Untargeted metabolomics reveals plasma metabolic dysregulation in schizophrenia Authors: Bing Han, Baie Feng, Yanyan Ge, Tongyuan Wang, Li Liu, Peng Sun, Xianzi Chu, Yanfen Zhang, Xinhong Du, Feng Wang, Tianqi Shen Published: 2026 ABSTRACT AbstractObjectiveTo screen significantly differential metabolites between schizophrenia (SZ) patients and healthy controls using univariate analysis combined with orthogonal partial least squares discriminant analysis (OPLS-DA), and to provide experimental evidence for disease-related research.MethodsPlasma samples were collected from 30 SZ patients and 30 healthy controls for untargeted metabolomics analysis. T-test together with OPLS-DA was applied to screen differential metabolites, and pathway enrichment analysis was performed to identify the core metabolic pathways involved.ResultsA total of 1,129 metabolites were identified by untargeted metabolomics. Using the thresholds of t-test P < 0.05 and VIP > 1.0(with false discovery rate < 0.05), 237 significantly differential metabolites were screened out. Three-level chemical taxonomy annotation showed that differential metabolites were mainly enriched in lipids and lipid-like molecules, organic acids and derivatives, and organoheterocyclic compounds (collectively accounting for nearly 70%), with amino acids, peptides and analogues being the core subclass. MetPA and MSEA pathway enrichment analyses revealed that differential metabolites were predominantly enriched in amino acid metabolism (biosynthesis of amino acids, alanine, aspartate and glutamate metabolism, glycine, serine and threonine metabolism), energy metabolism (glycolysis and TCA cycle), and neurotransmitter-related pathways (GABAergic synapse).ConclusionDifferential metabolites in the plasma of SZ patients are mainly concentrated in amino acid metabolism, energy metabolism, and neurotransmitter-related pathways. These findings suggest that SZ is associated with altered plasma metabolites, particularly in amino acid, energy, lipid, and neurotransmitter-related metabolic pathways. Further studies with larger cohorts and independent validation are needed before these metabolites can be considered candidate biomarkers.Trial registrationNot applicable.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12888-026-08441-y.Keywords: Schizophrenia, Metabolomics, Plasma, Biomarkers Introduction Schizophrenia (SZ) is a severe mental disorder with a global prevalence of approximately 1%. Clinically, it is characterized by impairments in cognitive function, emotional experience, and social competence. The disease predominantly affects young and middle-aged adults and tends to follow a recurrent and chronic course, imposing a substantial economic burden on patients, their families, and society. The etiological factors of SZ are complex and multifactorial. Currently, the prevailing pathogenic hypotheses in academia include the neurotransmitter dysregulation hypothesis and the neurodevelopmental disorder hypothesis; however, the exact molecular mechanisms underlying this disease remain incompletely elucidated. Metabolomic studies have confirmed that the disruption of metabolic networks represents a critical component in the pathogenesis of SZ. Even minor fluctuations in metabolite levels may disturb the homeostasis of central neurotransmitters, impair normal neuronal function, and thereby contribute to the pathophysiological progression of the disease [1, 2]. Relevant studies have demonstrated that patients with SZ exhibit abnormal expression of multiple metabolites, and these differentially expressed metabolites are closely associated with cognitive impairment and structural brain abnormalities [3]. A substantial body of evidence indicates that differential metabolites not only directly reflect the metabolic disturbances in patients with SZ but also hold potential as candidate biomarkers for disease diagnosis [4]. Therefore, the identification and in-depth analysis of differential metabolites between patients with SZ and healthy controls are of significant importance for uncovering the molecular mechanisms of the disease, exploring precise diagnostic indicators, and developing targeted therapeutic agents, while also providing novel research perspectives for individualized prevention and treatment of mental disorders. Metabolomic analyses of psychiatric disorders typically generate datasets characterized by high information content, elevated dimensionality, and relatively small sample sizes. Conventional univariate analytical methods struggle to identify biomarkers that can reliably reflect intergroup differences from vast metabolic profiling data, which has emerged as a significant challenge in the current advancement of metabolomics [5]. Orthogonal projections to latent structures discriminant analysis (OPLS-DA), as a novel multivariate statistical technique, offers the core advantage of effectively filtering out data variability in independent variables that is unrelated to the categorical response variable. This process concentrates the classification information primarily within a single predictive component, thereby simplifying model structure, enhancing interpretability, and