Untargeted serum metabolomic profiling in patients with generalized and ocular myasthenia gravis Authors: Bin Wei, Haoyu Yuan, Jiawei He, Yuxiang Hu, Jing Li, Xiaorong Wu Published: 2026 ABSTRACT AbstractBackgroundMyasthenia Gravis (MG) is divided into ocular (OMG) and generalized (GMG) subtypes. While clinical diagnosis is well-established, understanding the underlying biochemical mechanisms and metabolic shifts during disease progression remains challenging; untargeted metabolomics offers a novel perspective to explore these systemic alterations.ObjectiveTo characterize the serum metabolic landscape of MG patients and identify potential metabolic signatures associated with disease subtypes (OMG and GMG) via untargeted metabolomics.Methods91 participants (41 GMG, 22 OMG, 28 healthy controls [HC]) were enrolled. Fasting serum samples were analyzed by LC-MS/MS. Multivariate analyses (PCA, PLS-DA/OPLS-DA), differential metabolite screening (VIP > 1.0, p < 0.05), and KEGG pathway enrichment were performed.ResultsHC and MG groups showed distinct metabolic profiles. MG had 515 (175 up, 340 down) and 368 (146 up, 222 down) differential metabolites in positive/negative ion modes, respectively. Key perturbed pathways included glycerophospholipid, sphingolipid metabolism, and unsaturated fatty acid biosynthesis. Ten representative metabolites (e.g., ubiquinone, cortisol) differed significantly among groups; clustering analysis revealed distinct metabolite abundance trajectories across HC, OMG, and GMG.ConclusionMG is associated with notable systemic metabolic dysregulation, particularly in lipid-related pathways. Rather than serving as immediate diagnostic tools, these integrative metabolic signatures provide a crucial biochemical framework for understanding disease pathogenesis and offer valuable clues for future hypothesis-driven research and prospective validation.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12886-026-04843-4.Keywords: Myasthenia Gravis (MG), Untargeted metabolomics, Metabolic signatures, Lipid metabolism Introduction Myasthenia Gravis (MG) is a multifaceted autoimmune condition defined by disrupted neuromuscular junction activity, manifesting clinically as muscle weakness and increased fatigability [1]. MG is classified based on the distribution of affected muscles into ocular MG (OMG) and generalized MG (GMG) [2]. OMG predominantly affects extraocular muscles, resulting in symptoms such as ptosis and diplopia; however, a significant subset of these patients advances to GMG, where other skeletal muscles are involved, potentially leading to more pronounced systemic symptoms and increasing the risk for life-threatening complications [3]. Currently, MG diagnosis and management rely primarily on evaluations of clinical presentations, antibody assessments, immunosuppressant therapy, and thymectomy [4]. However, MG exhibits a variable clinical course, and reliable biomarkers for predicting progression, monitoring therapy, or distinguishing between OMG and GMG are still lacking. Metabolomics offers a holistic systems biology tool for deciphering disease mechanisms and identifying novel biomarkers by quantitatively and qualitatively analyzing small-molecule metabolites within biological samples—the metabolome [5]. Serum untargeted metabolomics is exceptionally suited to investigate complex disorders, as it facilitates simultaneous detection of myriad unknown metabolites, thus enabling comprehensive mapping of metabolic disturbances associated with disease [6]. Such methodologies have been utilized in autoimmune disease research to uncover metabolomic patterns linked to disease activity, prognosis, and responses to therapeutic intervention [7]. For instance, extensive metabolomics analyses in asthma have led to the identification and validation of 17 steroid metabolites significantly reduced among patients [8]. Although MG etiology involves immune cell functional and compositional abnormalities [9], use of untargeted serum metabolomics in MG remains a relatively emerging field but demonstrates substantial promise. Initial investigations have begun to characterize metabolomic profiles in MG serological samples to discover novel biomarkers and monitor disease status [10, 11]. A major focus of these studies is to clarify the metabolic peculiarities of MG and pinpoint specific metabolites or metabolic pathways relevant for diagnosis, disease severity, progression, and therapeutic response—particularly with immunosuppressants such as corticosteroids or biological agents like rituximab [12]. This project employs untargeted serum metabolomics to systematically characterize metabolic traits