Characterization of antioxidant qualities and exploration of mechanisms in different colored onions: Insights from metabolomics and systems biology Authors: Chenglong He, Johane Johari Mkunga, Dongsheng Zhang, Muhammad Iqbal Choudhary, Wenxi Zhai, Xinrong Xie, Liyue Fei, Chunhui Shan, Xinquan Yang, Wenchao Cai Published: 2026 ABSTRACT AbstractOnion color correlates with nutritional value, however, systematic studies assessing antioxidant properties across different colored varieties remain limited. This research integrates chemometrics, metabolomics, and systems biology to evaluate antioxidant quality and underlying mechanisms. Purple onions exhibited the highest total phenolic and flavonoid content, along with superior water-soluble free radical scavenging and reduction capacity. In contrast, white onions excelled in scavenging fat-soluble free radicals and superoxide anions. Key marker metabolites identified included fatty acyls for purple onions, flavonoids for white onions, and carboxylic acids for yellow onions. The in vivo antioxidant effects were mediated through core targets SRC, AKT1, EGFR, MMP9, and TNF, primarily via the VEGF signaling pathway, which was preliminarily validated through molecular docking studies. These findings elucidate the metabolic basis of color-related antioxidant activity, thereby supporting the development of functional foods.Keywords: Onion, Color, Metabolomics, Systems biology, Antioxidant quality characteristics Introduction Redox reactions continuously occur in the human body to maintain internal homeostasis. Oxidative stress refers to the production and accumulation of Reactive Oxygen Species (ROS) in cells. The balance between cellular antioxidant defenses and oxidants can be easily disrupted by aging and external environmental factors, leading to oxidative stress, which causes cellular damage and is linked to the pathogenesis of cardiovascular diseases, cancers, and other illnesses (Zheng et al., 2024). With the improvement of consumer living standards and the popularization of nutritional concepts, fruits and vegetables—natural dietary resources—are receiving increasing attention in the field of antioxidants (Peng et al., 2025). The functional components of these foods mainly derive from secondary metabolites such as polysaccharides, polyphenols, terpenoids, and alkaloids (Cai et al., 2025). These constituents are known to effectively scavenge free radicals (Fei et al., 2025). It has become a research hotspot in the field of food science and nutrition to screen horticultural crops rich in natural antioxidants and clarify their antioxidant mechanism, which provides a theoretical basis for the development of natural, safe and efficient functional foods. Therefore, it is crucial to conduct accurate research on the health effects of these functional components to create high-value fruit and vegetable products. Onion (Allium cepa L.) is an important vegetable crop in the Allium genus that is widely cultivated worldwide (Zhao et al., 2024). It is characterized by a distinctive flavor profile and a large fleshy bulb that serves as an edible section. Onions have long been utilized globally as a spice or consumed raw, possessing significant economic value (Yeshiwas et al., 2024). They are rich in chemical constituents with various pharmacological functions, including antibacterial and anti-inflammatory properties, which hold great potential for healthcare applications (Shabir et al., 2022). The antioxidant properties of onions are particularly noteworthy, as they are abundant in natural antioxidants such as flavonoids, which play a critical role in reducing free radical-induced damage in the human body (Roldán-Marín et al., 2009) and have been shown to significantly reduce oxidative stress levels through numerous clinical and preclinical studies (Gupta et al., 2025). Bulb color is an essential phenotypic trait of onions closely related to the accumulation of key functional metabolites, representing one of the most important characteristics linked to their nutritional value and commercial appeal (Han et al., 2023). Kaur et al. (2020) analyzed the physical and chemical properties of 30 onion genotypes cultivated in the subtropical region of India and found that onion genotypes could be classified based on color, with varieties of the same color showing similar trends in metabolite accumulation. Özcan et al. (2022) noted differences in macroelement and trace element contents among onions of varying colors. However, most of the existing studies focus on the comparison of single antioxidant index or nutritional components between onions of different colors, and lack systematic analysis of the specific correlation between metabolite profiles and antioxidant activity; moreover, the research on the antioxidant mechanism of onion is mostly limited to in vitro activity verification, and there is a lack of in-depth excavation of molecular targets, signaling pathways and metabolite-target interaction networks in vivo. These have formed a key research gap: the metabolic basis of onion color-related antioxidant activity and its possible antioxidant regulation mechanism in vivo are not yet clear, which restricts the precise development and high-value utilization of onion-derived antioxidant functional foods. The abundant metabolites in plants not only play a crucial role in growth, development, and color phenotypes but are also closely associated with health effects in humans (Bhuiyan, 2026). Metabolomics primarily involves a global analysis of small molecules and metabolic networks within organisms. Compared to transcriptome analysis, metabolomics can rapidly identify differential metabolites across different samples (He, Peng et al., 2025). However, a singular focus on metabolomics presents challenges, such as insufficient exploration of potential functional substances, unclear molecular mechanisms of multi-target regulation, and vagueness concerning metabolic pathways and in vivo metabolites. Traditional evaluations of food nutritional value focus mainly on individual nutrients. However, the recent advancement of databases allows bioinformatics—based on systems biology—to efficiently identify active food ingredients and their potential biological targets, offering a multidimensional analytical approach and a comprehensive understanding (Shen et al., 2024). By constructing network models, the relationship between food components and human health can be systematically analyzed. For instance, He, Mkunga, Zhang, Mao et al. (2025) utilized bioinformatics methods to demonstrate that the antioxidant mechanisms in mulberry wine are highly correlated with the PI3K/Akt signaling pathway. Therefore, combining metabolomics with systems biology analysis can deepen our understanding of the mechanisms of functional factors in food products and