improving discriminatory performance [4, 6]. The application of OPLS-DA to metabolomic data analysis enables the identification of differential metabolites that contribute substantially to intergroup classification, thereby providing robust support for uncovering critical information regarding disease-associated alterations in metabolic patterns [7, 8]. In this study, untargeted metabolomic profiling of plasma samples from patients with SZ and healthy controls was performed using ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS). On this basis, univariate statistical analysis combined with multivariate pattern recognition methods was employed to identify significantly differential metabolites in the patient group. Leveraging the Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathway database, metabolite set enrichment analysis (MSEA) and pathway topology analysis (MetPA) were conducted to screen for pathways enriched with differential metabolites. By exploring alterations in the metabolic profiling patterns of patients with SZ, this study investigated the molecular mechanisms of SZ from a systems-level perspective, with the aim to provide preliminary metabolomic evidence that may inform future research on the pathophysiology and potential biomarker development for this disease. Materials and methods Study participants This study was approved by the Ethics Committee of the No. 984 hospital of PLA (No. 2025060506; Date: 5 June, 2025). All participants or their legally authorized representatives were fully informed of the study objectives, procedures, and potential risks prior to enrollment, and all provided written informed consent voluntarily. The SZ group comprised 30 patients with SZ recruited from the No. 984 Hospital of PLA. The inclusion criteria were as follows: (1) fulfillment of the diagnostic criteria for SZ according to the International Classification of Diseases, 10th Revision (ICD-10); and (2) age between 18 and 65 years. The control group was matched to the 30 patients with SZ for sex, age, and body mass index (difference < 1 kg/m²) from healthy individuals undergoing physical examination at the same hospital. Psychiatric symptom severity was independently assessed by two physicians from the Third and Fourth Departments of Psychiatry at the No. 984 Hospital of PLA using the Brief Psychiatric Rating Scale (BPRS) and the Positive and Negative Syndrome Scale (PANSS). The exclusion criteria for all participants were as follows: (1) neurological disorders or other psychiatric disorders; (2) history of major physical illness; (3) history of substance (drugs, alcohol, etc.) abuse or dependence; (4) comorbid endocrine disorders such as diabetes mellitus, hypertension, or hyperlipidemia; and (5) pregnancy or lactation. Sample collection and metabolite extraction Fasting venous blood (5 mL) was collected into lithium heparin anticoagulant tubes and centrifuged at 3000 g for 15 min to obtain plasma, which was aliquoted and stored at -80℃. For metabolite extraction, plasma samples were thawed at 4℃, mixed with pre-cooled methanol/acetonitrile/water (2:2:1, v/v/v), vortexed, ultrasonicated at low temperature for 30 min, placed at -20℃ for 10 min, and centrifuged at 14,000 g for 20 min at 4℃. The supernatant was dried under vacuum. Before MS analysis, the residue was reconstituted in 100 µL acetonitrile/water (1:1, v/v), vortexed, and centrifuged at 14,000 g for 15 min at 4℃; the supernatant was used for analysis. UHPLC-MS/MS analysis Chromatographic separation was performed on a Vanquish LC UHPLC system, and mass spectrometry was conducted using an Orbitrap Exploris™ 480 (Thermo) in both positive and negative electrospray ionization (ESI) modes. Detailed chromatographic and mass spectrometric parameters are provided in Supplementary file 1. Samples were randomly assigned to the injection sequence to minimize systematic bias. A pooled quality control (QC) sample was injected after every 10 experimental samples to monitor instrument stability and reproducibility. Two to three blank injections were performed at the beginning and end of each analytical batch to assess potential carryover. All samples were analyzed in a single batch, and no batch correction was applied. The raw data were processed using XCMS 4.4.0 (Scripps, La Jolla, CA, USA) for feature detection, peak alignment, and retention time correction. Data normalization was performed using the support vector regression (SVR) method. Data processing and quality control Raw data were converted to .mzXML format using ProteoWizard, and peak alignment, retention time correction, and peak area extraction were performed with XCMS software. Metabolite identification was achieved by matching retention time, mass accuracy (< 10 ppm), MS/MS spectra, and collision energy against an in-house database (Shanghai Applied Protein Technology); all identifications were at least Level 2. Data preprocessing included removing features with > 50% missing values, KNN imputation, and filtering features with RSD > 50%. QC samples were prepared by pooling equal aliquots of all study samples and injected periodically throughout the analytical run. QC metrics included: total ion current (TIC) overlay, clustering of QC samples in principal component analysis(PCA), Pearson correlation among QC samples, Hotelling’s T2 test, multivariate control chart (MCC), and the proportion of QC features with RSD ≤ 30%. Statistical analysis Statistical analyses were performed in three parts: differential metabolite screening, pathway enrichment analysis, and correlation analysis. Normally distributed continuous variables were presented as mean ± standard deviation (SD), non-normally distributed continuous variables as median (interquartile range), and categorical variables as frequency (percentage). All statistical analyses and graphical visualizations were performed using R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P < 0.05 was considered statistically significant. Differential metabolite screening Principal component analysis (PCA) was first performed as an unsupervised method to visualize overall sample distribution and detect outliers. Subsequently, supervised OPLS-DA was employed to maximize inter-group separation. Model robustness was evaluated by Q² from 7-fold cross-validation (Q² > 0.5 considered robust), and overfitting was tested by permutation test (n = 200). Univariate analysis was performed using independent-samples t-test with false discovery rate (FDR) correction using the Benjamini-Hochberg method; significance was set at P < 0.05 and FDR < 0.05. Differential metabolites were defined by VIP > 1.0, FDR < 0.05, and fold change > 1.5 or < 2/3. Volcano plots and hierarchical clustering heatmaps were generated based on the differential metabolite profiles to visualize group separation and expression patterns. Pathway enrichment analysis Differential metabolites were submitted to the MetaboAnalyst platform for MetPA topology analysis (Pathway Impact) and MSEA enrichment analysis based on the KEGG database. A pathway was considered significantly enriched when FDR < 0.05. Bubble plots and bar plots were generated to visualize the enrichment results. Correlation analysis To explore the relationship between differential metabolites and clinical characteristics, Spearman correlation analyses were performed between metabolite abundance and clinical symptom scores (BPRS and PANSS). Correlation results were expressed as Spearman’s rank correlation coefficient (ρ) and P values. Bar plots were generated to visualize the correlation results. Due to the limited sample size, we were unable to incorporate clinical confounders into multivariate regression models. Results A total of 60 participants were enrolled in this study, including 30 patients with SZ (all were chronic SZ and receiving antipsychotic medication at the time of enrollment) and 30 healthy controls. The mean age of the overall cohort was 46.3 ± 10.9 years, and 38 participants (63.3%) were male. The body mass index (BMI) was 22.29 ± 8.1 kg/m² in the SZ group and 22.87 ± 7.8 kg/m² in the healthy control group. The mean BPRS score in the SZ group was 38.0 ± 6.9 points, and the median PANSS total score was 99 (89, 104) points. The mean chlorpromazine equivalent dose in the SZ group was 500 (300, 900) mg/day. Detailed demographic and clinical characteristics of all participants are provided in Supplementary Table S1. Quality control and experimental stability Six QC metrics were used to assess instrument stability and experimental reproducibility. Overlaid total ion current (TIC) chromatograms of QC samples showed excellent consistency in peak intensity and retention time in both positive and negative ion modes [9]. In PCA score plots, QC samples clustered tightly (Fig. 1-CD), and Pearson correlation coefficients among QC samples were all > 0.9, indicating good repeatability. Hotelling’s T² test placed all QC samples within the 99% confidence interval (Fig. 1-AB). Multivariate control charts (MCC) showed that fluctuations of QC samples were within ± 3 standard deviations. The proportion of QC features with RSD ≤ 30% exceeded 70%. These results demonstrate high instrument stability and reliable data quality for subsequent analysis [10–12] (Supplementary file 2). Assessment of experimental quality control Overview of identified metabolites A total of 1129 metabolites were identified (646 in positive ion mode, 483 in negative ion mode). Three-level chemical taxonomy (SuperClass, Class, SubClass) was applied. At the SuperClass level, the top three categories were lipids and lipid-like molecules (25.6%), organic acids and derivatives (25.07%), and organoheterocyclic compounds (19.57%), together accounting for 70% of total identified metabolites (Fig. 2). At the Class level, carboxylic acids and derivatives (20.19%), fatty acids (10.36%), and glycerophospholipids (7.26%) were most abundant. At the SubClass level, amino acids and peptides (18.25%), fatty acids and conjugates (6.02%), and glycerophospholipids (4.43%) dominated. These results indicate widespread lipid remodeling, amino acid and energy metabolic disturbances in SZ. Overview of chemical taxonomy of identified metabolites PCA analysis Unsupervised PCA was first performed to assess the intrinsic grouping tendency of the samples. As shown in Fig. 1C and D (positive and