distinguishing OMG from GMG. The goal is to discover biomarkers specific to each MG subtype, thus advancing early diagnosis, improving risk assessment for progression from OMG to GMG, and aiding development of precision therapeutic approaches. By juxtaposing the metabolomic profiles of OMG and GMG, this study seeks to uncover major mechanisms of disease evolution and provide objective evidence for accurate subtype discrimination in clinical contexts. Methods Subjects and serum samples Patients were recruited from the ophthalmology and neurology outpatient clinics and inpatient departments of the First Affiliated Hospital of Nanchang University and Xiangya Hospital of Central South University. The study population was divided into three groups: 41 individuals diagnosed with GMG, 22 with OMG, and 28 Healthy Controls (HC). All enrolled patients strictly met the established clinical diagnostic criteria for Myasthenia Gravis. For seronegative patients (negative for both AChR and MuSK antibodies), the diagnosis was based on: (1) typical fluctuating muscle weakness and fatigability; (2) a positive neostigmine test (quantitative score improvement ≥ 1) or abnormal repetitive nerve stimulation (RNS) demonstrating a > 10% decrement; and (3) rigorous exclusion of other differential diagnoses. Furthermore, all seronegative patients were followed for at least 12 months to confirm the persistence of the MG phenotype without subsequent antibody conversion. Written informed consent was obtained from all participants. Blood samples were collected during morning hours following overnight fasting, processed by centrifugation, and the separated serum aliquots were stored at − 80 °C until further analysis. Instrumentation and conditions Metabolomic analysis was conducted with a Liquid Chromatography–Mass Spectrometry/Mass Spectrometry (LC-MS/MS) system. The liquid chromatography setup included components such as a binary Nexera XR LC-20AD pump (Shimadzu, Kyoto, Japan) and a Nexera XR SIL-20AC autosampler (Shimadzu, Kyoto, Japan). Chromatographic separation was carried out using an ACQUITY UPLC HSS T3 column (1.8 μm, 2.1 × 100 mm). The mobile phase utilized for LC separation consisted of solvent A (water with 0.1% formic acid) and solvent B (acetonitrile with 0.1% formic acid). The gradient elution protocol was as follows: 0.01 min, 5% B; 2 min, 5% B; 4 min, 30% B; 8 min, 50% B; 10 min, 80% B; 14 min, 100% B; 15 min, 100% B; 15.1 min, 5% B; and 16 min, 5% B. The flow rate was set to 0.35 mL/min and the column temperature maintained at 45 °C. Injection volume was standardized at 3 µL. Mass spectrometric detection was performed using a QTRAP 5500 LC-MS/MS system (Sciex), operated in positive and/or negative electrospray ionization (ESI) mode. Key instrumental parameters included spray voltages of 3800 V (+) and 3200 V (–), capillary temperature of 320 °C, sheath gas flow at 35 arbitrary units, auxiliary gas flow at 8 arbitrary units, and an S lens RF level of 50. The system scanned a mass range between 100 m/z and 1200 m/z, with a primary scan resolution of 70,000 and secondary scan resolution of 17,500, utilizing collision energies of 10, 20, and 40 eV. Metabolite extraction Serum samples (100 µL) were transferred to Eppendorf tubes and resuspended with pre-chilled 80% methanol by vigorous vortexing. The samples were incubated on ice for 5 min, then centrifuged at 15,000 g and 4 °C for 20 min. A portion of the supernatant was diluted to a final concentration of 53% methanol with LC-MS grade water. Samples were placed in fresh tubes and centrifuged again at 15,000 g and 4 °C for 20 min. The resulting supernatant was used for LC-MS/MS injection and subsequent analysis. LC-MS/MS analysis Prepared serum samples were introduced into the LC-MS/MS system. Rigorous quality control procedures were enacted to ensure analytical reliability and data reproducibility, including preparation of pooled quality control (QC) samples by mixing equal aliquots of all study samples. QC samples were routinely injected to monitor signal intensity and retention time stability. Analytical variance was assessed using NIST plasma reference material. Blank solvent injections were performed after every ten samples, both before and following pooled QC sample injection, to detect and minimize sample carryover. Data processing and metabolite identification Raw data from the LC-MS/MS platform were processed using MultiQuant 3.0.3 software (Sciex), facilitating peak alignment, detection, and quantification of metabolites. Metabolite identification utilized mass-to-charge (m/z) ratios and retention times, in conjunction with established databases such