clarify the in vivo biochemical processes related to healthy effects (He, Zhang, Mkunga et al., 2025). The integration of metabolomics and systems biology can realize the organic combination of “metabolite screening” and “mechanism excavation”, which is a powerful technical means to solve the above research gaps and clarify the in vivo antioxidant mechanism of food functional components (He et al., 2025, Shen et al., 2024). The purpose of this study is to fill the above research gaps and provide a scientific basis for the precise development of onion-derived functional foods. Taking purple, yellow and white onions as the research objects, chemometrics, non-targeted metabolomics and system biology techniques were integrated to systematically analyze the differences in metabolite profiles and in vitro antioxidant activity specificity of onions of different colors, identify common key antioxidant metabolites in different onions and predict key antioxidant mechanisms, and preliminarily verify the binding affinity of metabolites to targets through molecular docking technology. This study achieved a progressive analysis from in vitro antioxidant activity characterization to in vivo molecular mechanism mining, breaking through the limitations of single technology application and focusing only on in vitro research in previous related studies. The innovation of this study is reflected in three aspects: First, the color-specific antioxidant activity characteristics of purple, yellow and white onions and their corresponding metabolite basis were clarified by metabolomics technology for the first time; secondly, the core targets and regulatory networks of onion antioxidant effect in vivo were systematically identified, which realized the extension from in vitro research to in vivo mechanism analysis and provides a new method and research perspective for the quality evaluation and high-value utilization of onions. The third is to integrate multiple technologies to establish an interdisciplinary technical system for analyzing the antioxidant mechanism of horticultural crops, which provides a reproducible research paradigm for the functional quality research of other colored fruits and vegetables. Material and methods Plant material Onions with different colors were collected upon reaching commercial maturity in Akesu, Xinjiang, China. All onions were cultivated under identical soil conditions and cultivation patterns to minimize variations from external factors. The onions were categorized as Purple Onion (PO), Yellow Onion (YO), and White Onion (WO). Five independent biological replicates were prepared for each cultivar (n = 5). The bulbs, approximately 2–3 mm in thickness, were carefully excised with a scalpel, frozen in liquid nitrogen, and stored at −80 °C for analysis. Chemical antioxidant assays of onions with different colors Samples were evaluated for antioxidant activity using several assays, including 2,2-Azinobis (3-ethylbenzothiazoline-6-sulfonic acid) (ABTS+) radical scavenging activity, 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging activity, ferric reducing antioxidant power (FRAP), superoxide anion scavenging capacity (SASC), total phenolic content (TPC), and total flavonoid content (TFC), utilizing commercial assay kits from Suzhou Grace Biotechnology Co., Ltd. (Suzhou, China), following the manufacturer's instructions. TPC was expressed as gallic acid equivalents (μg GAE/mL), TFC as rutin equivalents (mg RE/L), and FRAP as μmol FeSO4/mL; ABTS+, DPPH radical scavenging activity, and SASC were expressed as percentages. Ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS) analysis of onions with different colors The instrument utilized for UHPLC-MS analysis was an ultra-high-performance liquid chromatography tandem time-of-flight mass spectrometry system (UHPLC-Triple TOF 6600, AB SCIEX). Sample Pretreatment: The sample preparation protocol was as follows: precisely 25 ± 5 mg of each specimen was weighed into a 2 mL centrifuge tube containing a 6 mm grinding bead. Subsequently, 400 μL of extraction solvent (methanol: water, 4:1 v/v), supplemented with an internal standard (L-2-chlorophenylalanine at 0.02 mg/mL), was added. The mixture was homogenized using a cryogenic tissue grinder (-10 °C, 50 Hz, 6 min), followed by low-temperature ultrasonic extraction (5 °C, 40 kHz, 30 min). To ensure complete protein precipitation, the samples were incubated at −20 °C for 30 min. After centrifugation (13,000 ×g, 4 °C, 15 min), the supernatant was transferred to an injection vial with an insert for LC-MS/MS analysis. UHPLC-MS Detection: A chromatographic column (ACQUITY UPLC HSS T3, 100 mm × 2.1 mm i.d., 1.8 μm; Waters, Milford, USA) was operated at a column temperature of 40 °C. mobile phase A consisted of 5% aqueous acetonitrile containing 0.1% formic acid, while mobile phase B comprised 47.5% isopropanol, 47.5% acetonitrile, and 5% water (with 0.1% formic acid) at a flow rate of 0.4 mL/min. The injection volume was 10 μL. The samples were ionized by electrospray ionization, and mass spectrometry signals were collected in both positive and negative ion scanning modes, utilizing an electrospray voltage of +5.5 kV and − 4.5 kV, respectively. During analysis, one QC sample was inserted after every 5 to 15 analyzed samples to monitor assay quality. Compound identification was performed as described by He, Mkunga, Zhang, Choudhary et al. (2025). Systems biology analysis of antioxidant mechanisms exerted by onions in vivo Pearson correlation analysis was conducted on the expression levels of FRAP, SASC, DPPH, ABTS+, and expression levels of metabolites in onion samples. Metabolites that significantly and positively correlated with more than two in vitro antioxidant indicators (p < 0.01) were selected for systems biology analysis of antioxidant mechanisms in vivo. The SMILE formula of the target metabolite was obtained from PubChem and subsequently imported into Swiss Target Prediction to predict its targets. The species was set to “Homo sapiens,” and targets with “Possibility ≥ 0.2” in the derived results were retained to ensure high reliability. Cytoscape was then used to build the metabolite-target network. We used the keyword “Oxidative Stress” to search for oxidative stress-related targets in the GeneCards database, retaining credible targets with a relevance score of ≥10. These targets were then intersected with metabolic targets to create a Venn diagram. The intersection targets were imported into the String database to build a protein-protein interaction (PPI) network with a confidence level of 0.700. The resulting protein interaction network was saved as TSV format files, which were then imported into Cytoscape. The cytoHubba plugin was utilized to perform topology analysis based on Degree, Maximal Clique Centrality (MCC), Maximal Neighborhood Connectivity (MNC), and Edge Percolated Component (EPC) methods. The top five targets were identified as core targets, with the intersection across these three methods representing the final core target set. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the intersection targets were conducted to explain the antioxidant mechanisms. Additionally, molecular docking was employed to explore the binding affinity of small molecules to target protein binding sites, thereby preliminarily verifying analysis results. AutoDock Vina was used to calculate molecular binding energy, with the lowest binding energy reflecting molecular docking results. A binding energy less than 0 kcal/mol indicates that the target protein can spontaneously bind to the active ingredient. Furthermore, a binding energy less than -5.0 kcal/mol suggests a strong interaction with the active ingredient, while smaller binding energies indicate greater stability in the docking results (Song et al., 2023). Statistical analysis One-way ANOVA with Tukey's post-hoc test, Pearson Correlation Analysis, Principal component analysis (PCA), partial least squares discrimination analysis (PLS-DA), KEGG enrichment analysis, linear discriminant analysis effect size (LEfSe), and Pearson correlation analysis were performed using R software (version 4.4.3, https://www.r-project.org/). All figures were generated using R software. Results and discussion Comparison of in vitro antioxidant activity of different colored onions Chemical antioxidant assays are among the most common and straightforward methods for assessing the in vitro antioxidant capacity of fruits and vegetables. Phenolic compounds and flavonoids serve as key antioxidant components in plants; the phenolic hydroxyl group within the polyphenol structure scavenges free radicals, while flavonoids, characterized by two aromatic rings and one oxygen heterocyclic ring, effectively scavenge free radicals, thereby exerting antioxidant effects (Cai et al., 2020; Cai et al., 2024). Thus, TPC and TFC can reflect the antioxidant capacity of the samples from a particular perspective. FRAP, SASC, ABTS+ free radical scavenging activity and DPPH free radical scavenging activity collectively provide insight into the antioxidant properties of different colored onion samples. In this study, we conducted a comparative analysis of antioxidant activity among various colored onions by measuring TPC, TFC, FRAP, SASC, ABTS+ free radical scavenging activity, and DPPH free radical scavenging activity (Fig. 1A-F). and further explored the internal association of antioxidant indices via Pearson correlation analysis (Fig. 1G). The results demonstrated that PO had the highest levels of TPC (1592.44 μg GAE/mL), TFC (0.42 mg RE/L), FRAP (7.94 μmol FeSO4/mL), and ABTS+ radical scavenging activity (92.34%)., with significant differences observed compared to both YO and WO (p < 0.001). SASC and DPPH showcased their highest levels in WO (55.82% and 44.93%, respectively), with significant differences noted between PO and YO (p < 0.05). Fig. 1G revealed a highly significant positive correlation (p < 0.001) between TPC, TFC and FRAP, ABTS+ free radical scavenging activity in all three onion varieties, with the correlation coefficient exceeding 0.85; in contrast, SASC and DPPH free radical scavenging activity presented a significant positive correlation (p < 0.05) independent of the above indices. Comparison of TPC (A), TFC (B), ABTS+ (C), FRAP (D), SASC (E), DPPH (F), and correlation heatmap of antioxidant parameters (G) in PO, WO and YO. Heatmap of Pearson Correlation Analysis of Antioxidant and Basic Indices (G). Data are presented as mean ± SD (n = 5). * represents p < 0.05, ** represents p < 0.01, *** represents p < 0.001, **** represents p < 0.0001(One-way ANOVA with Tukey's post-hoc test). Antioxidant activity typically results from the synergistic effects of multiple antioxidant components, and the interactions among various antioxidants in onions of different colors vary, leading to distinct antioxidant properties. WO exhibited significant potential for DPPH radical scavenging activity and SASC, suggesting its unique ability to neutralize free radicals in both fat-soluble environments and against superoxide anions. In contrast, PO had the highest levels of TPC and TFC, establishing a solid basis for its comprehensive antioxidant capacity, with superior scavenging of ABTS+ radicals and FRAP, indicating a unique capability to scavenge free radicals in water-soluble environments and to function as a reducing agent. The strong positive correlation between TPC, TFC, FRAP and ABTS+ free radical scavenging activity indicates that phenols and flavonoids in onion are the core substances that jointly regulate the scavenging ability and reducing ability of water-soluble free radicals. The accumulation of these two types of substances can simultaneously enhance the two water-soluble antioxidant activities (Lim et al., 2024). The significant correlation between SASC and DPPH free radical scavenging activity confirmed that the scavenging of lipid-soluble free radicals and superoxide anions in onion was mediated by the same type of antioxidant components, which formed an independent regulatory system different from the water-soluble antioxidant system (Karpinska & Foyer, 2024).The metabolites' types and contents of plants under different color phenotypes will display differences, which, in turn, affect their antioxidant capacity (Zhang et al., 2023). To investigate the reasons for the differences in antioxidant capacity among different colored onions, we analyzed the relationship between their metabolites and antioxidant capacity. PCA and PLS-DA of different colored onion metabolites We analyzed the metabolite compositions of onions of three different colors using an untargeted metabolomics approach. A total of 1442 metabolites were detected in PO, 1314 in YO, and 1397 in WO. The metabolomics data from different colored onion samples were downscaled using unsupervised PCA, with the score plots shown in Fig. 2A. The relatively close distances within the groups of PO, YO, and WO, along with the greater distances among the groups, indicate that metabolite profiles differ more between colors than within the same color, which is a typical metabolic differentiation characteristic of horticultural crops with distinct phenotypic traits. PCA score plot (A). PLS-DA inertia bar chart (B), significance diagnostics (C), observation diagnostics (D), and score plot (E). To further analyze the similarities and differences among the groups, we employed supervised PLS-DA. Fig. 2B shows that the inertia bar chart demonstrated R2Y and Q2Y values of 0.990 and 0.981, respectively. R2Y and Q2Y evaluate the explanatory and predictive capabilities of the PLS-DA model, respectively, indicating good predictive ability. The