Serum and urine metabolic fingerprints enable diagnosis and prognosis for IgA nephropathy

Abstract

IgA nephropathy (IgAN), the most common primary glomerulonephritis worldwide, displays a pronounced geographical variation and progresses to end-stage renal disease within 20 years in 20–40 % of patients. However, current clinical management remains challenged by the invasive nature of kidney biopsy. To meet the urgent need for non-invasive strategies, we developed a dual-fluid metabolic profiling approach for early diagnosis and prognostic evaluation. By leveraging nanoparticle-enhanced laser desorption/ionization mass spectrometry (NPELDI-MS) to obtain metabolic fingerprints from serum and urine, combined with machine learning integration, our model achieved high diagnostic performance (AUC = 0.81–0.95) in distinguishing IgAN from healthy donors. Moreover, we tracked dynamic changes in key metabolites throughout treatment, revealing distinct metabolic pathways among different prognostic subgroups (p < 0.05). This study highlights the promising application of metabolic profiling as a non-invasive tool for precise management of IgAN.

Graphical abstract

IgA nephropathy (IgAN), the most common primary glomerulonephritis worldwide, displays a pronounced geographical variation and progresses to end-stage renal disease within 20 years in 20–40 % of patients. However, current clinical management remains challenged by the invasive nature of kidney biopsy. To meet the urgent need for non-invasive strategies, we developed a dual-fluid metabolic profiling approach for early diagnosis and prognostic evaluation. By leveraging nanoparticle-enhanced laser desorption/ionization mass spectrometry (NPELDI-MS) to obtain metabolic fingerprints from serum and urine, combined with machine learning integration, our model achieved high diagnostic performance (AUC = 0.81–0.95) in distinguishing IgAN from healthy donors. Moreover, we tracked dynamic changes in key metabolites throughout treatment, revealing distinct metabolic pathways among different prognostic subgroups. This study highlights the promising application of metabolic profiling as a non-invasive tool for precise management of IgAN.

1. Introduction

Chronic kidney disease (CKD), is characterized by persistent structural kidney abnormalities and/or impaired renal function, affecting over 850 million individuals globally [1,2]. Among the etiologies, IgA nephropathy (IgAN) stands out as the most prevalent primary glomerulonephritis worldwide, characterized by IgA1-dominant immune complex deposition in the glomerular mesangium, resulting in progressive renal inflammation, hematuria, proteinuria, and eventual kidney failure [3]. Globally, IgAN exhibits a striking geographical disparity, accounting for up to 40 % of primary glomerular diseases in East Asia compared to 20 % in Europe and 10 % in North America [[4], [5], [6]]. Current research indicates that IgA nephropathy occurs in roughly 25 to 50 people per million each year. Among these individuals, between 20 % and 40 % advance to end-stage kidney failure within a decade or two, even with conventional treatment [7].

The diagnosis and prognostication of IgAN remain heavily reliant on invasive renal biopsy, the current gold standard [8]. However, it is important to note that the procedure carries inherent risks, including bleeding complications in 1–40 % of cases [9], sampling variability leading to diagnostic inaccuracy in up to 30 % of cases, and an inability to serially monitor disease activity [[10], [11], [12]]. Non-invasive alternatives, such as urine protein-creatinine ratio (UPCR) or the estimated glomerular filtration rate (eGFR), lack specificity for IgAN and fail to differentiate between disease subtypes or to predict progression with any degree of accuracy [13]. Consequently, there is an urgent need for minimally invasive, sensitive, and specific tools for the precise diagnosis, subtyping, and risk stratification of IgAN [14].

Liquid biopsy, particularly metabolomic profiling, has emerged as a useful approach for dynamic disease monitoring, enabling non-invasive acquisition of multi-omics signatures critical for differential diagnosis and prognostic stratification [[15], [16], [17]]. Metabolites, as downstream outputs of cellular processes, offer direct insights into dysregulated pathways in IgAN, including aberrant glycosylation of IgA1, complement activation, and oxidative stress [18,19]. Crucially, serum metabolomics captures systemic perturbations, while urinary metabolomics directly reflects renal tissue alterations and tubular dysfunction [20,21]. Meanwhile, metabolic biomarkers promise to the non-invasive tools for the detection of IgAN, with sample collection protocols demonstrating a 90 % patient compliance rate [22], in comparison to the 60 % observed for renal biopsy [23]. Integrating these dual-fluid metabolic profiles presents a promising strategy to enhance diagnostic accuracy and prognostic precision in IgAN, thus overcoming limitations of single-biofluid analyses reported in earlier metabolomic studies.