as the LUG and HMDB. Quantitative data arrays were normalized, beginning with internal standard calibration—e.g., if L-2-chlorophenylalanine served as the internal standard, signal intensities for each detected peak within samples were segmented and normalized accordingly. Redundant data points and peaks were then merged to construct a comprehensive data matrix. Metabolites exhibiting a coefficient of variation (CV) above 30% for relative peak areas in QC samples were excluded. Data analysis The resultant data matrix was analyzed in the R programming environment for advanced statistical assessment. Principal Components Analysis (PCA) was performed to visualize sample distribution and assess analytical stability. For group discrimination, Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) and PLS-DA were utilized. Model validity was verified via seven-fold cross-validation and a 200-response permutation test to reduce overfitting risk. Variable importance for group differentiation was established using variable importance in projection (VIP) values derived from the OPLS-DA model. Group differences in metabolites were further verified via two-tailed Student’s t-tests. Metabolites with VIP > 1.0 and p-value < 0.05 were considered statistically significant and selected for differential analysis. Visualization of differential metabolite abundance was achieved through volcano plots using ggplot2 in R, relying on log2(FoldChange) and –log10(p-value) statistics. For clustering heat maps, z-score normalization was applied to differential metabolite intensity, with subsequent plotting using the Pheatmap package in R. Pearson correlation analysis was performed using the cor() function, and statistical significance was estimated by cor.mtest(). Significance was defined as p-value < 0.05, and correlation plots were rendered via the corrplot package. Functional annotation and pathway enrichment of differential metabolites were implemented using the KEGG database (https://www.genome.jp/kegg/pathway.html). Significant pathway enrichment was defined when the ratio x/n exceeded y/N and the pathway p-value was < 0.05, indicating meaningful involvement of the corresponding metabolic pathway. Results Clinical characteristics The study enrolled a total of 91 participants: 41 GMG patients and 22 OMG patients and 28 controls. Table 1 displays the basic information, MG classification, and antibodies status of the patients. Table 1Characteristics of MG patients and controlscharacteristicHC(n = 28)OMG(n = 22)GMG(n = 41)P-valueAge (years)51.21 ± 17.2547.14 ± 18.5450.49 ± 14.400.651Male (%)42.8636.3651.220.508Type (Early onset/Late onset)N/A11/1120/21N/AWith thymoma (Yes/No)N/A2/208/33N/AAntibodies (P/N)*N/A7/1536/5N/AReceived immunotherapy (Yes/No)N/A8/1427/14N/A* Among the antibody-positive MG patients, only 2 GMG patients tested positive for muscle-specific tyrosine kinase (MuSK) antibodies, while the remaining patients were positive for acetylcholine receptor (AChR) antibodiesGMG = generalized myasthenia gravis; OMG = ocular myasthenia gravis; HC = healthy control; P/N = positive/negative; N/A not applicable Characteristics of MG patients and controls * Among the antibody-positive MG patients, only 2 GMG patients tested positive for muscle-specific tyrosine kinase (MuSK) antibodies, while the remaining patients were positive for acetylcholine receptor (AChR) antibodies GMG = generalized myasthenia gravis; OMG = ocular myasthenia gravis; HC = healthy control; P/N = positive/negative; N/A not applicable Multivariate analysis of metabolite profiles between HC and MG groups To comprehensively map the metabolic alterations, a two-step exploratory analytical strategy was employed. First, to identify the overarching metabolic changes associated with the disease state, all MG patients (OMG and GMG combined) were compared to healthy controls (HC). Subsequently, subtype-specific differences were explored by separating the cohort into three groups (HC, OMG, and GMG). Figure 1A (positive ion mode) demonstrates a clear segregation between HC and MG groups on the PCA score plot, with principal component 1 (PC1) explaining 53.43% of total variance, supporting distinct metabolic states across groups. Similarly, in negative ion mode (Fig. 1B), PCA distinguishes groups effectively, with PC1 accounting for 46.49% of variance and PC2 for 17.25%. To substantiate these separations, PLS-DA was performed, yielding robust discrimination between groups. The validity of both PLS-DA and OPLS-DA models was confirmed through permutation testing (Fig. 1C for POS; 1D for NEG), with Q2 values of permutated models considerably lower than those of