Skip to main content
Download PDF

Multi-omic serum analysis reveals ferroptosis pathways and diagnostic molecular signatures associated with Moyamoya diseases

Journal of Neuroinflammation volume 22, Article number: 123 (2025) Cite this article

Abstract

Moyamoya disease (MMD) is a rare cerebrovascular disease in humans. Although early revascularization can improve symptoms, it cannot reverse the progression of the disease. The current diagnosis still relies on traditional a Digital Subtraction Angiography (DSA) examination, which is invasive and expensive, leading to delayed diagnosis and affecting treatment timing and patient prognosis. The ability to diagnose MMD early and develop personalized treatment plans can significantly improve the prognosis of patients. Here, we have introduced the research on MMD biomarkers. By integrating proteomics and metabolomics data, we have successfully identified over 1700 features from more than 60 serum samples collected at the onset of symptoms in MMD patients. We use multiple computational strategies to interpret complex information in serum, providing a comprehensive perspective for early diagnosis of MMD. Diagnostic ability of our biomarker is significantly better than previous studies, especially when used in combination. In the study of molecular mechanisms, we found that the ferroptosis pathway was significant disruption in MMD patients, which was also confirmed by transcriptomics data. Finally, we validated the metabolites and proteins associated with ferroptosis pathways, as well as the biomarkers screened by machine learning, using another independent MMD cohort. Our research provides important clues for the diagnosis of MMD, and this assay can identify MMD early, thereby promoting stronger monitoring and intervention.

Graphical abstract

Introduction

The global burden of Moyamoya disease (MMD) demonstrates a marked upward trajectory, particularly within Asian populations. Serial national epidemiological surveys in Japan revealed a near doubling of diagnosed cases from 3,900 (95% CI: 3,500-4,400) in 1994 to 7,700 (95% CI: 6,300-9,300) in 2003, accompanied by rising prevalence (3.16 to 6.03 per 100,000 population) and incidence rates (0.35 to 0.54 per 100,000 population) [1, 2]. This progressive cerebrovascular disorder exhibits striking clinical heterogeneity: while approximately 70% of patients achieve favorable outcomes through timely surgical revascularization (direct/indirect/combined approaches guided by digital subtraction angiography [DSA] and perfusion imaging) [3], 20–30% experience rapid disease progression culminating in catastrophic infarction or hemorrhage despite maximal intervention [4]. Such unpredictable clinical trajectories underscore the urgent need for deeper pathophysiological insights.

The advancement of MMD research faces dual constraints: (1) Limited availability of pathophysiologically relevant animal models, and (2) Practical challenges in obtaining human vascular specimens for mechanistic studies [5]. Current diagnostic paradigms remain anchored in late-stage digital subtraction angiography (DSA) findings, potentially delaying therapeutic intervention until irreversible vascular remodeling occurs. This diagnostic latency may compromise the efficacy of revascularization procedures, particularly in rapidly progressive cases.

Our previous research has confirmed that patients with MMD exhibit notable elevations in pro-inflammatory and immunosuppressive capacities, alongside substantial decreases in anti-inflammatory responses and immune regulation. Several metabolites and proteins associated with these mechanisms have been identified [6]. While these findings serve as the foundation for advancing ‘precision medicine’ in MMD, the etiology of MMD is highly intricate and may extend beyond immune parameters. For example, Geng et al. utilized non-targeted gas chromatography-mass spectrometry to examine serum metabolic biomarkers for MMD, such as L-isoleucine and urea [7]. Additionally, lipidomic analysis has indicated reduced serum levels of complex membrane glycosphingolipids in MMD patients [8]. Moreover, the limited sample size hinders broader generalization of the findings.

This study established a molecular profile of serum samples from MMD patients collected at initial symptom onset, revealing early-stage metabolic and proteomic perturbations. By integrating untargeted metabolomics and quantitative proteomics in a cohort of > 60 age- and sex-matched participants (MMD patients vs. healthy controls), we identified candidate biomarkers with individual diagnostic potential and enhanced discriminatory capacity in multi-marker panels. Systems biology approaches mapped disease-associated Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, with computational network analysis highlighting ferroptosis-related multi-omic mechanisms—specifically uracil and L-glutamic acid metabolism —supported by proteomic evidence of transferrin receptor protein-1 (TFRC) and Apolipoprotein A2 (APOA2). While these findings principally advance mechanistic understanding by implicating ferroptotic stress in early MMD pathogenesis, they could inform the development of biologically stratified subtyping frameworks and guide therapeutic hypotheses targeting ferroptosis modulation. Subsequent integration of transcriptomic data to validate these ferroptosis-associated signatures across extended cohorts may further elucidate pathway dynamics, establishing a foundation for mechanism-driven interventions aimed at mitigating ferroptosis-mediated vascular injury in MMD.

Results

Study population characteristics

Analysis of test (HC = 30/MMD = 30) and validation (HC = 13/MMD = 17) cohorts revealed comparable baseline characteristics between MMD patients and HCs. Age (test:42 ± 9 vs. 41 ± 10; validation:39 ± 10 vs. 39 ± 12 years, p = 0.664) and BMI (test:21.98 ± 1.97 vs. 21.42 ± 1.96; validation:21.62 ± 1.91 vs. 21.26 ± 2.04 kg/m², p = 0.608) showed no significant differences. Sex ratios were balanced (female:46.7–61.5% HC vs. 47.1–53.3% MMD, p = 0.808). MMD cohorts exhibited similar Suzuki staging distributions (early-stage 0-II:43.3–52.9% vs. advanced III-VI:47.1–56.7%, p > 0.05) and symptom profiles (ischemic:35.3–36.7%; hemorrhagic:29.4–33.3%, p > 0.05), confirming matched demographic/clinical baselines for valid biomarker comparisons (Table S1).

Overview of multi-omic analysis of MMD patient serum

Our integrated multi-omics strategy combined proteomic and metabolomic analyses to systematically characterize molecular perturbations in Moyamoya disease (MMD) (Fig. 1A). Proteomic screening identified 755 differentially expressed serum proteins (FDR < 1%, Table S2), establishing a candidate pool for phenotypic prediction. Subsequent metabolomic profiling of the same cohort quantified 1,029 metabolites (Table S3), creating a 1,784-feature matrix for biomarker discovery. Cross-omics analysis employed binary comparisons, hierarchical clustering, and network modeling to delineate disease-associated molecular signatures.

Unsupervised clustering of proteomic data revealed pronounced segregation between MMD patients and controls (Fig. 1B), confirming systemic molecular dysregulation. From this signature, we identified Q5NV69 as a top discriminative biomarker, achieving an AUC of 0.9489 in ROC analysis (Fig. 1C-D) – surpassing established biomarkers like APOE (AUC:0.703) [9] and MMP-9 (AUC:0.730) [10]. This example demonstrates the power of unbiased proteomics in biomarker discovery.

Fig. 1
figure 1

Multi-omic analysis of MMD patient serum. (A) Workflow for analyzing serum samples in SaB patients. (B) Hierarchical clustering (using Pearson correlation) of proteins identified in all samples. (C) Comparison of Q5NV69 abundance in control samples and MMD samples. (D) Receiver Operating Characteristic (ROC) curve analysis of Q5NV69

Full size image

Identification of high-confidence biomarkers for predicting MMD

To systematically identify high-confidence biomarkers for Moyamoya disease (MMD), we implemented Random Forest (RF) machine learning—a robust nonparametric algorithm optimized for high-dimensional omics data analysis [11]. RF’s capacity to accommodate variables with heterogeneous scales/sparsity patterns [12, 13], and enable patient stratification through data-driven subgroups [14] proved critical for ranking proteomic/metabolomic candidates (Figs. 2 A, D).