As a preeminent tool in analytical science, mass spectrometry (MS) is central to the fundamental nature of molecular analysis, offering unparalleled sensitivity and mass resolution compared to conventional biochemical or spectroscopic techniques. Among advanced MS methodologies, nanoparticle-enhanced laser desorption/ionization mass spectrometry (NPELDI-MS) constitutes an approach for the comprehensive metabolic profiling of complex biofluids, including plasma, serum, and saliva, enabling detection of low-abundance metabolites down to the pmol level [[24], [25], [26]]. LDI MS enables efficient analysis of biofluids with minimal sample pretreatment (∼minutes/sample) and high sensitivity (∼pmol), by using defined matrix materials on the microarray chip to selectively trap metabolites [27,28]. Compared to conventional liquid/gas chromatography-MS (LC/GC-MS), NPELDI-MS demonstrates significant technical advantages in minimal sample preparation, ultra-low sample consumption, rapid analysis cycles, and superior throughput capacity [29]. Based on these, NPELDI-MS has been identified as a promising technology for clinical diagnosis and prognosis, particularly in IgAN.

Herein, we applied NPELDI-MS to acquire high-fidelity serum metabolic fingerprints (SMF) and urine metabolic fingerprints (UMF) from a well-defined IgAN cohort (Fig. 1) by solid ferric oxide particles prepared via a modified solvothermal method [30,31]. Using machine learning, we developed a dual-fluid metabolic model that accurately distinguished IgAN patients from healthy donors (HD), yielding an area under the curve (AUC) of 0.81–0.95. Furthermore, we monitored the dynamic disturbances of key metabolites during the process of therapy, uncovering distinct metabolic trajectories across prognostic subgroups (p < 0.05). Our work efforts could accelerate the development of metabolomic methods for coupled serum-urine samples, thereby laying a foundation for improving IgAN management strategies.

Fig. 1.

Schematic overview of biofluid-based metabolic profiling for early diagnosis and prognosis of IgA nephropathy (IgAN). A multi-step workflow was employed, beginning with a rigorous cohort design and sample collection, including serum and urine samples, followed by the sample loading in LDI MS to generate precise metabolic fingerprints. The data were then analyzed using advanced machine learning algorithms to achieve feature selection for identifying the key biomarkers, which was applied in IgAN early diagnosis and prognosis.

2. Result

2.1. Materials characteristics and study design

We developed the ferric NPELDI-MS platform for the analysis of clinical SMFs and UMFs, which could offer swift processing (∼20 s per specimen), requires only 1 μL of sample per test, and high throughput, accommodating 384 samples on one chip. To fabricate the highly efficient microarray, ferric nanoparticles (NPs) were synthesized using a cost-effective solution-thermal method based on our established protocols [32,33]. The resulting ferric NPs exhibited polycrystalline and surface-roughened morphologies (Fig. 2B), characteristics that support efficient laser desorption/ionization (LDI) of low-concentration metabolites. Further structure validation confirmed the high-quality polycrystalline nature of the NPs (Fig. 2C and Fig. S1A). Meanwhile, ferric NPs enhanced metabolite photo-thermal desorption from strong UV absorption (Fig. S1B) and had good stability [34] in the solutions (polydispersity index was 0.237, Fig. S1C), achieving a good linear relationship in the detection of metabolites (Pearson r > 0.9, Fig. S2). To evaluate the reproducibility of the NPELDI-MS platform for metabolite analysis, we calculated the coefficients of variation (CVs) for five representative metabolites, including alanine (Ala, 6.18 %), proline (Pro, 6.76 %), glutamic acid (Glu, 2.04 %), glucose (Glc, 4.10 %), and sucrose (Suc, 7.12 %). Using LDI-MS, all CVs yielded values < 10 %, demonstrating robust performance of NPELDI-MS (Fig. S3).

Fig. 2.

Materials characteristics and study design. (A) Digital image of the microarray achieving 384 samples per chip. The distance between two adjacent spots was 5 mm. (B) Scanning electron microscopy (SEM) image displayed surface cavities of nanoparticles, with a scale bar measuring 50 nm. (C) Transmission electron microscopy (TEM) image displayed the high-quality polycrystalline nature, with a scale bar measuring 50 nm. D) Study design of participant enrollment, including 200 healthy donors (HDs) and 102 IgAN patients. E) Longitudinal sample collection of 32 IgAN patients at three time points for IgAN prognosis. F) Typical mass spectra and intragroup similarity scores of the serum metabolic fingerprints (SMFs) in the m/z range of 100–500 Da, showed frequency distributions for both HD and IgAN groups. G) Typical mass spectra and intragroup similarity scores of the urine metabolic fingerprints (UMFs) in the m/z range of 100–500 Da, showed frequency distributions for both HD and IgAN groups.