the original, reinforcing model reliability and excluding overfitting. PLS-DA intercepts were R2 = 0.19, Q2 = − 0.32 (Fig. 1C, POS), whereas OPLS-DA intercepts were R2 = 0.23, Q2 = − 0.33 (Fig. 1D, NEG). Multivariate analysis of metabolite profiles between HC and MG groups. (A, B) Principal component analysis (PCA) score plots showing the distribution and separation of HC and MG samples based on metabolic profiles. Panel A displays the results for positive ion mode (POS); panel B for negative ion mode (NEG). Ellipses represent the 95% confidence intervals for each group. (C, D) Permutation test results for the PLS-DA and OPLS-DA models, demonstrating the robustness and validity of the discriminant analyses. Panel C shows results for POS mode, while panel D shows results for NEG mode. Blue circles indicate R2 values, and red squares indicate Q2 values. HC = healthy control; MG = myasthenia gravis Differential metabolite analysis between HC and MG patients In positive ion mode (Fig. 2A), analysis identified 175 metabolites with increased abundance and 340 with decreased abundance in MG patients compared to HC, with 466 metabolites showing no significant alteration. Negative ion mode (Fig. 2B) revealed 146 upregulated and 222 downregulated metabolites, and 365 with no significant difference. The twenty most upregulated and the twenty most downregulated metabolites, ranked by log2(Fold Change) and VIP value, are presented in bar charts for POS (Fig. 2C) and NEG (Fig. 2D) modes. The complete lists of differential metabolites and the detailed KEGG pathway enrichment data for both positive and negative ion modes are provided in Supplementary Tables S1 and S2, respectively. Differential metabolite analysis between HC and MG patients. Volcano plots in positive ion mode (POS, panel A) and negative ion mode (NEG, panel B) illustrate changes in plasma metabolite levels between the HC and MG groups. Red dots represent significantly upregulated metabolites, blue dots indicate significantly downregulated metabolites, and grey dots correspond to metabolites without significant changes. Bar plots in POS (panel C) and NEG (panel D) modes highlight the top 20 upregulated (red) and top 20 downregulated (blue) metabolites, respectively. HC = healthy control; MG = myasthenia gravis; UP = upregulated; DW = downregulated Pathway enrichment and correlation network analysis of differential metabolites between HC and MG patients Figure 3A and B depict correlation network analyses for significantly altered metabolic pathways in positive and negative ion modes. Each node indicates a metabolic pathway, with node size proportional to –log10(p-value) for pathway enrichment. Connections represent overlapping or regulatory relationships among pathways, revealing the intricate interactions within the metabolic landscape. Pathway enrichment identified several perturbed pathways (Fig. 3C and D). In positive ion mode (Fig. 3C), glycerophospholipid metabolism, choline metabolism in cancer, and sphingolipid metabolism emerged as the most enriched, with glycerophospholipid metabolism displaying the highest quantity of altered metabolites. Negative ion mode (Fig. 3D) highlighted enrichment in biosynthesis of unsaturated fatty acids and fatty acid biosynthesis. The number of participating differential metabolites is reflected by dot size, while enrichment significance is indicated by the color intensity of each dot (–log10(P-value)). Pathway enrichment and correlation network analysis of differential metabolites between HC and MG patients. (A, B) Correlation networks of the top significantly altered metabolic pathways in positive ion mode (A) and negative ion mode (B). Each node represents a metabolic pathway, and the size of the node indicates the –log10 (P-value) of the enrichment. Red lines between nodes illustrate correlations among pathways, and node color distinguishes pathway classes. (C, D) Bubble plots showing the top enriched metabolic pathways in positive (C) and negative (D) ion modes between HC and MG. The x-axis represents the ratio of differential metabolites involved in each pathway. Dot size indicates the number of differential metabolites, and color represents the enrichment significance (–log10(P-value)). HC = healthy control; MG = myasthenia gravis K-means clustering analysis of metabolite expression profiles among HC, OMG, and GMG patients Figure 4A demonstrates eight distinct metabolite clusters characterized by unique expression trajectories across HC, OMG, and GMG groups. Several clusters (1, 2, 3, 5, 7, and 8) exhibited notable decreases in metabolite abundance from HC to OMG, with modest shifts from OMG to