Among protein biomarkers, Q5NV69 (V1-13 protein) (Fig. 2B), O00151 (PDZ and LIM domain protein 1) (Figure S1 A), and P07737 (Profilin-1) (Figure S1 B) were the highest ranked proteins with significantly increased serum levels. The key biomarkers exhibiting decreased serum levels were A2KBC1 (Anti-(ED-B) scFV) (Fig. 2C), V9HW34 (Epididymis luminal protein 213) (Figure S1 C), and A0A384MEF1 (Epididymis secretory sperm binding protein) (Figure S1 D). Applying a similar approach to the metabolomics data, we found that the highest ranked biomarkers were Com_8547_pos (Uracil) (Fig. 2E), Com_2282_pos (PC (14:0e/24:4)) (Figure S1 G), and Com_4604_neg (3-Hydroxydecanoic) (Figure S1 H) with significantly decreased serum levels. The top-ranked, identified metabolites include Com_749_pos (D-Erythro-sphingosine 1-phosphate acid) (Fig. 2F), Com_3196_neg (PC (4:0/16:3)), and Com_10094_neg (Glycerol-3-phosphate) (Figure S1E, F) with significantly increased serum levels. These molecules showed considerable diagnostic utility, similar to top-ranked protein biomarkers (ROC AUC = ~ 0.96; ROC AUC = ~ 0.90, respectively).

Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment of prioritized molecules revealed multi-system involvement in MMD pathogenesis (Figs. 2G-H). Shared proteomic/metabolomic pathways included cholesterol metabolism, ferroptosis, PPAR signaling, and actin cytoskeleton regulation. Disease-specific associations emerged through proteomic links to complement/coagulation cascades and TGF-β/HIF-1 signaling, versus metabolomic connections to ABC transporters and nitrogen metabolism. This multi-omics convergence demonstrates MMD’s pathophysiological complexity extends beyond classical inflammatory/immune paradigms.

The orthogonal nature of RF-derived biomarkers enabled synergistic prediction of MMD through combined proteomic-metabolomic signatures (Fig. 3A). Integrating the top two markers from each omics layer significantly enhanced diagnostic accuracy versus single-modality approaches. Cross-omics pathway mapping identified six core regulatory modules with mutual interactions: actin cytoskeleton regulation, ferroptosis, cholesterol/PPAR signaling, ECM-receptor interplay, and platelet activation (Fig. 3B). Extended multi-omics integration revealed broader pathway perturbations involving metabolic reprogramming (D-glutamine/taurine metabolism, nitrogen/bicarbonate handling) and neurovascular crosstalk (neuroactive ligand-receptor interactions, nicotine addiction pathways) (Fig. 3C). These coordinated multi-system alterations underscore the value of combinatorial biomarker strategies for deciphering MMD’s multifaceted pathophysiology.

Fig. 2
figure 2

Identification of high-confidence biomarkers for predicting MMD. (A) Top 25 RF proteins. (B) Abundance and ROC curve of Q5NV69. (C) Abundance and ROC curve of A2KBC1. (D) Top 25 RF metabolites. (E) Abundance and ROC curve of metabolite ID Com 8547 pos. (F) Abundance and ROC curve of metabolite ID Com 749 pos. (G) Dual-omic ROC curve (Combo: Q5NV69 + ID Com 8547 pos). (G) Top 25 RF proteins KEGG. (H) Top 25 RF metabolites KEGG

Full size image
Fig. 3
figure 3

Integration analysis of dual omics. (A) Dual-omic ROC curve (Combo: Q5NV69 + ID Com 8547 pos). (B) Top 25 RF proteins and metabolites intersection KEGG. (C) Top 25 RF proteins and metabolites integrated KEGG

Full size image

Identification of disease-associated module using WGCNA

To systematically characterize MMD-driven perturbations in serum proteomic and metabolomic profiles, we applied WGCNA to 60 patient samples. Proteomic analysis identified 601 highly variable proteins (SD > 0.5; initial 775) clustered into eight co-expression modules (soft threshold power = 10). The yellow module exhibited the strongest disease association (r = 0.64, p < 0.01) (Figs. 4A, S2). Among these modules, the yellow module was identified as the most disease-related (Correlation Coefficient (CC) = 0.64, P < 0.01) (Fig. 4A). The weighted network analysis highlighted A5YAK2 (Apolipoprotein C-IV), A2NYQ9 (Anti-folate binding protein), and A0A125U0V1 (MS-F1 heavy chain variable region) as proteins with higher weights (Fig. 4B), suggesting a close association with the MMD phenotype. Differential analysis revealed 117 upregulated proteins and 62 downregulated proteins (Fig. 4C). Moreover, 33 proteins were identified in both the differential and WGCNA analyses (Fig. 4D), indicating their significant involvement in the onset and advancement of MMD.

To investigate protein crosstalk, a functional association analysis was conducted on the intersecting proteins, revealing the interactions and regulatory relationships among them through Protein-Protein Interaction (PPI) analysis (Fig. 4E). Moreover, KEGG enrichment analysis associated the intersecting proteins with biological processes such as lysosome, HIF-1 signaling pathway, TGF-beta signaling pathway, PPAR signaling pathway, cholesterol metabolism, and ferroptosis (Fig. 4F). Notably, we observed an overlap between the KEGG enriched pathways of the intersecting proteins and the biomarkers selected through random forest analysis, indicating more significance of these pathway. Due to their association with MMD phenotype, we conducted Gene Ontology (GO) analysis on protein clusters associated with the MMD phenotype to ascertain their functional roles. The GO enrichment analysis indicated the involvement of the intersecting proteins in various biological processes, including epithelial cell proliferation in wound healing, modulation of lipoprotein and cholesterol metabolism, lipid transport regulation, acute-phase response, and high-density lipoprotein particle clearance and assembly (Fig. 4G). It is intriguing that these proteins demonstrate close connections to both inflammation and lipid metabolism, aligning with our prior research findings [6]. These findings demonstrate that in addition to inflammation and immunity, metabolic disorders also play an important role in MMD.

Metabolomic profiling of 781 filtered metabolites (SD > 0.5; initial 1029) identified 12 co-expression modules (power = 9), with the red module showing peak MMD correlation (Figs. 5A, S3). In the network analysis (Fig. 5B), metabolites Com 15,477 pos (Hesperetin), Com 18,535 pos (4-[1-(fur-2-oyl) pyrazol-5-yl]-5-methyl-1-phenylpyrazole), and Com 2373 neg (3-[(methoxycarbonyl)amino]-2,2,3-trimethylbutanoic acid) were found to have higher weights, indicating a significant correlation with the MMD phenotype. Differential analysis revealed 97 upregulated and 79 downregulated metabolites (Fig. 5C), with 28 metabolites shared between differential and WGCNA analyses (Fig. 5D), indicating their notable role in the initiation and progression of MMD. In Class I metabolite classification, Lipids and lipid-like molecules accounted for 32.14% of the metabolites, with the same proportion lacking a Class I classification (Fig. 5E). This indicates that the number of metabolites annotated in the current metabolite database is far less than the actual metabolites in the human body, which seriously hinders the progress of metabolomics research. In addition, functional association tools for interpreting metabolic data are currently lacking. The Sankey diagram for metabolite classification provides a detailed breakdown of II metabolite classifications (Fig. 5F). KEGG analysis associated the intersecting metabolites with biological processes including the Foxo signaling pathway, phospholipase D signaling pathway, gap junction, proximal tubule alanine, aspartate and glutamate metabolism, ferroptosis, purine metabolism, and biosynthesis of various antibiotics (Fig. 5G). Metabolites participating in these pathways hold significant value for further research.

Fig. 4
figure 4

Proteomic data analysis using WGCNA. (A) Module trait correlation analysis. (B) Yellow module hub protein weight network analysis. (C) Differential protein analysis. (D) Intersection of proteins between WGCNA and differential analysis. (E) Intersection protein PPI. (F) Intersection protein KEGG. (G) Intersection protein GO

Full size image
Fig. 5
figure 5

Metabolomics data analysis using WGCNA. (A) Module trait correlation analysis. (B) Red module hub protein weight network analysis. (C) Metabolomic differential analysis. (D) Intersection of metabolomics between WGCNA and differential analysis. (E) Metabolite class I classification. (F) Metabolite classification Sankey diagram. (G) Intersection metabolites KEGG

Full size image

Integrated multi-omics analysis

Integration based on the KEGG pathway can elucidate the regulatory relationship between proteins and metabolites. Therefore, we performed an integrated analysis that combined the KEGG pathway of intersections of proteins (WGCNA and differential proteins) and intersections of metabolites (WGCNA and differential metabolites), with the KEGG pathway of the top 25 proteins and metabolites identified through the RF. Our findings underscores the significance of the ferroptosis pathway (Fig. 6A). This discovery is noteworthy. We previously uncovered a strong correlation between MMD and ferroptosis [6]. However, due to the limited sample size, we were unable to conduct a deeper analysis. In this study, we identified three molecules implicated in the ferroptosis pathway: transferrin and L-glutamic acid showed a noteworthy increase, while uracil serum levels exhibited a decrease (Fig. 6B).