Employing a case-cohort design, we recruited a total of 367 individuals, comprising 167 IgAN patients. Based on predefined inclusion and exclusion criteria (Figs. 2D), 102 IgAN participants were enrolled into the final study cohort (see Table S1). All participants in the healthy donor (HD) group provided informed consent and confirmed the absence of kidney disease (Table S1). Serum and urine samples were collected from both IgAN and HD groups for subsequent metabolic fingerprinting using the NPELDI-MS platform. We next followed up 32 IgAN patients and collected paired blood and urine samples every 3 months for disease prognosis and risk evaluation (Fig. 2E).

Representative mass spectra for serum and urine samples from one HD participant and one IgAN patient were presented in Fig. 2F–G. The raw mass spectra for each sample contained approximately 145,000 data points across the 100–500 Da mass to charge (m/z) range. To assess spectral consistency within groups, we calculated the similarity score for each serum and urine spectrum among participants within the HD and IgAN groups (Fig. 2F–G). Notably, a significant majority of samples (over 90 %) yielded strong concordance values (≥0.8), which serves to validate the assay's performance and indicates excellent spectral consistency within every group.

2.2. Diagnosis of IgAN via SMFs and UMFs

For establishing the SMF and UMF databases, a total of 427 and 441 unique m/z signals were obtained from the initial mass spectra, respectively. Each database was built through localizing features based on the highest intensity peak within each aligned m/z bin (Fig. 3A). To assess the diagnostic potential of SMFs and UMFs in distinguishing IgAN from HD groups, we conducted machine learning algorithms for model development. The discovery cohort (SMF: IgAN/HD = 175/150; UMF: IgAN/HD = 175/150) was used for model training and hyperparameter optimization via 10-fold cross-validation (Fig. 3B). An initial pilot experiment with 12 participants (6 IgAN/6 HD) informed a power analysis. The analysis forecasted statistical powers of 0.813 and 0.816 at a 10 % FDR threshold for cohorts of 300 (150/group) and 200 (100/group) using serum and urine assays, respectively, attesting to the findings' reliability (Fig. S4(A-B)). The reliability of the optimized models was subsequently evaluated using an independent validation cohort (SMF: IgAN/HD = 58/50; UMF: IgAN/HD = 57/48; Fig. 3B).

Fig. 3.

Diagnosis of IgAN via SMFs and UMFs. (A) SMF (purple) and UMF (blue) of 427 and 441 m/z features were extracted from the raw mass spectrometry data. The intensities were displayed using a color scale after logarithmic transformation. (B) Workflow of training and evaluation of classification models via machine learning. Sample-probability plots for HDs (green dots) and IgAN (blue dots) in (C) discovery cohort and (D) validation cohort via SMFs. Sample-probability plots for HDs (green dots) and IgAN (blue dots) in (E) discovery cohort and (F) validation cohort via UMFs.ROC curves of different machine learning algorithms for IgAN diagnosis in (G) discovery cohort and (H) validation cohort via SMFs. ROC curves of different machine learning algorithms for IgAN diagnosis in (I) discovery cohort and (J) validation cohort via UMFs.

Specifically, we implemented four supervised machine learning algorithms, including Ridge Regression (RR), Least Absolute Shrinkage and Selection Operator Regression (LR), Random Forest (RF), and Neural Network (NN), to analyze SMFs and UMFs for IgAN diagnosis. The diagnostic ‘score’ was designated as the algorithm's estimated likelihood of an individual having IgAN. Each cohort's ‘average score’ reflected the mean value of these predictions. Notably, across all four models in the discovery cohorts, the diagnostic scores were significantly higher (p < 0.001) in the IgAN group (average scores: 0.689–0.815) compared to the HD group (average scores: 0.176–0.353) (Fig. 3C) via SMFs. In the validation cohort, the diagnostic scores showed the same trend between the IgAN and HD groups (Fig. 3D). Similarly, based on the UMFs, the diagnostic scores for IgAN patients (average scores: 0.790–0.890) were significantly higher (p < 0.001) compared to the HD group (average scores: 0.096–0.245) (Fig. 3E–F). These results indicate that alterations in serum and urinary metabolic profiles of IgAN patients hold significant diagnostic potential.