R2Y and Q2Y values were compared with corresponding values obtained from random permutation testing (Fig. 2C). Statistical analysis confirmed that the predicted R2Y and Q2Y values were less than the actual values (pR2Y = 0.05, pQ2 = 0.05). Furthermore, the metabolic variations among different colored onions were statistically significant according to Tukey's post-hoc test (p < 0.05). The observation diagnostic plot illustrates the distance of each sample in the projection plane (Fig. 2D), showing no outliers, which indicates the absence of anomalies in the model. Fig. 2E shows all PO samples were located in the first quadrant, YO samples in the second quadrant, and WO samples in the fourth quadrant, further validating the conclusions drawn from the PCA. In summary, the combined application of unsupervised PCA and supervised PLS-DA achieved a comprehensive characterization of the metabolic differences among PO, YO and WO. The consistent results of the two multivariate statistical methods not only revealed the obvious metabolic differentiation of onions with different color phenotypes, but also provided a statistical basis for the subsequent screening of differential metabolites and marker metabolites (He, Mkunga, Zhang, Choudhary et al., 2025). The clear metabolic clustering and no outlier characteristics also lay a solid foundation for the subsequent KEGG pathway enrichment analysis and systems biology research, ensuring the accuracy and scientificity of the follow-up mechanism exploration. Identification of metabolites significantly enriched in different color onions LEfSe analysis is a statistical approach used to identify high-dimensional biomarker features. It detects significant differences in feature abundance using the Kruskal-Wallis sum-rank test, facilitating the identification of biologically relevant differential metabolites among numerous compounds. We conducted LEfSe analysis on the metabolomics data to identify metabolites significantly enriched in different colored onions, applying screening criteria of p < 0.05 and LDA > 2.00 (Fig. 3A). Identification of enriched metabolites in onions of different colors (A) and class-level metabolite enrichment (B). Significantly enriched metabolites (p < 0.05 and LDA > 2.00) in PO (C), WO (D) and YO (E). At the classification level, each colored onion exhibited high expression of certain classes of metabolites that may serve as potential biomarkers (Fig. 3B). Specifically, isoform classes of metabolites that were significantly enriched in PO included isoflavonoids, prenol lipids, fatty acyls, pyrrolidines, phenols, tannins, azolidines, purine nucleotides, pyrimidine nucleosides, and pyrrolopyrimidine nucleosides and nucleotides. For WO, the significantly enriched classes consisted of purine nucleosides, flavonoids, cinnamic acids and derivatives, hydroxy acids and derivatives, anthracyclines, sulfoxides, and pyrans. YO demonstrated significant enrichment in lactones, sphingolipids, carboxylic acids and derivatives, pyridines and derivatives, anthracenes, pteridines and derivatives, organic carbonic acids and derivatives, furans, and benzoxazines. A total of 157 metabolites were identified as significantly enriched across the three onion colors, with 39 in PO (Fig. 3C), 39 in WO (Fig. 3D), and 45 in YO (Fig. 3E). Fatty acyls exhibited the highest LDA score of 3.92, suggesting their potential as biomarkers for PO. Among specific metabolites, gamma-L-glutamyl-S-(1-propenyl)-cysteine sulfoxide achieved the highest LDA value of 4.01, indicating its potential as a marker metabolite for PO. The phenolic hydroxyl groups in the structure of isoflavonoids provide hydrogen atoms that bind to free radicals, thereby scavenging them and exerting antioxidant effects. Prenol lipids, which are typically lipophilic, can integrate into cell membranes to enhance membrane stability and prevent lipid peroxidation (Levental et al., 2016). Pyrrolidines feature a cyclic structure with nitrogen atoms, which can react with free radicals, serving as scavengers. The phenolic hydroxyl groups in phenols and tannins are critical for their antioxidant effects, as they can donate hydrogen atoms to bind with free radicals, terminating the free radical chain reaction. Azolidines also possess a cyclic structure with nitrogen atoms capable of reacting with free radicals. Furthermore, pyrrolidines, phenols, and azolidines can inhibit metal ion-catalyzed oxidation reactions through chelation, thus enhancing the antioxidant capacity of PO. Tannins can protect proteins from oxidative stress by binding to them to form stable complexes. Purine nucleotides, pyrimidine nucleosides, and pyrrolopyrimidine nucleosides and nucleotides can enhance cellular antioxidant capacity by modulating the activity of antioxidant enzymes. Notably, fatty acyls, which contain a double bond structure, are prone to oxidation and free radical formation. However, under certain conditions, they can also act as antioxidants by donating hydrogen atoms or electrons to react with free radicals, thereby preventing oxidative chain reactions; thus, fatty acyls exhibit a dual role in the antioxidant capacity of PO. Flavonoids exhibit the highest LDA score of 4.64, indicating their potential as biomarkers for WO. Among the specific metabolites, quercetin 3-beta-D-glucoside showed the highest LDA value of 4.62, establishing its role as a marker metabolite for WO. Flavonoids possess multiple antioxidant mechanisms; their phenolic hydroxyl groups donate hydrogen atoms or electrons to neutralize free radicals, leading to direct scavenging of these reactive species (Shen et al., 2022). Additionally, flavonoids can inhibit oxidative enzymes to decrease the production of ROS, such as xanthine oxidase, which subsequently reduces superoxide anion levels. They also prevent oxidation reactions catalyzed by metal ions through chelation, thereby exerting an antioxidant effect and diminishing the catalytic activity of metal ions in lipid peroxidation (Shen et al., 2022). Cinnamic acids and their derivatives also demonstrate strong antioxidant activity and high scavenging capacity for superoxide anions (Gryko et al., 2021). They inhibit metal-catalyzed oxidation via chelation, contributing to their antioxidant effects, which may be a primary reason for the elevated SASC in WO (Shen et al., 2022). Hydroxy acids and derivatives typically feature functional groups, such as hydroxyl and carboxyl groups, which enable them to neutralize free radicals through hydrogen atom or electron donation, thereby preventing free radical chain reactions (Moussa & Al-Mamary, 2021). Sulfoxides possess sulfur atoms with strong electrophilic properties and electron-donating abilities that enable them to react with free radicals, offering additional antioxidant effects. Furthermore, sulfoxides can act as antioxidant