negative ion modes), a clear separation trend between SZ patients and healthy controls was observed along PC1, suggesting the presence of metabolic differences between the two groups. Subsequently, supervised OPLS-DA was conducted to maximize group discrimination and identify differential metabolites. The model parameters obtained from 7-fold cross-validation were as follows: for positive ion mode, R2Y(cum) = 0.994 and Q2(cum) = 0.983; for negative ion mode, R2Y(cum) = 0.987 and Q2(cum) = 0.969. While these high Q² values suggest good predictive ability in the training set, they should be interpreted with caution given the relatively modest sample size and the lack of an external validation cohort. To exclude the possibility of overfitting, a permutation test (n = 200) was performed, which confirmed that the model was not overfitted (Fig. 3). The Q2 values of the permuted models dropped to negative values with decreasing permutation retention (at full permutation, Q2 = − 0.8663 for positive ion mode and − 0.8121 for negative ion mode), whereas the original Q2 values remained significantly higher. This pattern indicates that the observed discrimination is unlikely to be a chance finding, supporting the robustness of the OPLS-DA models in our exploratory cohort. Validation of the OPLS-DA model Screening of differential metabolites Using thresholds of OPLS-DA VIP > 1.0 and t-test P < 0.05 (with FDR < 0.05), 237 significantly differential metabolites were identified, including 129 in positive ion mode (59 up-regulated, 70 down-regulated) and 108 in negative ion mode (43 up-regulated, 65 down-regulated). In positive ion mode, metabolites such as glutamine, serine, glycine, valine, and lysine (amino acid/neurotransmitter-related metabolites), as well as oxazepam and dihydrothymine (other key metabolites), were significantly altered. In negative ion mode, significant abnormalities were observed in metabolites including citric acid and malate (energy metabolism, TCA cycle-related); serine, glycine, D-aspartic acid, Thr-Glu dipeptide, Ser-Glu dipeptide, Glu-Gly, and N-acetylglutamate (neurotransmitter, amino acid metabolism-related); and uric acid (oxidative stress, inflammation-related). Key TCA cycle intermediates such as citric acid and malate were significantly downregulated, accompanied by abnormalities in serine, glycine, glutamate-related peptides, and uric acid levels. Complete results for all identified metabolites are provided in Supplementary Table S2. A volcano plot illustrated the distribution of differential metabolites (Fig. 4). Compared with healthy controls, a large number of significantly differential metabolites were identified in the plasma of SZ patients, among which both up-regulated (red dots) and down-regulated (blue dots) metabolites accounted for a high proportion, indicating extensive metabolic remodeling under the disease state. Further combined analysis of chemical taxonomy annotation and VIP values from the OPLS-DA model revealed that differential metabolites were mainly enriched in categories such as lipids and lipid-like molecules, organic acids and derivatives, and most metabolites in these categories exhibited high VIP values, making them key differential metabolites for distinguishing between the two groups. The hierarchical clustering heatmap of differential metabolites in positive and negative ion modes (Fig. 5) showed that the metabolite expression profiles of the SZ group and the healthy control group could be completely separated. The two groups formed distinct independent clusters, suggesting a systemic alteration in metabolite expression patterns under the disease state. Metabolites also exhibited clear bidirectional clustering characteristics: some metabolites were significantly up-regulated in the patient group, while others were significantly down-regulated, which corroborated the findings from the volcano plot analysis. This result further confirms the existence of widespread and stable abnormalities in the plasma metabolome of SZ patients. The chemical class distribution of differential metabolites was similar to that of total metabolites, dominated by lipids, organic acids, and heterocyclic compounds. Volcano plots of differential metabolites in positive and negative ion modes Hierarchical clustering heatmaps of differential metabolites in positive and negative ion modes Pathway enrichment analysis MetPA topology analysis As shown in Fig. 6, the differential metabolites were mainly enriched in amino acid metabolism, energy metabolism, and neurotransmitter-related pathways. The bar plot (Fig. 6-A) showed that pathways such as central carbon metabolism in cancer, alanine, aspartate and glutamate metabolism, biosynthesis of amino acids, and GABAergic synapse were statistically significant, and some pathways (e.g., GABAergic synapse) exhibited high enrichment factors, indicating a high concentration of differential metabolites in these pathways. MetPA topology analysis (Fig. 6-B) further revealed that glycine, serine and threonine metabolism pathways were not only significantly enriched but also possessed high topological impact values within the metabolic network, suggesting