Skip to main content
NCBI home page
Springer logoLink to Springer
. 2026 Feb 27;22(2):30. doi: 10.1007/s11306-026-02394-9

Comparative urinary metabolomics reveals unique and shared pathways in COVID-19 and liver diseases

Garima Juyal 1,✉, Fariya Khan 2,3, Sidra Siddiqui 2, Ayyub Rehman 2, Arumugam Madhumalar 2, Neda Mirsamadi 4, Gagan Deep Jhingan 4, Chhagan Bihari 5, Mohan Chandra Joshi 2,✉
PMCID: PMC12948825  PMID: 41758429

Abstract

Introduction

The COVID-19 pandemic and liver diseases both cause significant metabolic disturbances, yet the specific mechanisms driving these changes remain poorly understood.

Objectives

This study aimed to elucidate and compare the urinary metabolomic profiles of COVID-19 patients, individuals with liver diseases, and healthy controls to determine common and unique metabolic signatures between diseases.

Methods

Untargeted metabolomic profiling was performed using liquid chromatography–mass spectrometry (LC-MS) on urine samples from COVID-19 patients (n = 102), liver disease patients (n = 100), and healthy controls (n = 101). Differential metabolite abundance, pathway enrichment, network topology, and Random Forest–based machine learning analyses were performed.

Results

Both COVID-19 and liver disease exhibited extensive metabolic reprogramming. COVID-19 patients showed suppression of Vitamin B6 and purine metabolism, indicating impaired energy production and antioxidant defense. Liver disease patients exhibited reduced primary bile acid biosynthesis and pantothenate metabolism, reflecting hepatic dysfunction. Random Forest models robustly discriminated disease from healthy states, with binary models for COVID-19 and liver disease achieving AUCs of 0.998, and a multiclass model distinguishing all three groups with 91.9% accuracy. Both conditions shared perturbations in amino acid and steroid-related pathways, reflecting common systemic stress. Importantly, unique metabolites, such as N-Acetylvaline, Succinyladenosine, and S-adenosylhomocysteine in COVID-19, and 3-Hydroxysebacic acid, Asn-Trp, and bile acid derivatives in liver disease, emerged as highly specific biomarkers, highlighting systemic viral stress versus chronic hepatic metabolic adaptation and warranting future validation.

Conclusion

These findings enhance our understanding of disease-specific metabolic remodelling and point to potential biomarkers for diagnosis and therapeutic targeting.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11306-026-02394-9.

Keywords: COVID-19, Urine metabolomics, LC-MS, Liver disease, Vitamin B6 pathway

Introduction

Coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), has resulted in a global health crisis, with more than 523 million infections and over 6.2 million deaths worldwide (Zhou et al., 2020). Although initially characterized as a respiratory illness, COVID-19 is now recognized as a systemic disease affecting multiple organs, including the cardiovascular, neurological, renal, and hepatic systems (Liu et al., 2020; Niazkar et al., 2020; Richardson et al., 2020). Clinical outcomes range from asymptomatic infection to severe, life-threatening disease, with increased risk observed in elderly individuals and patients with comorbidities such as diabetes, obesity, and cardiovascular disorders (Mueller et al., 2020). Metabolic dysregulation has emerged as a central contributor to COVID-19 disease heterogeneity, severity, and long-term sequelae (Oropeza-Valdez et al., 2024). SARS-CoV-2 infection induces widespread alterations in immune metabolism, lipid homeostasis, amino acid utilization, and mitochondrial function, collectively influencing host immune responses and tissue damage (Sánchez et al., 2023). Consequently, metabolomics has become a powerful approach for identifying disease-associated biomarkers and for elucidating host–pathogen interactions in COVID-19. However, a major challenge in COVID-19 metabolomics-based biomarker discovery is the substantial overlap of metabolic alterations with other chronic diseases, particularly liver disorders such as chronic hepatitis, cirrhosis, and non-alcoholic steatohepatitis (NASH) (Buchynskyi et al., 2023; Imam et al., 2023). Liver diseases are themselves characterized by profound disruptions in amino acid metabolism, lipid metabolism, mitochondrial energy production, and gut–liver axis signaling pathways that are also commonly perturbed in COVID-19. As a result, many reported metabolic biomarkers lack disease specificity, limiting their translational and diagnostic utility.