GMG. In contrast, cluster 4 displayed peak expression in OMG followed by reduction in GMG, whereas cluster 6 showed increased levels in OMG stabilizing in GMG. These patterns suggest that specific groups of metabolites are associated with various stages of disease progression. In an alternative clustering strategy, three larger clusters were delineated (Fig. 4B), which mirrored previous results: clusters 1 and 2 demonstrated marked reduction from HC to OMG and slight increases toward GMG, while cluster 3 exhibited upregulation peaking in OMG. The comprehensive lists of differential metabolites among the three subgroups, along with the detailed K-means clustering numerical data, are available in Supplementary Tables S3–S6. K-means clustering analysis of metabolite expression profiles among HC, OMG, and GMG patients. (A) Eight clusters showing distinct patterns in standardized metabolite expression across HC, OMG, and GMG groups. Each subplot represents a cluster with red lines denoting individual metabolite expression profiles. (B) Three main clusters derived from an alternative K-means clustering, illustrating common trends in metabolite expression among the three groups. Each red line represents a single metabolite within the cluster. HC = healthy control, OMG = ocular myasthenia gravis, GMG = generalized myasthenia gravis Relative abundance of selected differential metabolites among GMG, OMG, and HC groups Figure 5 shows boxplots of ten representative differential metabolites among GMG, OMG, and HC groups as identified by LC-MS/MS in both positive (POS) and negative (NEG) ion modes. In POS mode (A–E), metabolites including Ubiquinone (Q2), SM(d18:0/24:1(15Z)), cortisol, oleic acid, and acetylcarnitine displayed statistically significant differences (P < 0.05) between groups. Similarly, in NEG mode (F–J), azelaic acid, threonic acid, 2-hydroxybutyric acid, suberic acid, and 13-hydroxymyristic acid also demonstrated significant group differences. These metabolites were notably altered not only between the MG subtypes and healthy controls, but also in pairwise MG subtype comparisons. Relative abundance of selected differential metabolites among GMG, OMG, and HC groups. (A-E) Representative differential metabolites detected in positive ion mode (POS): (A) Ubiquinone (Q2), (B) SM(d18:0/24:1(15Z)), (C) CORTISOL, (D) Oleic acid, (E) Acetylcarnitine. (F–J) Representative differential metabolites in negative ion mode (NEG): (F) Azelaic acid, (G) Threonic acid, (H) 2-Hydroxybutyric acid, (I) Suberic acid, (J) 13-Hydroxymyristic acid. Data are presented as boxplots (median, interquartile range, min/max), with each dot representing an individual sample. Significant differences between groups (P < 0.05) are indicated by different letters above the boxes (a, b, c). GMG = generalized myasthenia gravis; OMG = ocular myasthenia gravis; HC = healthy control Discussion This study utilized serum metabolomics analysis of 91 subjects—including GMG and OMG patients along with healthy controls—to uncover distinctive metabolic patterns that separate MG patients from controls. In the positive ion mode, 175 metabolites were found upregulated and 340 downregulated in MG, whereas in the negative ion mode, 146 metabolites were upregulated and 222 downregulated. KEGG pathway enrichment underscored significant perturbations in glycerophospholipid metabolism, sphingolipid metabolism, and unsaturated fatty acid biosynthesis. Clustering analysis delineated groups of metabolites with specific trends progressing from healthy controls to OMG and GMG. Ten key metabolites, such as ubiquinone (Q2), sphingomyelin SM(d18:0/24:1(15Z)), cortisol, and oleic acid, exhibited pronounced differences both across subgroups and between OMG and GMG. These results characterize a distinctive metabolic fingerprint for MG and point to novel metabolic targets that may further our understanding of disease pathogenesis and subtype differentiation. Glycerophospholipids constitute fundamental components of cellular membranes and play critical roles in cellular signal transduction, membrane fusion, and apoptosis [13]. The present findings demonstrate disturbed glycerophospholipid metabolism in MG patients, consistent with prior reports of metabolic disarray in disease states. For instance, untargeted metabolomics and lipidomics analyses of acute leukemia plasma have identified similar disruptions, including altered glycerophospholipid metabolism [14]. Furthermore, research in Sjögren’s syndrome using untargeted serum metabolomics revealed comparable dysregulation [15] Such cross-condition observations suggest that disordered glycerophospholipid metabolism