cofactors, participating in the body's redox cycles and enhancing overall antioxidant performance. The cyclic structure of pyrans imparts a degree of chemical stability, allowing them to absorb and dissipate free radical energy, preventing the propagation of free radical chain reactions. Certain pyrans containing o-dihydroxy structures can form quinone-type intermediates that more effectively scavenge free radicals. Additionally, pyranoids may inhibit metal-catalyzed oxidation through chelation, thereby enhancing antioxidant capacity. Metalloid elements within metalloidorganides can induce the expression of antioxidant-related genes, influencing the body's antioxidant properties (Takeda et al., 2019). Carboxylic acids and their derivatives have the highest LDA score of 5.73, suggesting their potential as metabolic markers for YO. Among the specific metabolites, gamma-glutamyl-gamma-glutamyl-S-methylcysteine displayed the highest LDA value of 4.47, serving as a marker metabolite for YO. Lactones, characterized by hydrophobic groups, stabilize lipid membranes and reduce lipid peroxidation. Benzoxazines, featuring a benzene ring and nitrogen atom, provide electrons to inhibit oxidative reactions, enhancing the antioxidant properties of plants. The furan ring in furans offers a conjugated system, facilitating electron transfer and antioxidant activity. Sphingolipids function as signaling molecules that regulate intracellular antioxidant enzymes, such as superoxide dismutase and catalase, thereby enhancing cellular antioxidant capacity. In summary, the metabolites characteristic of different colored onions confer unique antioxidant mechanisms and properties. The antioxidant capacity of PO relies on the synergistic action of multiple metabolite classes, including free radical scavenging (isoflavonoids), membrane stabilization (prenol lipids), and enzyme activity modulation (purine/pyrimidine-like nucleotides). The elevated SASC in WO is primarily attributed to cinnamic acids and derivatives, which efficiently scavenge superoxide anions and chelate metal ions. YO leverage metabolites such as lactones (which stabilize membrane structures), sphingolipids (which regulate antioxidant enzymes), and benzoxazines (which scavenge free radicals) to form a unique antioxidant network. KEGG pathway analysis of metabolic profiles of different colored onions The total metabolite profiles of different colored onions were enriched in the KEGG database. A total of 93 metabolic pathways were enriched in PO, of which 26 were significantly associated (p < 0.05) (Fig. 4A). In WO, 95 metabolic pathways were identified, with 28 significantly associated (p < 0.05) (Fig. 4B). YO exhibited 77 enriched metabolic pathways, of which only one was significantly associated (p < 0.05) (Fig. 4C). Pathways of significant enrichment of metabolites (p < 0.05) in PO (A), WO (B). The top 10 pathways of metabolites enrichment in YO (C). The same 25 significant pathways were observed in both PO and WO, with a particular focus on flavonoid biosynthesis, phenylpropanoid biosynthesis, metabolism of cysteine and methionine, ascorbate and aldarate metabolism, and biosynthesis of various plant secondary metabolites. These pathways are directly related to the generation of antioxidant activity in onions. Flavonoid biosynthesis in PO displayed the highest significance and may play a critical role in determining its antioxidant properties. The phenylpropanoid biosynthesis pathway is central for synthesizing antioxidants such as flavonoids, lignans, and coumarins. Phenylpropane metabolites, including caffeic acid and ferulic acid, function through direct free radical scavenging and enhancement of antioxidant enzyme activities. Cysteine, a vital component of glutathione (GSH), is one of the key water-soluble antioxidants in plants (Cheng et al., 2020). GSH plays a role in scavenging H₂O₂ via the ascorbic acid-glutathione cycle (ASA-GSH cycle) and helps maintain intracellular redox balance. Ascorbate is among the most important antioxidants in plants, capable of indirectly enhancing fat-soluble antioxidant capacity by regenerating vitamin E. The biosynthesis of various plant secondary metabolites encompasses the synthesis of flavonoids, terpenes, phenolics, and other antioxidants that exert antioxidant activity. The biosynthesis of cofactors involves synthesizing ubiquinone and vitamin E. Ubiquinone, a crucial component of the mitochondrial electron transport chain, protects membrane structure by scavenging lipid peroxidation products, while vitamin E inhibits lipid peroxidation by trapping lipoxygen radicals. From the WO enrichment analysis, three unique pathways emerged: flavone and flavonol biosynthesis, tyrosine metabolism, and cyanoamino acid metabolism. Among these, flavone and flavonol biosynthesis exhibited the highest significance in the enrichment results. This pathway may largely influence the formation of antioxidant quality in WO. Flavone and flavonol biosynthesis contributes to antioxidant properties through the production of flavonoids and flavonols, catalyzed by chalcone isomerase (CHI) and flavonoid synthase (FNS). Although TFC in WO is low, its antioxidant activity may relate to precursors from this pathway, such as phenylpropanes. For instance, cinnamic acid derivatives produced via the phenylpropane pathway can further convert to phenolic compounds with antioxidant activity through tyrosine metabolism. Consequently, the flavone and flavonol biosynthesis pathway may indirectly contribute to the antioxidant processes in WO by supplying precursors for other pathways. In YO, phenylalanine metabolism serves as the core driving pathway for generating its antioxidant activity, with multiple phenolic compounds produced through this pathway directly contributing to YO's antioxidant capacity. Systems biology analysis of onion antioxidant mechanisms To elucidate the key antioxidant metabolites in onions and the mechanisms of their in vivo antioxidant action, we conducted Pearson's correlation analysis on four in vitro antioxidant indices and metabolite expression. This analysis identified a total of 90 metabolites that exhibited highly significant positive correlations (p < 0.01) with more than two in vitro antioxidant activities (Fig. 5). These metabolites were initially classified as antioxidant metabolites in onions for subsequent correlation mechanism studies. Identification of metabolites significantly positively correlated with in vitro antioxidant indicators (p < 0.01). Identification of key antioxidant metabolites and their targets We identified 29 metabolites as key antioxidant compounds in onions for further analysis. Screening revealed a total of 29 metabolites associated with 131 reliable targets (post-de-emphasis), indicating that these metabolites are likely to play a functional role in vivo