The above findings outline the evaluation of molecular features associated with MMD ferroptosis, analyzed based on serum abundance. However, our data set lacks representation of key ferroptosis-related proteins, which are critical to most biomarker studies in this domain. These signaling molecules frequently register below the standard detection threshold in conventional serum proteomics experiments [15], even in recent efforts toward comprehensive serum proteomic profiling [16], yet they have proven to play a pivotal role in diseases. Hence, we developed a computational approach to evaluate the significance of key ferroptosis-related proteins from our proteomic data through the utilization of functional protein association networks (Fig. 6C). These 24 proteins have been confirmed in previous studies to be key proteins in the erroptosis-related pathway [17]. We identified TFRC, an ferroptosis-related gene, as a central player in MMD, and further revealed that six additional proteins (TF, APOA2, SAA4, APOC4, APOM, and ORM1) in MMD serum exert significant influence on the ferroptosis signaling pathway (Fig. 6D). Subsequently, we performed an analysis of abundance differences on these six proteins and observed a significant upregulation in MMD serum (Fig. 6E). The GO enrichment analysis unveiled that these six proteins are predominantly associated with acute phase inflammatory response and lipid metabolism, while the KEGG analysis indicated their involvement in cholesterol metabolism, ferroptosis, and the HIF signaling pathway(Fig. 6F). Interestingly, pathways beyond ferroptosis have been documented in MMD [18].

Despite thorough and extensive functional correlation analysis of metabolic and proteomic data, new core pathways have been identified. However, due to the complexity of metabolomics data and the lack of annotated databases, the functions of certain metabolites remain unclear. In order to further investigate the regulatory interactions between metabolomics and proteomics, we conducted a correlation analysis on differential metabolites and proteins. The detailed results of the analysis are detailed in Table S4. The clustering heatmap of differential metabolites and proteins with a |correlation coefficient|≥0.6 shows their mutual regulatory clusters (Figure S4). Many significant correlations between differential proteins and metabolites were found, which can be displayed as a network (P < 0.05) (Fig. 7A). To further emphasize the mutual regulatory connection between key differential proteins and metabolites, we established the criteria of a correlation coefficient exceeding 0.6 and a p-value below 0.05 for network reconstruction (Fig. 7B). Significantly, the analysis revealed the absence of uracil in this regulatory network. On the basis of previous network analysis, we added another parameter, namely Edgecount greater than 3, and found that the three previously discovered molecules involved in ferroptosis pathway were not in this network (Fig. 7C). This result is not unexpected. Adding thresholds in network analysis aims to optimize the network and isolate tightly connected components or nodes. The absence of three ferroptosis-related molecules in the resulting network does not imply their biological insignificance. Correlation analysis serves as the basis for network construction; unlike functional analysis, which delves into the roles and mechanisms of molecules in biological pathways. Next, we specifically reconstructed the network involving three molecules and observed a significant correlation between uracil and several proteins (Fig. 7D). However, the correlation between them is moderate, with a coefficient ranging between 0.4 and 0.5. Another noteworthy discovery is the predominance of a positive correlation between differential proteins and metabolites (Fig. 7E).

Fig. 6
figure 6

Integrated multi-omics analysis based on enrichment pathways. (A) Differential analysis, WGCNA, and RF intersection KEGG. (B) Relative abundance of proteins and metabolites related to the ferroptosis pathway in the MMD patients’ serum. (C) Ferroptosis associated with gene inference. (D) Generating a refined network. (E) Relative abundance of ferroptosis related proteins in MMD patients’ serum. (F) GO and KEGG of proteins related to the ferroptosis pathway in the MMD patients’ serum

Full size image
Fig. 7
figure 7

Integrated multi-omics analysis based on correlation analysis. (A) Correlation network of differential metabolites and proteins with significant correlation. (B) Correlation network of differential metabolites and proteins with |correlation coefficient|≥0.6 and P<0.05. (C) Correlation network of differential metabolites and proteins with |correlation coefficient|≥0.6, P<0.05, and edgecount ≥ 3. (D) Correlation network between proteins and metabolites in MMD patients’ serum associated with the ferroptosis pathway. (E) Circos plot of correlations across differential metabolites and proteins

Full size image

Exploring genes associated with ferroptosis through transcriptomic data in MMD patients

To investigate the relationship between ferroptosis-related genes and MMD, we performed transcriptomic data analysis in patients with MMD. Firstly, the data is normalized and further analyzed to remove batch effects (Fig. 8A-C). The heatmap analysis of the expression values of 24 ferroptosis-related genes is shown in Fig. 8D. The expression distribution of ferroptosis reveals notable distinctions between NFE2L2 and SLC1A5 in individuals with MMD compared to the control group (Fig. 8E). These results further confirm the important role of ferroptosis-related pathways in MMD, which is consistent with what was found in the proteome and metabolome.

Fig. 8
figure 8

Exploring genes associated with ferroptosis through transcriptomic data in MMD. (A-C). Normalization and batch effect correction of transcriptomic data. (D) Heat map of ferroptosis-related genes. (E) The expression distribution of ferroptosis in MMD tissues and aneurysm tissues

Full size image

Validation of proteins and metabolites associated with ferroptosis and identified by RF in independent cohorts

The analyses described above delineate the biomarkers and pathways in MMD; however, it remains ambiguous whether they are merely bystanders or actively contribute to disease outcomes. To bridge this gap, we leveraged an additional independent cohort of patients with MMD to assess the levels of ferroptosis-related proteins and metabolites in vivo.

Our multi-omics analysis revealed disrupted host protein and metabolic profiles, with significant elevations in transferrin and L-glutamic acid, and a marked decrease in uracil serum levels (Fig. 6B). Due to the absence of established cell and animal models for MMD, we opted for validation using an independent cohort (Fig. 9A). Our study revealed that the levels of the ferroptosis-related protein, transferrin receptor protein-1, and the ferroptosis-related metabolite, L-glutamic acid, were significantly elevated in patients diagnosed with MMD compared to those in the control group (p < 0.05) (Fig. 9B) (Tables S5, S6). This is consistent with our results in the testing cohort. Next, we validate the biomarkers identified through machine learning. In the proteomic dataset, Tβ4 was the only protein displaying significant differences. In the metabolomic data, Uracil, 3-Hydroxydecanoic acid, and Glycerol-3-phosphate exhibited notable variances (Fig. 9C) (Tables S5, S6). While the link between uracil and MMD has not been previously documented, studies have revealed a close connection between uracil and the immunological reaction [19]. This aligns with the understanding that MMD is an autoimmune disorder.

In the ROC analysis of validated biomarkers, six markers demonstrated varying diagnostic performances, with transferrin receptor protein-1 exhibiting the highest discriminatory capacity (AUC = 81.90), followed by glycerol-3-phosphate (AUC = 77.86), L-glutamic acid (AUC = 77.38), uracil (AUC = 75.11), 3-hydroxydecanoic acid (AUC = 75.34), and TP4 (AUC = 74.21). These results highlight the differential diagnostic potential of these biomarkers, with transferrin receptor protein-1 emerging as the most robust candidate, while others demonstrated moderate to strong discriminative capacities, collectively underscoring their potential utility in diagnostic applications. (Fig. 9D) (Tables S5, S6).

Fig. 9
figure 9

Validation of proteins and metabolites associated with ferroptosis and identified by random forest by independent cohorts. (A). Schematic for an independent cohort. (B) Validation of ferroptosis-related proteins and metabolites. (C) Validation of Biomarkers Identified by Machine Learning. It is noteworthy that uracil is not only the most significant biomarker for differences and RF top, but also an ferroptosis-related metabolite. (D) ROC analysis of validated biomarkers

Full size image

Discussion

Conventional strategies for delineating MMD biomarkers have principally relied on single omics analysis. Due to the absence of dependable animal and cellular models for MMD, a substantial portion of these biomarkers remain unconfirmed. In this study, we have instituted a novel MMD-related biomarker evaluation standard, scrutinizing a wider multi-omics landscape of MMD, and forgoing the presupposition that all clinically salient details are linked to immune and inflammatory responses. Furthermore, we have uncovered a new pathway implicated in MMD. By adopting multi-omics approaches, we have characterized several features and multivariate models that can effectively forecast MMD. These identified features, when integrated with the newly discovered pathways, can enhance prognostic value and assist in excavating molecular mechanisms. To bolster the reliability of biomarkers, we have expanded this study by incorporating additional computational analysis and conducting in vivo validation using independent cohorts.