that disturbances in these pathways may be core links in the metabolic imbalance of SZ patients. MetPA pathway topology analysis MSEA enrichment analysis MSEA identified significantly enriched metabolite sets among differential metabolites. Among the top 25 most significantly enriched metabolic pathways, As shown in Fig. 7, the Ubiquinone and other terpenoid-quinone biosynthesis pathway exhibited the highest enrichment ratio. Meanwhile, pathways such as Glutamatergic synapse, GABAergic synapse, Nitrogen metabolism, and Biotin metabolism also showed very high enrichment ratios and statistical significance (FDR < 1 × 10− 30). Further analysis revealed that these significantly enriched pathways mainly clustered into three modules: neurotransmitter metabolism (glutamatergic, GABAergic synapses), amino acid metabolism (histidine, lysine and nitrogen metabolism), and energy metabolism (ubiquinone biosynthesis, vitamin B6 metabolism). These findings suggest that plasma metabolic changes in SZ may be related to pathways involved in neurotransmitter metabolism, energy metabolism, and oxidative stress. However, plasma data alone cannot directly demonstrate central nervous system neurotransmitter imbalance. MSEA metabolite set enrichment analysis results Correlation analysis between differential metabolites and clinical characteristics To explore the clinical relevance of the identified differential metabolites, we performed Spearman correlation analysis between the abundance of these metabolites and the clinical symptom scores (BPRS and PANSS) of the patients. As shown in Fig. 8, among the top 20 metabolites most strongly correlated with BPRS, inosine showed the strongest positive correlation (ρ = 0.624, FDR = 0.019), while N-Acetyl-L-glutamine showed the strongest negative correlation (ρ = -0.655, FDR = 0.019). In contrast, no metabolite showed a significant correlation with PANSS scores after FDR correction. These results suggest that specific plasma metabolites may serve as potential biomarkers reflecting the severity of clinical symptoms, particularly as assessed by BPRS, in SZ patients. Correlation analysis between differential metabolites and clinical characteristics Discussion In this study, using untargeted metabolomics combined with univariate and multivariate statistical approaches, we systematically characterized the plasma metabolic alterations in patients with SZ. Based on chemical taxonomy annotation, differential metabolite screening, and dual pathway enrichment analyses (MetPA and MSEA), our findings suggest potential multi-layered metabolic remodeling correlated with disease status. Our results indicate extensive metabolic reshuffling in SZ patients, with differential metabolites prominently enriched in lipid metabolism, amino acid metabolism, energy metabolism, and neurotransmitter-related pathways [13]. These findings provide preliminary metabolomic evidence for understanding the pathophysiological mechanisms underlying SZ that may inform future mechanistic investigations. In terms of the chemical taxonomy distribution of differential metabolites, at the SuperClass level, the three most abundant categories were lipids and lipid-like molecules, organic acids and derivatives, and organoheterocyclic compounds, collectively accounting for nearly 70% of all differential metabolites. This observation suggests that the metabolic abnormalities in SZ present a systemic feature involving multiple categories and pathways. When stratifying metabolites by Class and SubClass chemical categories, carboxylic acids and derivatives, fatty acids, glycerophospholipids, amino acids, peptides and analogues were found to be the most prominently perturbed metabolite classes. Consistent with the above observations, amino acid and lipid metabolic shifts appear to be among the most markedly altered metabolic phenotypes detected in subjects with SZ [12, 14]. Importantly, these two classes of metabolic abnormalities are closely related to the synthesis of neurotransmitter precursors and also participate in the maintenance of cell membrane structural integrity, signal transduction, and mitochondrial energy supply [15]. The present data align well with earlier metabolomic investigations, offering partial evidence that shifts in lipid and amino acid metabolism could be involved in SZ-associated pathological mechanisms [16]. Regarding the screening of differential metabolites, a total of 237 significantly differential metabolites were identified in this study. Numerous metabolites with increased or decreased abundance were identified in both positive and negative ion modes. These observations imply that metabolic networks linked to SZ tend to show bidirectional metabolic shifts rather than uniform one-way alterations. Volcano plot and hierarchical clustering heatmap both demonstrated a clear separation of metabolic profiles between patients and healthy controls. Moreover, the differential metabolites consistently exhibited high VIP values, suggesting that they contribute critically to intergroup discrimination and thus may serve as candidate biomarkers [17] that warrant further validation in independent cohorts. In addition, the clustering