rather than merely demonstrate a statistical correlation. A metabolite-target network comprising 161 nodes and 250 edges was constructed using Cytoscape (Fig. 6A). The metabolite information was organized according to degree values, as summarized in Table 1. Metabolites-target network diagram (A). The intersection of Metabolite targets and Oxidative Stress targets (B). Table 1Target information of antioxidant metabolites.No.MetabolitesDegreeM1Myricetin87M2(R)-Laudanidine14M3Vitamin P13M46″-Malonylastragalin12M5Kaempferol 3-O-Galactoside11M6Myricetin 3-(3″,4”-Diacetylrhamnoside)11M7(−)-Armepavine10M8(S)-Reticuline9M9Protocatechuic Acid9M1013,16,19-Docosatrienoic Acid7M112-Aminoheptanedioic Acid6M12Diosmetin 7-Neohesperidoside5M13Fenofibric Acid5M14Nelumboside5M15Prodelphinidin B5M16Prodelphinidin B15M17Cyanidin4M18Cynaroside4M19Carboxymethyl Chitosan3M20Dihydroisorhamnetin3M21Ganoderiol A3M22Palmitoleamide3M23Patuletin3M24Sinensetin3M2513-Hode2M263′,5’-Cyclic Gmp2M27Alisol A2M28Delphinidin 3-(6’-Acetylgalactoside)2M29S-Sulfanylglutathione2 Target information of antioxidant metabolites. Additionally, a total of 979 reliable oxidative stress targets were obtained by searching and screening the GeneCards database. The intersection of these targets yielded 48 common targets (Fig. 6B), which may represent potential targets for the in vivo antioxidant actions of onions. Analysis of protein interactions Metabolic processes in organisms are regulated by a multitude of molecules and proteins that interact with one another within a complex network. By analyzing protein interactions, we can identify key nodes within this network to elucidate the relationships between the target proteins. A total of 48 targets were examined through protein interaction analysis, resulting in 42 nodes, 242 edges, an average number of neighbors of 5.095, and a network centralization of 0.279 (Fig. 7A). These findings indicate that the protein interactions exhibited moderate connectivity, suggesting frequent interactions but with some centralization, specifically, some central nodes significantly impact the structure and functionality of the network. These central nodes may represent proteins that play crucial roles in regulating biological processes. Protein-Protein Interaction network diagram (p<1.0e-16) (A). The core targets network diagram obtained by Degree (B), MNC (C), EPC (D) and MCC (E) methods (Top5). The intersection diagram of core targets obtained by three topological methods (F). Different topological methods, based on distinct biological principles and computational models, can capture various properties of protein networks from multiple perspectives. By employing intersection analysis, we can integrate the strengths of these methods to enhance the reliability of identifying core targets. The top five scored targets were identified following screening with four topological methods—Degree, MCC, MNC, and EPC (Fig. 7B-E). The intersection of these results yielded the core targets: SRC, AKT1, EGFR, MMP9, and TNF (Fig. 7F). These core targets are key switches in oxidative stress-mediated apoptosis pathways, suggesting that key antioxidant metabolites in onion not only scavenge free radicals in vitro, but also play a role by regulating the host ‘s antioxidant signaling network. However, merely identifying these core targets is insufficient to systematically explain the biological processes and pathways associated with antioxidant effects in vivo; therefore, the 48 antioxidant targets were further analyzed using GO and KEGG enrichment analyses. GO enrichment analysis GO enrichment analysis yielded a total of 1999 significantly correlated results (p < 0.05), comprising 1767 results at the biological process (BP) level, 78 results at the cellular component (CC) level, and 154 results at the molecular function (MF) level. The top 10 enrichment scores for each of the three levels were selected for visualization (Fig. 8A). GO function analysis histogram (A). KEGG enrichment analysis of dot plots of significant pathways (B), the classification of them (C). Molecular docking minimum binding energy heat map (D). At the BP level, the analysis mainly focused on cellular responses to chemical stress and ROS metabolic processes, regulation of inflammatory responses, oxidative stress responses, regulation of cysteine-type endopeptidase activity involved in apoptosis, andPI3K signaling pathways that help resist oxidative stress. Biological processes such as ROS metabolic processes and cellular responses to oxidative stress directly exert antioxidant functions and activate signaling pathways that counteract oxidative stress, such as the Nrf2 signaling pathway. As a key transcription factor, Nrf2 regulates the expression of several antioxidant genes, enhancing cellular antioxidant capacity and detoxification abilities, thereby mitigating oxidative stress damage and maintaining cellular homeostasis. The PI3K signaling pathway, the regulation of inflammatory responses, and the cellular response to oxidative stress primarily facilitate antioxidant-related signal transduction and gene expression regulation. Activation of PI3K phosphorylates downstream Akt, which enhances the expression of antioxidant enzymes by regulating the Nrf2 signaling pathway. Inflammatory responses are linked to oxidative stress; when inflammation occurs, it promotes ROS release, leading to excessive oxidative damage and creating a vicious cycle. Regulation of the inflammatory response acts antioxidant effects by inhibiting the NF-κB signaling pathway, thereby reducing pro-inflammatory cytokine production and, consequently, ROS release. Meanwhile, activation of the Nrf2 signaling pathway enhances the expression of antioxidant enzymes while decreasing inflammation-induced oxidative stress. The regulation of cysteine-type endopeptidase activity involved in apoptosis can limit the degradation of antioxidant proteins, maintaining intracellular levels of these proteins and enhancing cellular antioxidant capacity. At the CC level, cellular components such as membrane rafts, membrane microdomains, caveolae, extrinsic components of membranes, and vesicle lumen play significant roles in exerting antioxidant effects. Membrane rafts and microdomains are specialized lipid-rich regions of the cell membrane containing signaling molecules and receptors. In response to oxidative stress, signaling molecules within these microstructures can rapidly aggregate, activating related signaling pathways such as the Nrf2 pathway, which triggers antioxidant responses, regulates the expression of antioxidant genes, and enhances cellular antioxidant activity. The extrinsic components of membranes and the cytoplasmic side of the plasma membrane sense external oxidative stress and activate intracellular antioxidant defense mechanisms through signaling. The