In the discovery cohort, SMF-based models achieved areas under the receiver operating characteristic curve (AUCs) ranging from 0.84 to 0.91 (Fig. 3G). Consistent performance was observed in the validation cohort, with AUCs ranging from 0.85 to 0.95 (Fig. 3H), attesting to the superior generalization performance of our proposed models. Furthermore, receiver operating characteristic (ROC) evaluation was employed to confirm the diagnostic efficacy of our models, which achieved AUC values of 0.95–0.97 in the discovery cohort and 0.94–0.98 in the independent validation cohort (Fig. 3I–J). The high diagnostic performance observed for both SMF- and UMF-based models confirms that serum and urinary metabolite fingerprints exhibit excellent potential for IgAN diagnosis. Moreover, we performed permutation test yielding p < 0.001 (Fig S5), demonstrating the efficacy of the model in IgAN diagnosis without overfitting.

2.3. Feature selection for IgAN diagnosis

To enhance the predictive performance of the machine learning models, we implemented a feature selection strategy based on m/z signals. Starting from a pool of 427 and 441 m/z signals, molecular features were systematically selected using the coefficients derived from the LR algorithm. Crucially, the thresholds for these coefficients were dynamically optimized during the feature selection process to identify the most informative predictors. Subsequently, we filtered the corresponding metabolites by referencing both the SMF and UMF databases. Metabolites were retained based on stringent criteria: a mean intensity exceeding 800, a positive model score, and the p-value less than 0.05 (Fig. 4A).

Fig. 4.

Feature selection for IgAN diagnosis. (A) The illustration of feature selection via SMF and UMF by the criteria. The violin plot demonstrated differential expressions of the metabolic biomarkers between HDs and IgAN patients in (B) serum and (C) urine. *p < 0.05, **p < 0.01, ***p < 0.005. (D) ROC curves of the metabolic panel for discovery cohort and validation cohort to distinguish HDs from IgAN patients by serum biomarker panel. (E) ROC curves demonstrated the AUCs of each serum metabolic biomarker (AUC of 0.54–0.75). (F) ROC curves of the metabolic panel for discovery cohort and validation cohort to distinguish HDs from IgAN patients by the urine biomarker panel. (G) ROC curves demonstrated the AUCs of each urine metabolic biomarker (AUC of 0.70–0.87).

The approach yielded distinct diagnostic signatures for IgAN. Specifically, we identified five significant molecular features from the SMF database and four from the UMF database associated with IgAN diagnosis using Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FT-ICR-MS). The IgAN group exhibited significantly altered expression levels of glucose (Glc), lactic acid (LA), threonine (Thr), and valine (Val) based on SMF-identified metabolites (Fig. 4B–Table S2). Simultaneously, differential expressions were detected in the signal intensity of glucose (Glc), thymine (Thy), glycolic acid (GA), and urea via UMF-identified metabolites (Fig. 4C and Table S3).

We then evaluated the diagnostic efficacy of both individual biomarkers and a combined biomarker panel. Based on SMF-derived biomarkers, the combined biomarker panel demonstrated substantially improved diagnostic power, achieving an AUC of 0.85 (95 % confidence intervals (CI): 0.77–0.87) in the discovery cohort and 0.81 (95 % CI: 0.72–0.89) in the validation cohort (Fig. 3D). In contrast, individual SMF features showed limited diagnostic utility, with AUC values ranging only from 0.54 to 0.75 (Fig. 3E). In terms of UMF-derived biomarkers, the combined UMF biomarker panel showed exceptional diagnostic performance, achieving an AUC of 0.95 (95 % CI: 0.93–0.97) in the discovery cohort and 0.93 (95 % CI: 0.88–0.98) in the validation cohort (Fig. 3F). Although higher than their SMF counterparts, individual UMF features also had limited AUCs, ranging from 0.70 to 0.87 (Fig. 3G). These findings reveal that both serum and urine metabolic signatures capture non-overlapping aspects of IgAN pathophysiology, offering parallel avenues for biomarker development.