outcomes support the consistency of metabolite abundance changes between groups, implying that certain characteristic metabolic shifts could be detected in the peripheral blood of subjects diagnosed with SZ. To further elucidate the biological implications of the differential metabolites, we performed dual pathway enrichment analyses using MetPA and MSEA. MetPA topology analysis revealed that differential metabolites were significantly enriched in amino acid metabolism pathways, including biosynthesis of amino acids, alanine, aspartate and glutamate metabolism, and glycine, serine and threonine metabolism [18]. Simultaneously, energy metabolism pathways such as glycolysis and the TCA cycle, as well as pathways involved in neurotransmitter signaling, were also significantly enriched and exhibited high topological impact values [19]. Our observations imply that shifts in amino acid metabolism, disrupted neurotransmitter metabolic cascades, and compromised mitochondrial energy homeostasis could collectively participate in disease-related pathological mechanisms among individuals with SZ [18, 20, 21]. It is well established that glutamate and GABA serve as primary excitatory and inhibitory neurotransmitters of the central nervous system [22]. Perturbations of pathways related to these two transmitters could interfere with normal neural circuit activity and show correlative trends with positive, negative and cognitive symptoms among subjects with SZ [23]. MSEA enrichment analysis further indicated that pathways associated with mitochondrial respiratory chain function (ubiquinone biosynthesis), neurotransmitter signaling (glutamatergic and GABAergic synapses), as well as cofactor metabolism (biotin and vitamin B6 metabolism) presented relatively high enrichment ratios and robust statistical support. These pathways can be primarily categorized into three functional modules: neurotransmitter metabolism, amino acid and nitrogen metabolism, and energy metabolism along with oxidative stress-related metabolism. Among them, ubiquinone (coenzyme Q10) biosynthesis is a key component of the mitochondrial respiratory chain; its dysregulation points to impaired energy metabolism and enhanced oxidative stress [24]. Meanwhile, vitamin B6 metabolism and biotin metabolism, as cofactor-related pathways, may indirectly affect neurotransmitter synthesis and antioxidant capacity [25]. Collectively, these findings indicate that SZ is not merely a disorder of central neurotransmitter imbalance [26], but is also accompanied by systemic disturbances in amino acid, lipid, and energy metabolism. The metabolite changes observed in peripheral blood might only partially hint at pathological events linked to the central nervous system, and could represent a portion of systemic biological shifts that show correlation with SZ [27]. It should be noted that oxazepam, a benzodiazepine derivative detected in our untargeted profiling, was not routinely used in our patient cohort. However, occasional single-dose administration for acute symptom control cannot be completely ruled out, and may have contributed to the detectable plasma levels. Nevertheless, given that our LC-MS/MS analysis was performed without derivatization, the likelihood of analytical artifact is minimal. In contrast, dihydrothymine is a well-characterized endogenous intermediate of thymine catabolism, and its detection reflects normal metabolic turnover rather than exogenous exposure [28]. To explore the clinical relevance of these metabolic alterations, we performed Spearman correlation analyses between differential metabolites and symptom scores. Notably, inosine showed a significant positive correlation with BPRS (rho = 0.624, FDR = 0.019), while N-acetyl-L-glutamine showed a significant negative correlation (rho = − 0.655, FDR = 0.019). Inosine, a purine metabolite with neuroprotective properties, may reflect a compensatory response to oxidative stress and neuroinflammation in SZ [29]. N-acetyl-L-glutamine, closely linked to glutamate metabolism, may indicate impaired glutamine-glutamate cycling in patients with more severe symptoms [30]. No significant correlations were found with PANSS after FDR correction, which may be attributed to differences in the clinical dimensions assessed by the two scales and the modest sample size. These findings suggest that plasma metabolites, particularly inosine and N-acetyl-L-glutamine, may correlate with symptom severity in SZ [31], warranting further validation in larger cohorts. Interestingly, no significant correlations were observed between any metabolite and PANSS scores after FDR correction. This discrepancy may be explained by the distinct clinical dimensions captured by the two instruments: the BPRS is a broader psychiatric rating scale that assesses a wider range of symptoms, including affective and general psychopathology, whereas the PANSS focuses more specifically on positive, negative, and general symptoms with a different weighting structure. Additionally, the absence of significant PANSS correlations may also reflect the relatively modest sample size (n = 30) and the limited variability in PANSS scores within our cohort. Nevertheless, the