While preliminary explorations of MMD biomarkers have been reported [10, 20], their clinical utility remains limited due to small cohort sizes and lack of multicenter validation. Current diagnostic protocols continue to depend on DSA as the definitive gold standard, though its invasive nature and high cost limit accessibility in resource-constrained settings. Notably, non-invasive neuroimaging modalities—particularly high-resolution MRA and computed tomography angiography (CTA)—have been rigorously validated for MMD and moyamoya syndrome (MMS) evaluation [21, 22]. Nevertheless, diagnostic delays persist, potentially attributable to atypical radiographic and clinical presentations in early-stage disease. Although surgical revascularization provides symptomatic relief, therapeutic innovation has stagnated compared to advances in other neurovascular disorders. Crucially, while revascularization transiently improves perfusion parameters, it fails to address the underlying progressive vasculopathy [23]. The validated biomarkers demonstrated in this investigation offer a non-invasive diagnostic strategy with enhanced specificity for early-stage MMD detection, addressing critical diagnostic gaps that contribute to delayed clinical management while enabling more timely therapeutic intervention.

Beyond standard biomarker characterization, we implemented complementary computational workflows: Differential analysis identified MMD-dysregulated proteins/metabolites, while WGCNA pinpointed phenotype-correlated modules. Intersection analysis of these datasets yielded consensus biomarkers, which underwent KEGG pathway integration with RF-selected features. This multi-algorithm convergence revealed ferroptosis as a pathway mechanistically linked to human MMD—the first demonstration of this association, offering novel mechanistic insights into disease pathogenesis.

Notably, our top-ranked biomarker uracil—a natural RNA base also implicated in ferroptosis—participates in genomic integrity through cytosine deamination or dUTP misincorporation during DNA replication, constituting a common DNA lesion [24]. Recent studies indicate that incorporating uracil bases into viral genomic DNA intermediates during genome replication could serve as an innate immune defense mechanism against specific viruses [25]. Since Sheehy et al. discovered that human endogenous APOBEC3G protein serves as a restriction factor in inhibiting HIV-1 replication, the issue of innate immunity to the virus has received widespread attention [26]. The discovery of APOBEC family members highlights the core role of uracil as a barrier to infection, firstly as a participant in antibody diversification in adaptive immunity, and secondly as an effective antiviral drug itself [24]. This immunological nexus aligns with reported associations between MMD and viral infections [27, 28], as well as RNF213’s dual roles in MMD pathophysiology and antiviral response modulation. RNF213 mitigates enterovirus-related symptoms [29], while RNF213-deficient mice exhibit heightened susceptibility to RVFV infection compared to overexpression models [30]. We observed significantly reduced serum uracil levels in MMD patients, paralleling RNF213 mutation effects. Though mechanistic details require elucidation, this first-reported uracil-MMD association bridges ferroptosis, antiviral immunity, and vascular pathogenesis, offering a unified framework for future investigation.

L-glutamic acid, a central ferroptosis-related metabolite in MMD, acts as the brain’s primary excitatory neurotransmitter and free amino acid, predominantly synthesized from glutamine via the Krebs cycle [31]. It mediates synaptic plasticity [32], regulates dopaminergic neurotransmission [33], and contributes to ~ 30% of CNS synaptic activity [34]. Dysregulation of glutamate is implicated in neuropsychiatric disorders, epilepsy, and neurodegenerative diseases [35], with its excitotoxicity exacerbating ischemic injury in stroke and neurodegeneration [36]. Our study confirmed elevated glutamate levels in MMD patients, consistent with prior targeted metabolomics reports [37]. Although not a novel biomarker, independent cohort validation reinforces its pathogenic relevance. We propose that glutamate accumulation may drive MMD progression through ferroptosis-mediated mechanisms, linking metabolic disruption to cerebrovascular pathology.

Transferrin receptor 1 (TFRC) emerges as a critical ferroptosis regulator in MMD pathogenesis. Ferroptosis, an iron-dependent cell death pathway marked by lethal lipid peroxidation, has emerged as a pathogenic mechanism across multiple disorders, with TFRC-mediated iron influx driving Fe²⁺ accumulation and subsequent cellular demise [38]. This process is particularly implicated in ischemic neuronal damage observed in stroke and brain injuries [39]. Our multi-omics data revealed dysregulation of ferroptosis markers (uracil, TFRC) in MMD, suggesting its involvement in vascular stenosis progression. Mechanistically, endothelial ferroptosis from iron overload and GPX4 inactivation compromises vascular integrity, exposing VSMCs to inflammatory/hydrodynamic stress that stimulates intimal hyperplasia [40]. VSMC ferroptosis concurrently induces compensatory proliferation and ECM overproduction, mirroring atherosclerotic remodeling [41]. A hypoxia-ROS amplification loop perpetuates stenosis: chronic hypoperfusion stabilizes HIF-1α, enhancing iron uptake and lipid peroxidation to exacerbate ferroptosis and oxidative damage [42]. ECM stiffening via dysregulated 5-HETE-mediated arachidonic acid metabolism further promotes collagen deposition and fibrotic wall thickening [43]. These pathophysiological interactions collectively create a self-reinforcing cycle driving MMD progression. Our findings suggest ferroptosis inhibition holds therapeutic potential for targeting ischemia-related pathologies, particularly in MMD. While first characterized in this context, ferroptosis mechanisms now represent a critical focus for elucidating disease pathways and developing targeted interventions to disrupt vascular degeneration in MMD.

This study demonstrates the utility of serum proteomics for computational characterization of ferroptosis-related proteins, revealing significant alterations in TF, APOA2, SAA4, APOC4, APOM, and ORM1 functionally linked to ferroptosis pathways in MMD. The methodology enables systematic correlation of these proteins with proteomic signatures, validating ferroptosis as a central pathogenic mechanism. Advanced pathway-centric interrogation identified novel ferroptosis-associated candidates in MMD, with KEGG enrichment analysis linking these proteins to cholesterol metabolism – consistent with established connections between cholesterol homeostasis disruption and ferroptosis [44].

Complementing proteomic and metabolomic profiling, we expanded our investigation to transcriptomics, revealing ferroptosis-related differential expression of NFE2L2 and SLC1A5 in MMD patients. While the functional consequences require further validation, these transcriptional signatures corroborate ferroptosis pathway activation, reinforcing its mechanistic relevance in MMD pathogenesis.

Finally, Validation confirmed the diagnostic reliability of RF-selected biomarkers: Tβ4, a pleiotropic regulator of wound healing and inflammation [45], demonstrates ocular anti-inflammatory effects via neutrophil suppression and metalloproteinase modulation [46, 47], with elevated serum levels in inflammatory bowel disease [48] aligning with our observed Tβ4 upregulation in MMD. In addition, 3-Hydroxydecanoic acid (3HD), biosynthesized via PhaG-mediated pathways in E. coli [49], engages immune signaling through SD-RLK/LORE receptors in A. thaliana recognizing 3-OH-C10:0 [50]. Glycerol-3-phosphate, central to glycolysis and glycerolipid synthesis [51], underscores our metabolomic findings linking immune dysregulation and metabolic dysfunction in MMD pathogenesis.

These findings underscore the potential of multi-biomarker panels to refine moyamoya diagnosis and risk stratification. TfR1’s diagnostic prominence highlights its potential as a therapeutic target, given its role in vascular remodeling. However, validation in diverse cohorts is essential, particularly given moyamoya’s genetic and ethnic heterogeneity. Future studies should integrate these biomarkers with advanced imaging and genetic data to refine predictive models. Mechanistic studies exploring biomarker-disease pathways may unveil novel therapeutic targets, advancing precision medicine in moyamoya disease.

Overall, our primary objective is to set a high standard in the domain of MMD biomarkers by developing an accurate multi-omics diagnostic model for MMD occurrence. Intense future research with equivalent substantiation may unveil more theoretically viable discoveries, further enriching our comprehension of MMD pathogenesis. This study serves as the cornerstone for tools harnessing multiple biomarkers. Importantly, the credibility of these biomarkers is enhanced by the validation of independent datasets.