may be a common pathology, with implications for MG pathogenesis through influences on immune cell membrane integrity or cell signaling, thereby potentially aggravating autoimmune activity. Additional literature highlights dramatic changes in glycerophospholipid metabolism under diverse physiological and pathological contexts, such as weaning piglets [16], and ovarian cyst formation [17]. Sphingolipids represent another vital lipid class within cell membranes, contributing to processes like cell growth, differentiation, apoptosis, and inflammation [18]. Significant impairment of sphingolipid metabolic pathways was observed in MG patients in this study. This is in line with research documenting sphingolipid metabolic anomalies in cardiovascular disease [19], inflammatory diseases [20], and neurodegenerative diseases [21]. Notably, plasma metabolomics of atherosclerosis progression in apolipoprotein E-deficient mice revealed changes in both glycerophospholipid and sphingolipid metabolism [19]. In rat models of Alzheimer’s disease, neuro acid demonstrated therapeutic effect by modulating sphingolipid pathways [21]. Evidence from atopic dermatitis studies also indicates sphingolipid pathway involvement in disease development [20]. Given their essential role in immune regulation, sphingolipid metabolic imbalance may alter the behavior of autoimmune cells, thereby fostering MG onset and progression. Unsaturated fatty acids are essential for the maintenance of cell membrane fluidity, engagement in inflammatory processes, and functioning as signaling mediators [22]. The current work identified disrupted biosynthesis of unsaturated fatty acids in MG patients, concordant with previous investigations in diverse pathologies. For example, serum and liver metabolomics in iron-deficiency anemic rats revealed affected unsaturated fatty acid metabolism [23]. Disturbances in unsaturated fatty acid biosynthesis have also been reported in colorectal cancer [24]. Considering their relationship to inflammation and immunomodulation, imbalance in unsaturated fatty acid metabolism may underlie abnormal inflammatory signaling, thereby impacting MG pathology. As further evidence, dandelion extract has been shown to suppress triple-negative breast cancer cell growth by affecting glycerophospholipid and unsaturated fatty acid metabolism [25]. In this study, significant differences were observed in key metabolites—ubiquinone (Q2), sphingomyelin SM(d18:0/24:1(15Z)), cortisol, and oleic acid—across OMG, GMG, and HC groups, suggesting their potential involvement in the metabolic dysregulation underlying MG and offering clues for clinical subtype differentiation. Ubiquinone (Q2), analogous to coenzyme Q10, functions in mitochondrial electron transport and antioxidative defense, thereby providing an index of cellular oxidative status [26]. Since mitochondrial dysfunction and oxidative stress are crucial in MG pathogenesis, altered ubiquinone levels may indicate underlying disruptions in energy metabolism. Sphingomyelin SM(d18:0/24:1(15Z)), a critical membrane lipid prevalent in neural tissues, facilitates membrane integrity and cell signaling; its altered abundance may reflect structural or functional perturbations of neuromuscular junctions or immune cell membranes, aligning with MG pathology [27]. Cortisol, a chief glucocorticoid hormone regulating stress and immune responses, exhibits variation in MG possibly due to immune dysregulation or influence of glucocorticoid therapy, thus supporting its role as a biomarker [8]. Oleic acid, a monounsaturated fatty acid involved in systemic lipid regulation, revealed changes indicative of lipid metabolism and inflammation in MG, consistent with findings in related disorders [28]. Collectively, these metabolites encompass mitochondrial, membrane, immune, and lipid metabolic impairments implicated in MG, providing a biochemical rationale for future hypothesis-driven research. Larger validation studies and integrative multi-omics will be essential for elucidating their function and potential as therapeutic targets in MG. Notably, the clustering analysis revealed intriguing U-shaped or V-shaped expression trajectories (e.g., Cluster 4 and Cluster 6), where certain metabolite levels peaked or dipped in the OMG phase and reversed or stabilized in the GMG phase. One possible biological explanation is that these transient alterations, particularly in specific lipids, might reflect early compensatory responses to local inflammation at the neuromuscular junction during the OMG phase, which are subsequently exhausted or overwhelmed as the disease generalizes into GMG. Alternatively, these patterns