significant correlations with BPRS highlight the potential utility of plasma metabolomics in reflecting clinically relevant symptom dimensions, particularly general psychopathology and overall illness severity. Taken together, these findings suggest that specific peripheral metabolites, especially inosine and N-acetyl-L-glutamine, may serve as candidate biomarkers for monitoring symptom severity in SZ, although future large-scale studies are warranted to validate these associations. Several limitations should be acknowledged. First, the sample size (n = 30 per group) is modest, limiting statistical power and precluding subgroup analyses. Second, as a single-center study without external validation, the generalizability of our findings requires confirmation in larger, multicenter cohorts. Third, although we applied strict exclusion criteria and recorded multiple confounding factors at patient enrollment, no statistical corrections for these covariates were conducted in our analysis. Accordingly, we cannot fully exclude the confounding effects of antipsychotic medications and lifestyle factors (e.g., diet, smoking) on the plasma metabolome; the cross-sectional design further prevents distinguishing drug-related from disease-intrinsic metabolic changes. Fourth, plasma metabolomics, while reflecting systemic metabolism and capturing metabolites that cross the blood–brain barrier, cannot fully represent regional brain metabolic dynamics. Fifth, as an untargeted screening, our findings are exploratory and hypothesis-generating rather than conclusive. Finally, some metabolites were identified at Level 2 (spectral similarity) rather than Level 1 (authentic standards); targeted validation using reference standards is therefore needed to confirm key candidates. Overall, measurable metabolic variations in plasma were detected in subjects diagnosed with SZ. Lipids, lipid-like compounds and amino acids constituted the majority of differential metabolites, which primarily altered signaling cascades of amino acid metabolism, energy metabolism and neurotransmitter function. This study documents plasma metabolic changes that show correlation with SZ, and these observational results could serve as preliminary supporting data for follow-up mechanistic investigations and biomarker validation work for this disease. Conclusion Untargeted metabolomic profiling in the present study detects extensive plasma metabolic shifts in subjects with SZ. Pathways involving amino acid metabolism, energy metabolism and neurotransmitter function display prominent enrichment of altered metabolites. These observational findings may supply limited reference information for follow-up mechanistic studies and preliminary biomarker screening efforts. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Acknowledgements The authors extend their sincere gratitude to all patients with schizophrenia and healthy volunteers who participated in this study. Their understanding, cooperation, and willingness to contribute to research are greatly appreciated. Author contributions Bing Han, Baie Feng and Yanyan Ge: Conceptualization, Methodology, Software, Validation, Project administration, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Writing - Review & Editing.Tongyuan Wang, Li Liu and Peng Sun: Software, Formal analysis, Visualization, Writing - Original Draft.Xianzi Chu and Yanfen Zhang: Investigation, Data Curation, Writing - Original Draft.Feng Wang and Xinhong Du: Conceptualization, Methodology, Software, Writing - Review & Editing.Tianqi Shen: Conceptualization, Methodology, Writing - Original Draft, Writing - Review & Editing. Funding This study was supported by Intramural Cultivation Grant of the No. 984 hospital of the PLA(202506-03) and Joint Logistics Medical High-Quality Specialty Program. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Researchers who wish to access the data are required to submit a request to the corresponding author, detailing the purpose and intended use of the data. The corresponding author will evaluate each request and determine whether to grant access to the data. Declarations Ethics approval and consent to participate This study was conducted in accordance with the Declaration of Helsinki, approved by the Ethics Committee of the No. 984 Hospital of the PLA (approval number: 2025060506; date: June 5, 2025). All participants or their legally authorized representatives were fully informed of the study objectives, procedures, and potential risks prior to enrollment, and all provided written informed consent voluntarily. Consent for publication 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. Bing Han, Baie Feng and Yanyan Ge contributed equally to this work. Contributor Information Xinhong Du, Email: anaesthe984@aliyun.com. Feng Wang, Email: 14715461@qq.com. Tianqi Shen, Email: shenanea984@163.com. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. Researchers who wish to access the data are required to submit a request to the corresponding author, detailing the purpose and intended use of the data. The corresponding author will evaluate each request and determine whether to grant access to the data.