Urine metabolomics offers a unique opportunity to address this challenge. Beyond its non-invasive nature, urine reflects the cumulative output of systemic metabolism, hepatic detoxification, renal filtration, and gut microbial co-metabolism. Importantly, urine captures downstream metabolic end-products that integrate liver function, immune activation, and oxidative stress, making it particularly suitable for differentiating metabolic signatures arising from viral infection versus intrinsic hepatic dysfunction. While numerous studies have characterized urinary metabolic alterations in COVID-19, including disruptions in amino acid metabolism, lipid metabolism, energy pathways, and immune-regulatory processes such as the tryptophan–kynurenine axis (Badawy, 2023; Tsuji et al., 2023; Ansone et al., 2024), similarly metabolomic perturbations have also been reported in liver diseases, including elevated kynurenic acid, sphinganine, indole derivatives, and altered purine metabolism (Tan et al., 2025). Despite these advances, most metabolomics studies have examined COVID-19 and liver diseases independently or relative only to healthy controls. Direct, comparative urine metabolomic analyses between COVID-19 and liver disease patients are limited, particularly in population-specific cohorts. Consequently, it remains unclear which urinary metabolic alterations are disease-specific and which reflect shared systemic inflammation, mitochondrial dysfunction, or immune–metabolic stress. This lack of comparative frameworks represents a critical gap that hinders the identification of specific, non-invasive biomarkers capable of distinguishing infectious disease from chronic hepatic pathology.

In this study, we address this gap by performing discovery-phase LC–MS-based urine metabolomic profiling in COVID-19 patients, liver disease patients, and healthy controls from an Indian population. By directly comparing these groups, we aimed to (i) identify shared and disease-specific metabolic signatures, (ii) evaluate the utility of urine metabolomics in distinguishing COVID-19 from chronic liver disease, and (iii) advance the discovery of non-invasive, population-relevant metabolic biomarkers. This comparative approach provides mechanistic insight into overlapping and distinct metabolic pathways underlying infectious and chronic hepatic conditions, while strengthening the translational potential of urine-based metabolomics for clinical biomarker development.

Materials and methods

Study design

Urine samples from both COVID-19 and liver disease patients were obtained from the ILBS biobank (SI Table-1). The COVID-19 samples were collected from individuals hospitalized in New Delhi hospitals between January 2021 and December 2022 and subsequently stored at the ILBS biobank. Diagnosis of COVID-19 was confirmed by RT-PCR of nasopharyngeal/oropharyngeal swabs following Indian Council of Medical Research (ICMR) guidelines. Adults (≥ 18 years) with PCR-confirmed SARS-CoV-2 infection and no pre-existing chronic liver or kidney disease were included. Urine specimens were obtained within the first week of hospitalization, consistent with ILBS protocols used in the COPLA-II and ICAM1 studies, to reduce variability related to disease progression or treatment (Bajpai et al., 2022; Kaur et al., 2024). Urine samples from patients with liver cirrhosis, non-alcoholic fatty liver disease (NAFLD), acute liver failure, chronic hepatitis, or jaundice were collected during the same period and archived under comparable conditions. Liver disease was diagnosed using standard EASL criteria based on clinical, biochemical, and imaging findings (European Association for the Study of the Liver, 2018). Patients with concurrent SARS-CoV-2 infection, renal impairment, or systemic infections were excluded. The study was approved by the Institutional Ethics Committee, Jamia Millia Islamia (IEC No. 1/10/290/JMI/IEC/2020), and written informed consent was obtained from all participants and principle investigator of COPLA-II trial and ICAM-1 study. The COVID-19 samples originated from hospitalized patients representing moderate-to-severe disease. The healthy control samples were from the biobank (National Liver Disease Biobank) where we already have the cohort which was duly approved by ICE. All urine samples were collected, processed, and stored under identical conditions to ensure consistency and minimize pre-analytical variability.