2.4. Enrichment analysis and prognosis of IgAN treatment

For the metabolic enrichment analysis, we conducted pathway enrichment analysis of serum and urinary metabolites using MetaboAnalyst 5.0 (https://www.metaboanalyst.ca/). Seven serum-derived pathways met the significance thresholds (enrichment ratio >0, hit number ≥1): 1) Valine, leucine and isoleucine biosynthesis, 2) Neomycin, kanamycin and gentamicin biosynthesis, 3) Starch and sucrose metabolism, 4) Pantothenate and CoA biosynthesis, 5) Galactose metabolism, 6) Glycine, serine and threonine metabolism, and 7) Valine, leucine and isoleucine degradation via serum biomarkers (Fig. 5A and Table S4). Based on the urine biomarkers, six urinary pathways demonstrated significance, including: 1) Neomycin, kanamycin, and gentamicin biosynthesis, 2) Arginine biosynthesis, 3) Starch and sucrose metabolism, 4) Galactose metabolism, 5) Pyrimidine metabolism, 6) Purine metabolism (Fig. 5B and Table S5).

Fig. 5.

Enrichment analysis and prognosis of IgAN treatment. Pathway enrichment analysis of biomarker panels via (A) serum and (B) urine. (C) Longitudinal tracking of serum biomarkers revealed distinct prognostic patterns, including lactic acid, threonine, glucose, and valine. (D) Longitudinal tracking of urinary biomarkers revealed distinct prognostic patterns, including glucose, glycolic acid, urea, and thymine. *p < 0.05, **p < 0.01, ***p < 0.005.

Specifically, in valine metabolism, IgAN patients frequently exhibit intestinal dysbiosis, which disrupts valine synthesis and metabolism (e.g., through microbial-derived metabolites interfering with host amino acid metabolism) [35]. The metabolic disturbance, coupled with renal cell injury, forces valine to participate in energy supply via the gluconeogenesis pathway, accelerating its consumption [36,37]. In the neomycin, kanamycin, and gentamicin biosynthesis, aminoglycoside antibiotics influence the progression of IgAN through a tripartite mechanism encompassing direct nephrotoxic injury, dysregulation of the gut microbiota-immune axis, and potential disruption of IgA1 glycosylation [38]. In terms of galactose metabolism, disordered galactose metabolism drives IgAN progression through a cascade initiated by C1GALT1/Cosmc dysfunction, leading to elevated Gd-IgA1 synthesis, immune-complex deposition, and subsequent glomerular inflammation [39,40].

Longitudinal tracking of serum and urinary biomarkers provides valuable insights into the prognostic trajectories of IgAN progression (Fig. 5C). In particular, lactic acid and valine levels gradually decreased in the clinical improvement group (better group), while they initially decreased and subsequently increased in the disease progression group (worse group). Threonine and glucose levels decreased in the better group, but remained stable in the worse group. These findings indicate the potential of serum biomarkers in predicting disease course and guiding personalized therapeutic interventions in IgAN.

For the urinary biomarkers, both glycolic acid and urea demonstrated parallel trajectories (Fig. 5D). In the worse group, their levels initially underwent a pronounced decline followed by stabilization at a sustained plateau. Conversely, the better group showed an initial marked elevation in metabolite concentrations, which subsequently declined to baseline levels. Although glucose levels increased in both groups, the trajectories were statistically distinct (p < 0.05). Conversely, thymine profiles demonstrated convergence between the improvement and progression groups during therapy, showing no significant divergence. These integrated findings of key metabolite dynamics, pathway perturbations, and prognostic trajectories reveal actionable targets for elucidating IgAN progression mechanisms, developing molecular therapies, and designing early clinical interventions.

3. Discussion

Early diagnosis and prognostic assessment of IgAN remain major global healthcare challenges. Compared with conventional biomarkers, renal biopsy, or some new biomarkers such as Gd-IgA1 [41] and urinary miRNA panels [42], our metabolomics platform offers a non-invasive, early diagnostic approach that captures systemic metabolic alterations reflecting immune activation and oxidative stress proceeding overt renal injury. This multidimensional profiling enables both disease diagnosis and prognosis stratification, complementing current clinical assessments. Biomarker discovery in renal diseases increasingly spans genomic, proteomic, and metabolomic levels, with biofluid-derived markers demonstrating growing diagnostic and prognostic utility [43,44]. However, upstream nucleic acid and protein biomarkers, though moderately discriminative (AUC = 0.8) in IgAN using signal-amplification assays, remains limited by analytical complexity [10,45]. Metabolomic biomarkers, as functional downstream products, provide real-time phenotypic insights with greater clinical feasibility. While previous studies have suggested diagnostic potential for metabolic markers in IgAN, conclusive validation is lacking. Our rigorous study design, which includes appropriate biofluid selection and cohort characterization, is fundamental for deriving precise diagnostic insights. By employing feature selection refined by machine learning, we achieved enhanced diagnostic discrimination (AUC 0.81–0.95) and significant metabolic differences between prognosis groups (p < 0.05). These findings highlight the translation potential of our integrated platform—combining appropriate biofluid selection, advanced computational methods, and robust cohort design—for clinical IgAN management.