This study has several limitations that warrant consideration. First, the relatively small sample size may constrain the statistical power to detect subtle biomarker differences and generalizability of findings. Second, all samples were derived from a single clinical center, introducing potential selection bias and limiting external validity; future multi-center studies are required to confirm the robustness of our conclusions. Third, while mass spectrometry enabled discovery of novel biomarkers (e.g., Q5NV69/V1-13 protein), the lack of standardized ELISA reagents for these targets precluded independent validation in the external cohort—a critical step for clinical translation. To address this, we prioritized platform-homogeneous comparisons within the test cohort and explicitly recommend future development of certified assays for these candidates.

STAR methods

Availability of resources

Designated lead contact for additional information and requests regarding resources and reagents, please contact the designated lead, Qingbao Guo (guo18291908296@163.com), who will address and fulfill your inquiries.

Materials availability

This study did not yield any novel reagents.

Our team provides proteomic data through an accessible online repository (https://www.iprox.org/; DOI: IPX0009899000) for researchers to access sera from moyamoya disease patients. Metabolomic findings were created and stored at the Fifth Medical Center of the People’s Liberation Army General Hospital in China. For access to the supportive data of this research, kindly reach out to the corresponding authors. The R scripts utilized for the analysis in this manuscript can be obtained by request.

Experimental subjects details

Human Studies. After obtaining approval from the Ethics Review Committee of the Fifth Medical Center (Approval number: ky-2020-9-22), serum samples were obtained from patients at the Fifth Medical Center of the People’s Liberation Army General Hospital (PLAGH). This institution is a leading tertiary nursing academic medical center, and is a major MMD diagnosis and treatment institution in China, having treated over 10,000 MMD patients. Patients aged between 18 and 59, regardless of gender, are eligible to be included in the biobank study and their specimens collected as part of standard medical treatment after obtaining written informed consent. Further details regarding the collection of subject data are outlined in the subsequent Method Details section.

Methods details

Patients and Samples

The study did not analyze a continuous sample of MMD patients but opted to select MMD individuals (i.e., hospitalized; n = 30) and healthy controls (partners or immediate family members of patients; n = 30) from the MMD biobank for multi-component serum analysis. External validation included the extraction of MMD patients (i.e., inpatients; n = 13) and healthy controls (partners or immediate relatives of patients; n = 17) from the biobank for serum ELISA protein quantification and targeted metabolomics analysis.

The patient’ s serum sample was collected on the morning of the second day of the initial hospital admission and preserved at -80 °C for subsequent analysis.

Study design and blood samples

This study recruited adults with MMD and normal controls from our center as research subjects. The inclusion criteria were as follows: (1) Patients diagnosed with MMD based on digital subtraction angiography (DSA) according to the diagnostic criteria for MMD [3]; (2) Age > 18 years; (3) Informed consent to participate in the study. Exclusion criteria: Systemic diseases that may interfere with the research results, such as hypertension/diabetes, MMD syndrome, or other central nervous system diseases like central nervous system tumors, severe brain trauma history, prior craniotomy, and similar conditions.

Proteomic methods

Library construction

A low-abundance peptide separation was performed using a Rigol L3000 HPLC system with a Waters BEH C18 column (4.6 × 250 mm, 5 μm) at a flow rate of 1 mL/min and a column temperature of 50 °C. Elution was carried out with a gradient of mobile phases A (2% acetonitrile with ammonium hydroxide, pH 10.0) and B (98% acetonitrile with ammonium hydroxide, pH 10.0), starting at 3% B and increasing in steps: 3–8% B (0.1 min), 8–18% B (11.9 min), 18–32% B (11 min), 32–45% B (7 min), 45–80% B (3 min), hold at 80% B (5 min), peak at 85% B (0.1 min), and back to 5% B (6.9 min). Eluates were monitored at 214 nm, collected every minute, and grouped into four fractions, covering both high and low-abundance peptides. These fractions were vacuum-dried, reconstituted in 0.1% formic acid (FA) in water, and mixed with 0.2 µL of standard peptides (iRT kit, Biognosys) for analysis.

For the transition library, DDA mode shotgun proteomic analyses were conducted using an EASY-nLC™ 1200 system connected to a Thermo Fisher Scientific Orbitrap Q-Exactive HF-X mass spectrometer. Peptides from the fractions were dissolved in 0.1% FA and separated on a homemade C18 NanoTrap column (2 cm × 100 μm, 3 μm) and an analytical column (15 cm × 150 μm, 1.9 μm) with a 120-min linear eluent B gradient (0.1% FA in 80% acetonitrile) in eluent A (0.1% FA in water) at a flow rate of 600 nL/min. The gradient steps included 5–10% B (2 min), 10–40% B (105 min), 40–50% B (5 min), 50–90% B (3 min), and 90–100% B (5 min).

The Q-Exactive HF-X mass spectrometer operated in positive polarity mode with a spray voltage of 2.3 kV and a capillary temperature of 320 °C. Full MS scans ranged from 350 m/z to 1500 m/z at a resolution of 60,000 (at 200 m/z), with an AGC target of 3 × 10^6 and a maximum ion injection time of 20 ms. The 40 most abundant precursor ions were fragmented by HCD analysis at a resolution of 15,000 (at 200 m/z), with an AGC target of 5 × 10^4 and a maximum ion injection time of 45 ms. The normalized collision energy was set at 27%, with an intensity threshold of 2.2 × 10^4 and a dynamic exclusion period of 40 s.

MS analysis: DIA mode

For peptide analysis in serum samples, they were combined with standard peptides and introduced into an EASY-nLC™ 1200 UHPLC system connected to an Orbitrap Q-Exactive HF-X mass spectrometer set to run in DIA mode. The liquid conditions were consistent with those utilized in the DDA model. In the DIA acquisition process, the MS1 resolution was adjusted to 60,000× (at 200 m/z), while the MS2 resolution was configured at 30,000× (at 200 m/z). The m/z range of 350–1500 was divided into 30 acquisition windows. The full-scan AGC target was set at 3 × 106 with a 50 ms injection time. DIA parameters encompassed a normalized collision energy (NCE) of 27%, a target value of 1 × 106, and an automatic maximum injection time to facilitate continuous MS operation in parallel ion filling and detection mode.

Proteome data analysis

In the study, data analysis and visualization of the DIA data were conducted using the Proteome Discoverer platform (PD 2.2, Thermo Fisher Scientific), Biognosys Spectronaut v.9.0, and the R statistical framework. Biognosys Spectronaut v.9 facilitated MS2-based label-free quantification of the raw DIA data, following the method outlined by Bruder et al. [52]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) database was employed to examine protein families and pathways, with the enrichment pipeline [53] utilized for KEGG enrichment analysis.

Protein extraction and peptide Preparation

The methodology of protein extraction and peptide preparation adhered to standard laboratory protocols. Each sample was primarily lysed using DB lysis buffer (comprising 8 M urea, 100 mM TEAB at pH 8.5), and subjected to centrifugation at 12,000 rpm for a period of 20 min at 4 °C. Subsequently, a pooled sample was constructed for the creation of a spectral library for DIA protein identification, by amalgamating 20–30µL of serum from every sample. Data-Dependent Acquisition protein identification process (DDA) comprised two split duplicates from the pooled sample, one of which was treated with the Bio-Rad ProteoMiner TM Protein Enrichment Kit for the removal of abundantly present proteins. These duplicates were subjected to treatment using 2 mM DTT at a temperature of 56 °C for one hour, followed by alkylation through iodoacetic acid under dark conditions at ambient temperature for another hour. Upon completion of acetone precipitation, the protein pellet was dissolved in 0.1 M of triethylammonium bicarbonate (pH 8.5) and 8 M of urea buffer. Consequently, the supernatant from each sample containing exactly 0.1 mg of protein was digested using Promega Trypsin Gold at 37 °C for 16 h. The sample then underwent a desalting process to remove urea content, and dehydration was achieved through vacuum centrifugation.

Metabolomics methods

Metabolomics extraction

The serum samples, each 100 µL, were mixed with 80% pre-chilled methanol and incubated on ice for 5 min. Subsequently, centrifugation at 15,000×g and 4 °C for 20 min was conducted. The resulting supernatant was diluted to a 53% methanol final concentration with LC-MS grade water, followed by a second centrifugation at 15,000×g and 4 °C for 20 min.