vesicle lumen, which houses antioxidant substances such as glutathione, protects the vesicle membrane and its contents from oxidative damage during vesicle transport, participating in the regulation of intracellular redox homeostasis. Antioxidant enzymes and substances located on organelle membranes help prevent ROS leakage into the cytoplasm, thereby protecting the organelle's structure and function while maintaining intracellular redox homeostasis. At the MF level, several molecular functions, including phosphatase binding, transmembrane receptor protein tyrosine kinase activity, protein tyrosine kinase activity, heme binding, and ligand-activated transcription factor activity, contribute to antioxidant effects. Protein tyrosine kinase activity and related activities primarily engage in signal transduction and regulation; for example, proteins binding to phosphatases can regulate their activity, impacting intracellular redox homeostasis. Metabolites can inhibit protein phosphatase activity by binding to them, subsequently reducing ROS production or increasing antioxidant enzyme activity. Additionally, specific binding proteins associated with phosphatases may affect the activation of antioxidant-related signaling pathways and antioxidant enzyme activity. Activities such as alditol: NADP+ 1-oxidoreductase and heme binding influence antioxidant enzyme activity through metabolic regulation, contributing to maintaining intracellular NADPH levels by catalyzing redox reactions with NADP+. NADPH serves as a critical reducing agent within cells, participating in antioxidant defense functions by supplying reducing equivalents to antioxidant enzymes and enhancing cellular antioxidant capacity. Heme-binding proteins help maintain intracellular redox balance, preventing over-oxidation of heme and the subsequent generation of ROS. Similarly, tetrapyrrole-binding proteins are involved in the metabolism and transport of tetrapyrrole compounds, such as heme and chlorophyll, maintaining their homeostasis and preventing over-oxidation. Nuclear receptor activity, including ligand-activated transcription factor activity, influences transcriptional regulation and gene expression; nuclear receptors play crucial roles in antioxidant stress responses. For instance, peroxisome proliferator-activated receptor γ (PPARγ) activates antioxidant enzyme expression, thus reducing oxidative stress. The transcription factor Nrf2 activates under oxidative stress conditions and binds to antioxidant response elements (ARE) to promote the expression of a series of antioxidant genes, ultimately enhancing cellular antioxidant capacity. KEGG enrichment analysis The KEGG enrichment analysis identified a total of 221 signaling pathways, of which 128 were significantly associated (p < 0.05). The top 20 pathways were selected for visualization (Fig. 8B) and first-level classification (Fig. 8C), which encompassed four major categories: Human Diseases (12), Environmental Information Processing (4), Organismal Systems (3), and Cellular Processes (1). We focused on the three pathway categories excluding Human Diseases, with the VEGF signaling pathway showing the highest significance. VEGF stimulates angiogenesis, enhances blood perfusion and oxygen supply to tissues, and reduces ROS production due to hypoxia, thereby alleviating oxidative stress and providing antioxidant protection to tissues (Byeon et al., 2010). The PI3K pathway converts phosphatidylinositol to PIP3, which recruits and activates Akt, leading to the phosphorylation of various downstream target proteins, such as the FoxO transcription factor. This transcription factor regulates the expression of antioxidant enzymes, including superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx), thus enhancing cellular antioxidant capacity. Additionally, activation of the C-type lectin receptor signaling pathway influences antioxidant enzyme expression, impacting cellular antioxidant capacity (Geijtenbeek & Gringhuis, 2009). The ErbB signaling pathway regulates cell proliferation, differentiation, and survival by activating downstream pathways such as MAPK and PI3K-Akt, which subsequently upregulate antioxidant enzymes and inhibit oxidative stress-induced apoptosis (Yang et al., 2022). Ovarian steroidogenesis involves the synthesis and secretion of various hormones, such as estrogens and progesterones, which are critical for the physiological functions of the female reproductive system and are closely related to oxidative stress resistance. Estrogen can upregulate the expression of antioxidant enzymes, including SOD and CAT, while inhibiting oxidative stress-related inflammatory factors, thus mitigating oxidative damage in ovarian tissues. In multiple tissues and organs, including the cardiovascular and nervous systems, estrogen protects cells from oxidative stress-induced apoptosis, senescence, and dysfunction (d'Anglemont de Tassigny, 2017). Rap1, a member of the Ras superfamily of small GTPases, regulates processes such as cell adhesion, migration, and cytoskeletal reorganization and plays a role in the cellular response to oxidative stress. For example, Rap1 regulates adhesion plaque formation and interactions between cells and the extracellular matrix, enhancing cellular mechanical stability and resistance to oxidative stress-induced damage (Yang et al., 2022). The Relaxin signaling pathway activates downstream signaling cascades, such as cAMP-PKA, via G-protein-coupled receptors, regulating cell proliferation, differentiation, and metabolism. In the vascular system, Relaxin protects endothelial cells from oxidative stress damage, maintains vascular integrity, regulates vascular tone, and improves blood circulation, thereby reducing tissue hypoxia and ROS generation caused by poor perfusion (Ng et al., 2018). In summary, onions form an antioxidant network with the VEGF signaling pathway as the core, collaborating with multiple pathways to regulate the expression of antioxidant enzymes in vivo. This network maintains intracellular redox homeostasis, with each pathway exhibiting distinct antioxidant mechanisms tailored to the specific needs of different tissues. Molecular docking preliminary verification of the antioxidant mechanism The top five metabolites, based on Degree ranking in the “Metabolite-Target” network, were validated via molecular docking against each of the five core targets. In this context, the ligand refers to the metabolite and the receptor to the target protein. The molecular docking binding energies of the active ingredients and their corresponding targets are illustrated in Fig. 8D, where darker squares indicate lower binding energy. The results demonstrated that the binding energies of all tested molecules were below 0 kcal/mol, implying favorable binding characteristics that facilitate the formation of stable binding