might be partially driven by differences in medical management, as GMG patients are more frequently prescribed higher doses of systemic corticosteroids or other immunosuppressants, which can significantly alter metabolic profiles. Without longitudinal data, these hypotheses remain speculative and warrant future prospective investigation. Several limitations of this study should be acknowledged. First, as a retrospective study utilizing historical serum samples, we were unable to retrieve precise clinical scores (e.g., exact QMG or MG-ADL scores on the specific day of blood draw) or individualized, detailed records of disease duration and cumulative immunosuppressive therapy. Consequently, we could not statistically adjust for these critical confounders, and the observed metabolic alterations may partially reflect medication effects rather than pure disease pathology. Second, the sample size of the seronegative subgroups (15 OMG and 5 GMG) was too small to conduct a statistically robust comparison between seropositive and seronegative patients, leaving this important biological question for future studies. Third, the two-step exploratory analytical strategy was not subjected to further statistical adjustments for multiple comparisons. In summary, this research utilized untargeted serum metabolomics to delineate distinctive metabolic alterations in OMG, GMG, and HC. Disruption of pathways, including glycerophospholipid metabolism, sphingolipid metabolism, and unsaturated fatty acid biosynthesis, underscores lipid metabolism imbalance as a fundamental aspect of MG pathogenesis, influencing cell membrane integrity, neuromuscular junction functionality, and immune regulation. Key metabolites such as ubiquinone (Q2), sphingomyelin SM(d18:0/24:1(15Z)), cortisol, and oleic acid distinctly differed between study subgroups, reinforcing their relevance to the pathophysiological processes of specific MG subtypes. Ubiquinone serves as a marker of mitochondrial dysfunction and oxidative stress, sphingomyelin indicates membrane and cell signaling alterations, cortisol relates to immunological and stress pathways, and oleic acid implicates lipid and inflammatory imbalance. These integrative metabolic signatures provide a biochemical framework for future hypothesis-driven research. While they expand our understanding of the underlying disease mechanisms, further prospective validation is required to determine how these insights might eventually translate into optimized patient management. The results highlight the utility of metabolomics in elucidating the biochemical complexity of MG and establish a foundation for subsequent precision medicine research. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Acknowledgements The authors thank all the participant for their contribution in the study. Author contributions B.W., H.Y., and J.H. designed the study, collected data, performed data analysis, and interpreted the results. Y.H. assisted with statistical analysis and figure visualization and drafted the initial manuscript. X.W. and J.L. supervised the entire study, provided critical revisions, and finalized the manuscript. All authors reviewed and approved the final version of the manuscript. Funding This study received support from the National Natural Science Foundation of China (Grant No. 82160207; Grant No. 82371412), the Technology Plan of Jiangxi Provincial Health and Health Commission (202130156), the Young Scholar Project of the First Affiliated Hospital of Nanchang University (YFYPY202219), Jiangxi Science Education Society (2025KXJYS354), the Science and Key Projects of Jiangxi Youth Science Fund (No. 20202ACBL216008), and the leading clinical research project under the Jiangxi Province 2024 “Changyi Leading Research” Program (Grant No. YK14), entitled “Preliminary study on vitamin D-regulated regulatory T-cell subsets and related cytokines in ocular myasthenia gravis. Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate The protocol of this study was approved by the Ethical Committee of The First Affiliated Hospital of Nanchang University ((2023)CDYFYYLK(01–014)) and was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent to participate was obtained from all participants prior to inclusion in the study. Consent for publication Informed, Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Bin Wei and Haoyu Yuan contributed equally to this work. Contributor Information Jing Li, Email: lijing.neurology@aliyun.com. Xiaorong Wu, Email: wxr98021@126.com. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Data Availability Statement No datasets were generated or analysed during the current study.