For LC-MS/MS analysis, a ThermoFisher Hypersil Gold column (100 × 2.1 mm, 1.9 μm) paired with a Thermo Fisher Orbitrap Q-Exactive™ HF-X mass spectrometer was utilized. Samples underwent chromatography on the column with a 17-minute linear gradient at a flow rate of 0.2 mL/min. The raw UHPLC-MS/MS data were processed using Compound Discoverer software (v3.1, Thermo Fisher), along with R (v3.4.3), Python (v2.7.6), and CentOS (v6.6) for functions such as peak alignment, quantitation of metabolites, and normalization of peak intensities based on total spectral intensity. Peak data verification for precise and quantitative results was performed against the mzCloud, mzVault, and MassList databases.

Metabolome data analysis

Metabolites were annotated through the utilization of the KEGG database. Statistical significance (p-value) was determined using univariate analysis (t-test). Differential metabolites were identified based on criteria including VIP > 1, P < 0.05, and fold change ≥ 2.0 or FC ≤ 0.5. Volcano plots, plotting log2 (fold change) against log10 (P-value) of the metabolites, were generated using the ggplot2 package in R to filter metabolites of interest. Normalization of data was performed by calculating z-scores for the intensity areas of differential metabolites, followed by the creation of cluster heatmaps using the Pheatmap package in R.

Correlation analysis among the differential metabolites was conducted using the cor() function in the R language, employing the Pearson method. Statistically significant correlations were determined with the cor.Mtest() in R at P < 0.05. Correlation plots were crafted utilizing the Corrplot package in the R software. Enrichment analysis of metabolic pathways for the differential metabolites was carried out by assessing whether the ratios met the threshold x/n > y/N to identify enriched pathways. Statistically significant enrichment of metabolic pathways was established through this criterion.

Statistical analyses of MS data

Binary comparison: Initially, binary comparison is employed to identify biomarkers associated with the prediction of MMD. Two types of binary analysis were employed: T-test, a validated approach for biomarker selection [54], and the random forest (RF) feature selection method, known for mitigating biases associated with individual feature selection methods. The screening of differential metabolites primarily involves assessing three parameters: Variable Importance in the Projection (VIP), fold change (FC), and P-value. VIP denotes the Variable Importance in the Projection of the first principal component of the PLS-DA model and indicates the metabolites’ contribution to grouping [55]. FC, signifies the ratio of mean values among all biological replicates within each metabolite in the comparison group. P-value, derived from a T-test, reflects the degree of significant difference. The thresholds set are VIP > 1.0, FC > 1.5, or FC < 0.667, and P-value < 0.05 [56, 57]. Upregulated proteins were identified if the FC was greater than 1.2 and P-value was less than 0.05. Conversely, downregulated proteins were defined by an FC of less than 0.83 and a P-value below 0.05 [58]. RF algorithm proposed by Leo Breiman is widely acclaimed among many machine learning algorithms [14]. This method constructs many decision trees with low correlation and introduces randomness in feature selection through bagging technique [59]. Each decision tree consists of multiple nodes, with each node representing the split rule based on a single variable. The collective “voting” of decision trees ultimately determines the classification outcome. When establishing a Random Forest, it forms the training data for each tree by continuously repeating subsampling of samples. Among them, approximately one-third of the data not selected is considered out-of-bag data (OOB), which is used for model validation. The number of decision trees is adjusted to minimize the error rate as much as possible and prevent the model from overfitting the training data [60, 61]. The performance of the model created based on Random Forest can be estimated by validating using OOB data. At the same time, the error rate of OOB can also estimate the model’s generalization ability. Compared to traditional statistical methods, RF stands out for its high accuracy, strong ability to prevent overfitting, applicability to large-scale datasets and multivariable features, interpretability of results, handling of data missingness and outliers, and efficient data processing advantages [62]. The RF algorithm was implemented using the randomForest package and the model’s performance was evaluated through a 5-fold cross-validation technique.

Construction of weighted gene co-expression network analysis (WGCNA) networks and identification of modules

In recent decades, a plethora of computational methods have been developed for analyzing complex high-throughput data. Mounting evidence suggests that employing network-based analysis strategies focused on networks, rather than individual genes or proteins, can more comprehensively capture cellular response mechanisms. The Weighted Gene Co-expression Network Analysis (WGCNA) methodology is adept at constructing gene co-expression networks, unveiling correlations among genes through an analysis of patterns and associations in gene expression data. WGCNA facilitates the identification of gene clusters and the elucidation of associations between these clusters and specific phenotypes, thereby aiding in the interpretation of the biological significance of gene expression data [63]. WGCNA enables global-level analysis and prioritizes the capture of low-abundance genes and subtle expression changes to ensure no information is overlooked. Recent advancements indicate that the scope of WGCNA application has expanded into the realms of proteomics and metabolomics data analysis [64]. To investigate co-expression patterns among metabolites and their relevance to clinical features, we opted to analyze identified metabolites using the “WGCNA” package in the R programming language. The network construction involved setting the minimum module size to 10 and the threshold for merging modules to 0.25, while maintaining all other parameters at their default values in the package.

Multi-omic data integration

Multi-omics integration analysis plays a crucial role in unraveling complex biological mechanisms by combining information from different molecular levels. In metabolomics and proteomics studies, two main approaches are commonly adopted. Firstly, correlation analysis based on relative abundance provides insights into the relationships between metabolites and proteins across various biological conditions [65]. Secondly, utilizing KEGG pathway analysis allows for the exploration of functional regulatory connections between metabolites and proteins sharing the same biochemical pathways [66]. This integrative framework enhances our understanding of cellular dynamics and can uncover key interactions involved in metabolic processes. The specific method has been previously described in detail.

Based on the network-based inference of iron death proteins. Knowledge-based networks are used to infer the relative contributions of major ferroptosis-related proteins to observed proteomic changes. Firstly, the list of ferroptosis-related genes (ACSL4, ALOX15, ATP5MC3, CARS, CDKN1A, CISD1, CS, DPP4, EMC2, FANCD2, FDFT1, GLS2, GPX4, HSPA5, HSPB1, LPCAT3, MT1G, NCOA4, NFE2L2, RPL8, SAT1, SLC1A5, SLC7A11, and TFRC) [17] is submitted to the STRING database tool along with the intersection of differential analysis and WGCNA proteins. The network is exported to a simple tabular output and visualized using Cytoscape. Subsequently, ferroptosis-related proteins are filtered to have at least five connections with the observed protein network. We obtained 6 proteins, further conducting relative abundance comparison and functional enrichment analysis on them.

To test the above ferroptosis-related protein inference method, we further explore transcriptomic data of MMD to validate whether the ferroptosis-related pathway in MMD is interfered. The specific method is as follows. Microarray data was obtained from the GEO database at http://www.ncbi.nih.gov/geo in MINiML format. The genes related to ferroptosis were sourced from the systematic analysis by Ze-Xian Liu et al. [17], focusing on the anomalies and functional implications of ferroptosis in cancer. The ggord package in R software was employed for generating PCA plots, while the pheatmap package was utilized for creating heatmaps. The significance of ferroptosis-related molecule expression distribution in MMD tissue and control tissue was assessed using the Wilcoxon test.

Validation of proteins and metabolites associated with ferroptosis and identified by random forest by independent cohorts

Targeted metabolomics assay

Thaw the sample gently on ice, then transfer 0.20mL of the thawed sample to a 2mL centrifuge tube. Add 1.5mL of a chloroform and methanol mixture (at a volume ratio of 2:1) to the tube and vortex mix for 1 min. Grind at 65 Hz for 180 s, followed by sonication of the sample at 4 °C for 30 min. Centrifuge at 3000 g for 10 min, transfer the supernatant to a clean tube, and evaporate it using nitrogen gas. Subsequently, re-suspend the dried sample in a mixture of isopropanol and methanol (100uL) and vortex mix for 60 s. Finally, centrifuge again at 12,000 g at 4 °C for 10 min, and dilute the supernatant 100-fold for subsequent analysis. Instrumentation: Employing a liquid chromatography-mass spectrometry system comprising the Waters Acquity UPLC chromatograph and the AB SCIEX 5500 QQQ-MS mass spectrometer. UPLC-QQQ-MS methodology: The chromatographic analysis utilizes the Acquity UPLC HSS T3 column (1.8 μm, 2.1 mm * 100 mm). Chromatographic parameters: The column temperature is held at 40 °C, with a mobile phase flow rate set at 0.30 mL/min. Transfer the prepared standard solution into the autosampler vial for injection; confirm the target compound by the retention time of the specific peak in the chromatogram.