conformations and enable corresponding biological functions, thereby preliminarily affirming the reliability of the antioxidant mechanism analysis. Molecules exhibiting binding energies less than -5.0 kcal/mol are considered to have strong binding affinity and high biological activity. The binding energies for Myricetin with the five core targets were recorded as -5.02, -5.39, -6.57, -6.91, and -5.40 kcal/mol, all of which are below -5.0 kcal/mol, indicating that Myricetin may play a significant role in the antioxidant mechanism of onions. Molecular docking was employed as a preliminary verification of the antioxidant mechanism through computer simulation, confirming that the analyzed metabolite-target interactions are not only theoretically feasible but also biologically active. The docking approach provides valuable insights into the binding modes, key interacting residues, and binding affinities of metabolite-target complexes, which are essential for understanding the molecular basis of antioxidant activity. The binding energy values obtained from molecular docking reflect the thermodynamic stability of the ligand-receptor complexes, with more negative values indicating stronger and more stable interactions. Myricetin is a flavonoid known for its extensive antioxidant activities; it protects cells from H₂O₂-induced damage by inhibiting the generation of ROS and activating antioxidant enzymes such as SOD, CAT, and GPx (Semwal et al., 2016). Furthermore, Myricetin can prevent oxidative stress-induced apoptosis by modulating the PI3K/Akt and MAPK signaling pathways, thereby safeguarding cells from oxidative damage. Additionally, Myricetin has the capacity to form complexes with divalent metal cations like copper (Cu2+) and iron (Fe2+), which mitigates metal ion-induced oxidative damage. This metal ion chelation helps impede free radical reactions catalyzed by metals, thus exerting antioxidant effects. Notably, Vitamin P and the core target EGFR exhibited the lowest predicted binding energy of -7.66 kcal/mol, suggesting an exceptionally stable interaction that may play a critical role. Further cellular and animal studies are necessary for subsequent validation. It must be admitted that this study has the following limitations. First of all, the samples of this study were derived from different color onion samples cultivated in Aksu area of China. The generalization of the research results to other onion varieties and geographical regions will be further improved in the next study by conducting a wider sample study under the same methodology. Secondly, the non-targeted metabolomics method used in the study depends on the spectral matching with the existing compound database, and the targeted metabolomics quantitative research will be carried out based on the research results in the future, so as to further deepen the research conclusions. Thirdly, the system biology analysis and molecular docking prediction in this study belong to bioinformatics methods, and the proposed compound-target-pathway network represents a prediction model based on bioinformatics integration. The actual in vivo interactions between these pathways may be more complex and may involve additional regulatory mechanisms not covered in this analysis. Therefore, we plan to conduct experimental verification in cells or animal models in the next step, and solve these limitations by integrating multi-omics verification and functional experiments. Conclusion This study systematically investigated the metabolite differences and antioxidant mechanisms of purple, yellow, and white onions using chemometrics, non-targeted metabolomics, and systems biology techniques. A total of 1442, 1314, and 1397 metabolites were identified in PO, YO, and WO), respectively, with 157 metabolites showing significant enrichment (p < 0.05, LDA > 2.00) among the three varieties. The results demonstrated that the in vitro antioxidant activities of the different colored onions were significantly specific, with PO exhibiting superior performance in TPC (1592.44 μg GAE/mL, p < 0.001), TFC (0.42 mg RE/L, p < 0.001), FRAP (7.94 μmol FeSO4/mL, p < 0.001), and ABTS+ radical scavenging ability (92.34%, p < 0.001) compared to YO and WO. In contrast, WO showed the highest DPPH radical scavenging activity (44.93%, p < 0.001) and SASC (55.82%, p < 0.05), indicating its unique lipophilic and superoxide anion scavenging capabilities. Fatty acyls were identified as a potential biomarker group for PO, flavonoids for WO, and carboxylic acids and derivatives for YO. KEGG pathway analysis indicated that the antioxidant traits associated with different colored onions varied in their formation mechanisms: PO primarily relied on flavonoid biosynthesis, WO on flavone and flavonol biosynthesis, and YO on phenylalanine metabolism. Moreover, 29 key antioxidant metabolites were screened through Pearson correlation analysis and systems biology methods. By targeting the core proteins SRC, AKT1, EGFR, MMP9, and TNF, the antioxidant metabolites formed an antioxidant network centered on the VEGF signaling pathway, with multiple pathways collaborating to regulate the expression of antioxidant enzymes in vivo and exert antioxidant effects. Molecular docking validation preliminarily verify the accuracy of the antioxidant mechanism and suggesting that myricetin may be a critical metabolite for the antioxidant effects of onions in vivo. The present study elucidates the metabolic basis and molecular mechanisms underlying the differences in antioxidant activity among various colored onions, providing scientific insights and new perspectives for evaluating onion quality, optimizing high-value utilization, and developing tailored nutritional products. The next step will be to further verify the antioxidant mechanism through cell experiments and animal experiments, and expand the sampling range of each color to enhance universality. CRediT authorship contribution statement Chenglong He: Writing – review & editing, Writing – original draft, Investigation, Formal analysis. Johane Johari Mkunga: Software, Methodology, Conceptualization. Dongsheng Zhang: Validation, Supervision, Project administration, Investigation. Muhammad Iqbal Choudhary: Methodology, Conceptualization. Wenxi Zhai: Methodology. Xinrong Xie: Conceptualization. Liyue Fei: Data curation. Chunhui Shan: Supervision, Software, Data curation. Xinquan Yang: Project administration. Wenchao Cai: Writing – review & editing, Supervision, Project administration, Funding acquisition. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments This work was financially supported by the NGHJG Project of Xinjiang Production and Construction Corps (2023AA503). Data availability Data will be made available on request. References Associated Data Data Availability Statement Data will be made available on request.