For biofluid selection, blood testing remains the most routine, standardized, and rapid approach for assessing hepatic and renal function. Compared with tissue biopsy, blood collection causes minimal trauma, resulting in less pain, lower risk, and high patient compliance. Urine sampling, as a minimally invasive alternative, is well suited for large-scale IgAN studies [46]. Emerging evidence links urinary metabolic profiles to alterations in glycolysis, tricarboxylic acid (TCA) cycle dynamics, and amino/fatty acid metabolism [47,48]. Accordingly, our study employed paired serum-urine sampling and demonstrated that integrated biofluid analysis achieved good performance. To our knowledge, this represents the first large-scale metabolomic validation showing the advantage of combined biofluids over single-source (serum or urine) approaches in IgAN.

Previous metabolomic studies on IgAN have been limited by relatively small cohorts, typically involving 50 to 200 participants [49]. In contrast, our study included 302 well-charaterized individuals under standardized collection protocols. To distinguish IgAN patients from HDs, we analyzed 433 serum samples and 430 urine samples using NPELDI-MS. For prognostic evaluation, longitudinal metabolic monitoring was conducted with 96 paired serum-urine samples collected at three time points (every three months) during treatment of 32 IgAN patients. Machine learning-based variable selection and predictive modeling achieved AUC values exceeding 0.8. These results highlight the value of our metabolic platform for IgAN management and underscore its potential for broad clinical application through optimized biofluid selection, advanced analytics, and robust cohort design.

Finally, we identified the diagnostic biomarker panels comprising four serum metabolites (glucose, lactic acid, threonine, and valine) and four urinary metabolites (glucose, thymine, glycolic acid, and urea). Pathway analysis was subsequently performed to elucidate the potential relationships between these metabolites and IgAN. Consistent with our findings, several of the implicated pathways have been reported in previous studies [50]. The development of pathway-based analyses and prognostic trajectories is essential for accurately identifying high-risk patients and guiding therapeutic decision-making. Although liquid biomarkers such as miRNAs and cfDNA/ctDNA have demonstrated prognostic value for survival outcomes [49,51], our study introduces metabolic dynamics as a novel class of prognostic indicators. The SMF-UMF prediction model, based on NPELDI-MS technology, enables real-time and large-scale risk stratification, providing a practical framework to improve IgAN management through early intervention.

Nevertheless, several limitations should be acknowledged. The single-center design may limit generalizability, and the biological significance of the observed metabolic alterations requires further investigation. Future multicenter studies with larger and more diverse cohorts are needed to validate these findings, and integration with genomic, proteomic, and clinical data could further refine predictive modeling for precision medicine in IgAN. Advances in analytical sensitivity, such as engineered nanostructures in NPELDI-MS, may enable population-scale screening and longitudinal monitoring. Regarding clinical translation, implementing this platform in routine practice warrants consideration. Although the workflow involves multi-omics integration and computational modeling, advances in high-throughput technologies and cloud-based analytics have reduced cost and turnaround time. With further standardization and automation, the platform could be scaled for real-time monitoring in larger cohorts. Future studies should assess cost-effectiveness, workflow optimization, and integration into diagnostic pipelines to support practical application.

CRediT authorship contribution statement

Beili Wang: Writing – original draft, Investigation, Funding acquisition, Data curation, Conceptualization. Ruimin Wang: Writing – original draft, Investigation, Data curation, Conceptualization. Wenqi Shao: Writing – original draft, Investigation, Data curation, Conceptualization. Baishen Pan: Visualization, Software, Data curation. Xiaoqiang Ding: Validation, Software, Methodology, Data curation. Juxiang Zhang: Visualization, Software, Methodology, Data curation. Yanxi Yang: Validation, Software, Methodology, Data curation. Yiqin Shi: Visualization, Software, Funding acquisition, Data curation. Jiao Wu: Writing – review & editing, Writing – original draft, Supervision, Methodology, Investigation, Funding acquisition, Conceptualization. Wei Guo: Writing – review & editing, Writing – original draft, Supervision, Methodology, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors have filed patents for both the technology and the use of the technology to detect bio-samples.

Appendix A. Supplementary data

The following is the supplementary data to this article:

Data availability

Data will be made available on request.