The areas of the peaks were quantified with the MultiQuant software [67], and the concentrations of individual components in the sample were determined using the established standard curve.

Bar graphs comparing the two datasets were generated using the ‘ggpubr’ package (v0.4.0) and the ‘ggplot2’ package (v4.4.2) in R software (v4.2.2).

Enzyme-linked immunosorbent assay

To allow for equilibration, leave the reagents and samples at room temperature for 120 min. Set up standard, test, and control wells. Sequentially add 50 uL of standard solution with varying concentrations to the standard wells. Place 50 uL of the sample solution in the test sample wells and 50 uL of sample dilution solution in the control wells. Subsequently, apply 100 uL of horseradish peroxidase (HRP) labeled antibody to each well, seal the plate with a lid, and incubate at 37 °C for 60 min. Once incubation is complete, remove the liquid from each well and pat dry with absorbent paper. Fill each well with a wash buffer (350 uL), discard after 20 s, and repeat this washing process 5 times. Following the wash steps, add 50 uL each of substrate A and substrate B to every well, then place the plate in the dark at 37 °C for a 15-minute incubation. Finally, add 50 uL of the stop solution and measure the optical density (OD) at 450 nm of each well within 15 min.

Data availability

No datasets were generated or analysed during the current study.

References

  1. Wakai K, et al. Epidemiological features of Moyamoya disease in Japan: findings from a nationwide survey. Clin Neurol Neurosurg. 1997;99(Suppl 2):S1–5.

    Article  PubMed  Google Scholar 

  2. Kuriyama S, et al. Prevalence and clinicoepidemiological features of Moyamoya disease in Japan: findings from a nationwide epidemiological survey. Stroke. 2008;39(1):42–7.

    Article  PubMed  Google Scholar 

  3. Fujimura M, et al. 2021 Japanese guidelines for the management of Moyamoya disease: guidelines from the research committee on Moyamoya disease and Japan stroke society. Neurol Med Chir (Tokyo). 2022;62(4):165–70.

    Article  PubMed  Google Scholar 

  4. Arias EJ, et al. Advances and surgical considerations in the treatment of Moyamoya disease. Neurosurgery. 2014;74(Suppl 1):S116–25.

    Article  PubMed  Google Scholar 

  5. Carrozzini T et al. Proteome profiling of the dura mater in patients with Moyamoya angiopathy. Int J Mol Sci. 2023; 24(13).

  6. Guo Q, et al. Proteomic and metabolomic characterizations of Moyamoya disease patient Sera. Brain Behav. 2023;13(12):e3328.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Geng C, et al. Metabolomic profiling revealed potential biomarkers in patients with Moyamoya disease. Front Neurosci. 2020;14:308.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Dei Cas M et al. Plasma lipid profiling contributes to untangle the complexity of Moyamoya arteriopathy. Int J Mol Sci. 2021; 22(24).

  9. Wu H, et al. APOE as potential biomarkers of Moyamoya disease. Front Neurol. 2023;14:1156894.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Lu J, et al. MMP-9 as a biomarker for predicting hemorrhagic strokes in Moyamoya disease. Front Neurol. 2021;12:721118.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Ishwaran H, et al. Random survival forests. Ann Appl Stat. 2008;2:841–60.

  12. Chen X, Ishwaran H. Random forests for genomic data analysis. Genomics. 2012;99(6):323–9.

    Article  CAS  PubMed  Google Scholar 

  13. Cutler DR, et al. Random forests for classification in ecology. Ecology. 2007;88(11):2783–92.

    Article  PubMed  Google Scholar 

  14. Breiman L. Random forests. Mach Learn. 2001;45:5–32.

    Article  Google Scholar 

  15. Geyer PE, et al. Revisiting biomarker discovery by plasma proteomics. Mol Syst Biol. 2017;13(9):942.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Dey KK, et al. Deep undepleted human serum proteome profiling toward biomarker discovery for Alzheimer’s disease. Clin Proteom. 2019;16:16.

    Article  Google Scholar 

  17. Liu Z, et al. Systematic analysis of the aberrances and functional implications of ferroptosis in Cancer. iScience. 2020;23(7):101302.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Asselman C, et al. Moyamoya disease emerging as an immune-related angiopathy. Trends Mol Med. 2022;28(11):939–50.

    Article  PubMed  Google Scholar 

  19. Hagen L, et al. Genomic uracil and human disease. Exp Cell Res. 2006;312(14):2666–72.

    Article  CAS  PubMed  Google Scholar 

  20. Fang YC et al. Pathological Circulating factors in Moyamoya disease. Int J Mol Sci. 2021; 22(4).

  21. Yamada I, et al. High-resolution turbo magnetic resonance angiography for diagnosis of Moyamoya disease. Stroke. 2001;32(8):1825–31.

    Article  CAS  PubMed  Google Scholar 

  22. Gonzalez NR, et al. Adult Moyamoya disease and syndrome: current perspectives and future directions: A scientific statement from the American heart association/american stroke association. Stroke. 2023;54(10):e465–79.

    Article  PubMed  Google Scholar 

  23. Scott RM, Smith ER. Moyamoya disease and Moyamoya syndrome. N Engl J Med. 2009;360(12):1226–37.

    Article  CAS  PubMed  Google Scholar 

  24. Sire J, et al. Uracil within DNA: an actor of antiviral immunity. Retrovirology. 2008;5:45.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Nilsen H, et al. Nuclear and mitochondrial uracil-DNA glycosylases are generated by alternative splicing and transcription from different positions in the UNG gene. Nucleic Acids Res. 1997;25(4):750–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Sheehy AM, et al. Isolation of a human gene that inhibits HIV-1 infection and is suppressed by the viral Vif protein. Nature. 2002;418(6898):646–50.

    Article  CAS  PubMed  Google Scholar 

  27. Hsiung GY, Sotero de M, Menezes. Moyamoya syndrome in a patient with congenital human immunodeficiency virus infection. J Child Neurol. 1999;14(4):268–70.

    Article  CAS  PubMed  Google Scholar 

  28. Tanigawara T, et al. Studies on cytomegalovirus and Epstein-Barr virus infection in Moyamoya disease. Clin Neurol Neurosurg. 1997;99(Suppl 2):S225–8.

    Article  PubMed  Google Scholar 

  29. Echizenya I, et al. Reversible cerebral angiopathy after viral infection in a pediatric patient with genetic variant of RNF213. J Stroke Cerebrovasc Dis. 2020;29(2):104549.

    Article  PubMed  Google Scholar 

  30. Houzelstein D, et al. The ring finger protein 213 gene (Rnf213) contributes to rift Valley fever resistance in mice. Mamm Genome. 2021;32(1):30–7.

    Article  CAS  PubMed  Google Scholar 

  31. Yano S, Tokumitsu H, Soderling TR. Calcium promotes cell survival through CaM-K kinase activation of the protein-kinase-B pathway. Nature. 1998;396(6711):584–7.

    Article  CAS  PubMed  Google Scholar 

  32. McDonald JW, Johnston MV. Physiological and pathophysiological roles of excitatory amino acids during central nervous system development. Brain Res Brain Res Rev. 1990;15(1):41–70.

    Article  PubMed  Google Scholar 

  33. Cleland TA. Inhibitory glutamate receptor channels. Mol Neurobiol. 1996;13(2):97–136.

    Article  CAS  PubMed  Google Scholar 

  34. Watkins JC, Evans RH. Excitatory amino acid transmitters. Annu Rev Pharmacol Toxicol. 1981;21:165–204.

    Article  CAS  PubMed  Google Scholar 

  35. Duan X, et al. [Correlation between altered levels of neurotransmitters in the frontal lobe and hippocampus and behavioral abnormalities in a Clock(delta19) mutant mice modeling bipolar manic disorder]. Zhonghua Yi Xue Yi Chuan Xue Za Zhi. 2020;37(9):991–6.

    PubMed  Google Scholar 

  36. Amantea D, Bagetta G. Excitatory and inhibitory amino acid neurotransmitters in stroke: from neurotoxicity to ischemic tolerance. Curr Opin Pharmacol. 2017;35:111–9.

