and environment for statistical computing [Internet] R Foundation for Statistical Computing; 2024 [cited 2025 Nov 10]. Available from: https://www.R-project.org/
19. Pang Z, Lu Y, Zhou G, Hui F, Xu L, Viau C, et al. MetaboAnalyst 6.0: towards a unified platform for metabolomics data processing, analysis and interpretation. Nucleic Acids Res 2024;52(W1):W398-406.
20. Chen Y, Wang B, Zhao Y, Shao X, Wang M, Ma F, et al. Metabolomic machine learning predictor for diagnosis and prognosis of gastric cancer. Nat Commun 2024;15:1657.
21. Jung J, Jung Y, Bang EJ, Cho SI, Jang YJ, Kwak JM, et al. Noninvasive diagnosis and evaluation of curative surgery for gastric cancer by using NMR-based metabolomic profiling. Ann Surg Oncol 2014;21(Suppl 4):S736-42.
22. Matsumoto T, Fukuzawa M, Itoi T, Sugimoto M, Aizawa Y, Sunamura M, et al. Targeted metabolomic profiling of plasma samples in gastric cancer by liquid chromatography-mass spectrometry. Digestion 2023;104:97-108.
23. Liu Y, Zhang Z, Wang J, Chen C, Tang X, Zhu J, et al. Metabolic reprogramming results in abnormal glycolysis in gastric cancer: a review. Onco Targets Ther 2019;12:1195-204.
24. Ursu S, Ciocan A, Ursu CP, Gherman CD, Ciocan RA, Pop RS, et al. Role of metabolomics in pathogenesis and prompt diagnosis of gastric cancer metastasis: a systematic review. Diagnostics (Basel) 2023;13:3401.
25. Pratama AM, Sharma M, Naidu S, Bommel H, Prabhuswamimath SC, Madhusudhan T, et al. Peroxisomes and PPARs: emerging role as master regulators of cancer metabolism. Mol Metab 2024;90:102044.
26. Xiong H, Liu R, Xu K, Chen X, Huang L, Shou Y, et al. Branched-chain amino acid and cancer: metabolism, immune microenvironment and therapeutic targets. J Transl Med 2025;23:636.
27. Xu E, Ji B, Jin K, Chen Y. Branched-chain amino acids catabolism and cancer progression: focus on therapeutic interventions. Front Oncol 2023;13:1220638.
28. Yu L, Bao S, Zhu F, Xu Y, Liu Y, Jiang R, et al. Association between branched-chain amino acid levels and gastric cancer risk: large-scale prospective cohort study. Front Nutr 2024;11:1479800.
30. Shin WS, Xie F, Chen B, Yu P, Yu J, To KF, et al. Updated epidemiology of gastric cancer in Asia: decreased incidence but still a big challenge. Cancers (Basel) 2023;15:2639.
Assessing the impact of metabolomic markers on gastric cancer risk: a two-sample Mendelian randomization study
Figure 1
Flowchart for identification of instrumental variables.
Figure 2
Funnel plots for Mendelian randomization (MR) analysis of hexanoylcarnitine (A), cis-4-decenoylcarnitine (B), pregnanediol disulfate (C), acetylcarnitine (D), prolyl-hydroxyproline (E), and X-18914 metabolites and gastric cancer (F). The horizontal axis represents the estimated effect of each single-nucleotide polymorphism on the exposure variable and the vertical axis represents the precision or uncertainty of these estimates. IVW, inverse-variance weighted; SE, standard error.
Figure 3
Scatter plots for Mendelian randomization (MR) analysis of hexanoylcarnitine (A), cis-4-decenoylcarnitine (B), pregnanediol disulfate (C), acetylcarnitine (D), prolyl-hydroxyproline (E), and X-18914 metabolites and gastric cancer (F). The horizontal axis denotes the impact of instrumental variables on metabolites, and the vertical axis represents the effect of instrumental variables on gastric cancer. Each black dot represents an individual single-nucleotide polymorphism (SNP), and the vertical and horizontal lines represent the corresponding 95% confidence intervals. The slope of the line represents the estimated causal effect of the various MR methods. IVW, inverse-variance weighted.
Figure 4
Enriched pathways of metabolites positively associated with gastric cancer. IMD, inherited metabolic disorder.
Figure 5
Enriched pathways of metabolites negatively associated with gastric cancer. CoA, coenzyme A; ESR, estrogen receptor.
Graphical abstract
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Graphical abstract
Assessing the impact of metabolomic markers on gastric cancer risk: a two-sample Mendelian randomization study
Genetically predicted metabolites positively associated with gastric cancer risk