Urine sample storage and processing

All samples were collected aseptically, centrifuged, aliquoted, and stored at − 80 °C in the ILBS biobank until analysis. Metabolite extraction and cleanup were performed concurrently for the entire cohort, thereby minimizing variability associated with solvent composition, reagent stability, and operator handling. To prevent degradation or storage-induced artifacts (e.g., oxidation, precipitation, or freeze–thaw effects), no extracted metabolites were stored; instead, each sample was injected directly onto the LC–MS/MS system immediately following preparation.

Liquid chromatography mass spectrometry analysis (LC/MS)

For metabolite extraction, 200 µL of urine was mixed with 200 µL of acetonitrile, vortexed for 30 s, and centrifuged at 14,000 × g for 10 min. The supernatant was vacuum-dried, and the resulting residue was reconstituted in 200 µL of 2% acetonitrile. All LC–MS experiments were performed on a Dionex Ultimate 3000 UHPLC system (Thermo Fisher Scientific, MA, USA) equipped with a binary solvent delivery manager and autosampler, coupled to a Thermo Orbitrap Exploris 240 mass spectrometer (Thermo Fisher Scientific, Sunnyvale, CA, USA) operated in both positive (ESI⁺) and negative (ESI⁻) electrospray ionization modes in separate runs. Chromatographic separation was achieved using solvent A (water with 0.1% v/v formic acid) and solvent B (acetonitrile with 0.1% v/v formic acid) under the following gradient program at a constant flow rate of 0.35 mL/min: 0 min (2% B), 0–2 min (2% B), 2–5 min (2–20% B), 5–9 min (20–100% B), 9–13 min (100% B), 13–13.1 min (2% B), and 13.1–16 min (2% B). The injection volume was 10 µL, yielding a total run time of 16 min per sample. Mass spectrometric data acquisition was carried out under standardized conditions to ensure reproducibility. MS1 survey scans were collected across an m/z range of 100–1,500 at a resolution of 120,000 (at m/z 200). MS2 scans were acquired at a resolution of 30,000 (at m/z 200) with a normalized AGC target of 50% and auto injection time. Precursor ions were isolated with a 2.0 m/z isolation window and fragmented using higher-energy collisional dissociation (HCD) with normalized collision energy set at 35%. The ion source was operated with a spray voltage of 4 kV(+) and 3 kV (-) and a capillary temperature of 275 °C, while sheath and auxiliary gas flows were maintained at 30 and 10 arbitrary units, respectively.

For analytical QC, we employed a pooled quality-control (QC) strategy, which was generated by combining equal aliquots from all study samples, was injected periodically throughout the LC–MS/MS acquisition sequence. These QC runs enabled continuous assessment of mass accuracy, retention time stability, and signal intensity drift. Within Compound Discoverer (Thermo Scientific, v3.2), pooled QC data were integrated into the Retention Time Alignment and Compound Detection nodes to improve feature alignment across all samples. Furthermore, the Normalization node utilized pooled QC responses to correct for signal drift and systematic variability during batch acquisition. The performance of QC injections was additionally monitored in the Feature Grouping and Statistical Analysis nodes, ensuring that only robust and reproducible features were retained for downstream differential analysis.

The raw data files generated were analyzed using Compound Discoverer (CD) 3.3. For compound consolidation, a mass tolerance of 5 ppm and a retention time (RT) tolerance of 0.2 min were applied. The Metabolite annotation was performed in accordance with the Metabolomics Standards Initiative (MSI) framework. As authentic reference standards were not employed in this study, no features could be assigned as MSI Level 1. Instead, annotations were classified as MSI Level 2 when features matched both accurate mass (within a 5-ppm tolerance) and chromatographic retention behaviour with database-reported values, and supporting MS/MS spectra were consistent with library fragmentation patterns. Features annotated solely on the basis of accurate mass matching were designated as MSI Level 3. Annotation was supported by searches against multiple established repositories, including the Human Metabolome Database (HMDB) within Compound Discoverer. Both positive (ESI+) and negative (ESI−) ionization modes were analyzed separately, and the results were combined at a later stage, with duplicates removed to ensure data integrity.