    Article  CAS  PubMed  Google Scholar 

  37. Liu X, et al. Targeted metabolomics analysis of serum amino acid profiles in patients with Moyamoya disease. Amino Acids. 2022;54(1):137–46.

    Article  PubMed  Google Scholar 

  38. Pei Z, et al. Inhibition of ferroptosis and iron accumulation alleviates pulmonary fibrosis in a bleomycin model. Redox Biol. 2022;57:102509.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Jiang X, Stockwell BR, Conrad M. Ferroptosis: mechanisms, biology and role in disease. Nat Rev Mol Cell Biol. 2021;22(4):266–82.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Tuo QZ, et al. Thrombin induces ACSL4-dependent ferroptosis during cerebral ischemia/reperfusion. Signal Transduct Target Ther. 2022;7(1):59.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Baba Y, et al. Protective effects of the mechanistic target of Rapamycin against excess iron and ferroptosis in cardiomyocytes. Am J Physiol Heart Circ Physiol. 2018;314(3):H659–68.

    Article  PubMed  Google Scholar 

  42. Stockwell BR, et al. Ferroptosis: A regulated cell death Nexus linking metabolism, redox biology, and disease. Cell. 2017;171(2):273–85.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Yang WS, et al. Regulation of ferroptotic cancer cell death by GPX4. Cell. 2014;156(1–2):317–31.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Liu W, et al. Dysregulated cholesterol homeostasis results in resistance to ferroptosis increasing tumorigenicity and metastasis in cancer. Nat Commun. 2021;12(1):5103.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Renga G, et al. Thymosin Β4 limits inflammation through autophagy. Expert Opin Biol Ther. 2018;18(sup1):171–5.

    Article  CAS  PubMed  Google Scholar 

  46. Sosne G, et al. Thymosin Beta 4: A potential novel therapy for neurotrophic keratopathy, dry eye, and ocular surface diseases. Vitam Horm. 2016;102:277–306.

    Article  CAS  PubMed  Google Scholar 

  47. Sosne G, et al. Thymosin-beta4 modulates corneal matrix metalloproteinase levels and polymorphonuclear cell infiltration after alkali injury. Invest Ophthalmol Vis Sci. 2005;46(7):2388–95.

    Article  PubMed  Google Scholar 

  48. Mutchnick MG, et al. Defective in vitro gamma interferon production and elevated serum immunoreactive thymosin beta 4 levels in patients with inflammatory bowel disease. Clin Immunol Immunopathol. 1988;47(1):84–92.

    Article  CAS  PubMed  Google Scholar 

  49. Zheng Z, et al. Thioesterase II of Escherichia coli plays an important role in 3-hydroxydecanoic acid production. Appl Environ Microbiol. 2004;70(7):3807–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Eschrig S, et al. LORE receptor homomerization is required for 3-hydroxydecanoic acid-induced immune signaling and determines the natural variation of immunosensitivity within the Arabidopsis genus. New Phytol. 2024;242(5):2163–79.

    Article  CAS  PubMed  Google Scholar 

  51. Mandal MK, et al. Glycerol-3-phosphate and systemic immunity. Plant Signal Behav. 2011;6(11):1871–4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Bruderer R, et al. Extending the limits of quantitative proteome profiling with data-independent acquisition and application to acetaminophen-treated three-dimensional liver microtissues. Mol Cell Proteom. 2015;14(5):1400–10.

    Article  CAS  Google Scholar 

  53. Huang da W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2009;37(1):1–13.

    Article  PubMed  Google Scholar 

  54. Rao G, Sui J, Zhang J. Metabolomics reveals significant variations in metabolites and correlations regarding the maturation of walnuts (Juglans regia L). Biol Open. 2016;5(6):829–36.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Heischmann S, et al. Exploratory metabolomics profiling in the Kainic acid rat model reveals depletion of 25-Hydroxyvitamin D3 during epileptogenesis. Sci Rep. 2016;6:31424.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  56. Sreekumar A, et al. Metabolomic profiles delineate potential role for sarcosine in prostate cancer progression. Nature. 2009;457(7231):910–4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Haspel JA, et al. Circadian rhythm reprogramming during lung inflammation. Nat Commun. 2014;5:4753.

    Article  CAS  PubMed  Google Scholar 

  58. Hanash S. Disease proteomics. Nature. 2003;422(6928):226–32.

    Article  CAS  PubMed  Google Scholar 

  59. Breiman L. Bagging predictors Machine Learning 24 (2), 123–140 (1996) 10.1023. A: 1018054314350, 1996.

  60. Deo RC. Machine learning in medicine. Circulation. 2015;132(20):1920–30.

    Article  PubMed  PubMed Central  Google Scholar 

  61. Senders JT, et al. Machine learning and neurosurgical outcome prediction: A systematic review. World Neurosurg. 2018;109:476–e4861.

    Article  PubMed  Google Scholar 

  62. Rigatti SJ. Random forest. J Insur Med. 2017;47(1):31–9.

    Article  PubMed  Google Scholar 

  63. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559.

    Article  PubMed  PubMed Central  Google Scholar 

  64. Pei G, Chen L, Zhang W. WGCNA application to proteomic and metabolomic data analysis. Methods Enzymol. 2017;585:135–58.

    Article  CAS  PubMed  Google Scholar 

  65. Huang Y, et al. Integrative metabolomics and proteomics allow the global intracellular characterization of Bacillus subtilis cells and spores. J Proteome Res. 2024;23(2):596–608.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Horgusluoglu E, et al. Integrative metabolomics-genomics approach reveals key metabolic pathways and regulators of Alzheimer’s disease. Alzheimers Dement. 2022;18(6):1260–78.

    Article  CAS  PubMed  Google Scholar 

  67. Schreiber A. MultiQuant™ Software Version 3.0.

Download references

Acknowledgements

We acknowledge individuals who have contributed to the research or manuscript preparation but do not meet all authorship criteria.

Funding

This study was supported by grants from the National Natural Science Foundation of China (grant numbers: 82171280 and 82201451).

Author information

Author notes

  1. Qingbao Guo, Manli Xie and Xiaopeng Wang contributed equally to this work.

Authors and Affiliations

  1. Department of Neurosurgery, XI’AN NO.9 HOSPITAL, Shaanxi, 710054, China

    Qingbao Guo

  2. Department of Occupational Diseases, Xi’an Central Hospital, Shaanxi, 710003, China

    Manli Xie

  3. Medical School of Chinese PLA, Beijing, 100039, China

    Xiaopeng Wang, Gan Gao & Jingjie Li

  4. Department of Neurosurgery, the Fifth Medical Centre, Chinese PLA General Hospital, Beijing, 100071, China

    Cong Han & Xiangyang Bao

  5. Department of Neurosurgery, the Eighth Medical Centre, Chinese PLA General Hospital, Beijing, 100089, China

    Qian-Nan Wang

  6. Department of Neurosurgery, the First Medical Centre, Chinese PLA General Hospital, Beijing, 100039, China

    Lian Duan

Authors
  1. Qingbao Guo
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Manli Xie
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Xiaopeng Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Cong Han
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Gan Gao
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Qian-Nan Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Jingjie Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Lian Duan
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Xiangyang Bao
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Qingbao Guo, Lian Duan, and Xiangyang Bao designed the study. Qingbao Guo, Manli Xie, and Xiaopeng Wang wrote the manuscript and performed the bioinformatics analysis. Cong Han, Gan Gao, Qian-Nan Wang, and Jingjie Li contributed to the manuscript discussion, figures, and tables. All the authors contributed to the manuscript and approved the submitted version.

Corresponding authors

Correspondence to Qingbao Guo, Lian Duan or Xiangyang Bao.

Ethics declarations

Ethical approval

This study was approved by the Institutional Review Board and Ethics Committee of the Fifth Medical Center of the Chinese PLA General Hospital (approval number: ky-2020-9-22). Informed consent was obtained from all participants.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Guo, Q., Xie, M., Wang, X. et al. Multi-omic serum analysis reveals ferroptosis pathways and diagnostic molecular signatures associated with Moyamoya diseases. J Neuroinflammation 22, 123 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12974-025-03446-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12974-025-03446-y

Keywords

Journal of Neuroinflammation

ISSN: 1742-2094