Data analysis

The complete dataset comprising 303 urine samples (SI Table-2) was processed using MetaboAnalystv6.0 (www.metaboanalyst.ca) (Pang et al., 2024). Data preprocessing involved filtering out 149 variables based on an interquartile range (IQR) threshold of 25% and a low-abundance filter (mean intensity = 0%) to remove non-informative features. The remaining data were normalized by median and autoscaling (mean-centering followed by division by the standard deviation) to ensure comparability across samples. Both univariate and multivariate statistical analyses were performed to identify and visualize metabolomic variations between study groups. Multivariate analysis included Principal Component Analysis (PCA) for unsupervised clustering and hierarchical clustering using Euclidean distance and Ward’s linkage to assess group separation and metabolite-level relationships. Enrichment analysis (Over representation analysis) and network enrichment analysis of significantly upregulated and downregulated metabolites (FDR p < 0.1, log2FC > 2.0) between comparative groups (COVID-19 vs. healthy and Liver vs. Healthy) performed using default setting of MetaboAnalystv6.0. In network topology analysis enabled identification of hub network and interconnecting metabolites.

Random forest analysis

Random Forest (RF), an ensemble-based algorithm that constructs multiple independent decision trees on bootstrapped subsets of the data and aggregates their predictions (majority vote for classification), was used to minimizes overfitting and variance (Breiman, 2001). Random Forest Classifier (RFC) from scikit-learn (sklearn.ensemble.RandomForestClassifier) was used as a supervised ensemble learning method to classify metabolomic LC–MS intensity profiles. Normalized LC–MS metabolomic datasets representing each of the three groups were used as an input for RF classifier. The 2/3 of dataset (67%) was split as training subset, with the remaining 1/3 kept for testing (33%), to preserve the class balance. Number of decision trees in the forest were set to 500, which ensured ensemble stability and convergence. Each tree was trained on a random sample (with replacement) and diversity was ensured by enabling the bootstrap sampling. Out-of- Bag validation was activated to allow internal accuracy estimation. For feature subsampling strategy, at each split, only √N features (N- represents the number of metabolites) were considered, reducing correlation among trees. An OOB accuracy of 0.98 (error = 0.02) indicated minimal overfitting and robust internal validation. Further, ROC curves were generated to show the performance of the RF model. Each class-specific area under the ROC curve (AUC) exceeded 0.95, confirming excellent discriminative power. An accuracy of 97% and a high F1-score for the Covid and Liver datasets indicated the good performance of RF model. To enhance interpretability, SHapley Additive exPlanations (SHAP) values were computed. It further transformed the “black-box” ensemble into an interpretable biological framework, allowing direct mapping of statistical model outputs onto biochemical signatures. The SHAP framework quantifies each feature’s marginal contribution to the prediction probability for every class, thus, estimating the values corresponding to the contribution from each metabolite for the model prediction (Lundberg & Lee, 2017).

Results & discussion

The COVID-19 pandemic has caused global disruption, underscoring the urgent need for non-invasive biomarkers that can enable rapid diagnosis and effective disease monitoring (Wilson & Forse, 2023). While the pulmonary and immunological effects of SARS-CoV-2 infection are well documented, recent studies have also revealed substantial metabolic alterations in affected individuals, as evidenced by disrupted metabolite profiles (Oropeza-Valdez et al., 2024). These metabolic perturbations can vary across populations due to genetic, ethnic, and environmental factors and often overlap with other conditions, such as liver disease (Buchynskyi et al., 2024; Imam et al., 2023b). This overlap complicates the identification of COVID-19-specific biomarkers.In this study, we performed untargeted discovery metabolomics (LC-MS) to identify differentially abundant metabolites in each disease group compared to healthy controls, as well as between the two disease states. Subsequently, pathway enrichment and network analyses were conducted to determine under- and over-represented pathways. Using machine learning approaches, we identified unique metabolic signatures characteristic of each group (Fig. 1). Log2-transformed metabolite abundances showed consistent distributions across all groups, confirming the uniform data quality (SI Fig. 1). Dual hierarchical clustering revealed clear separation of samples across the three groups in both ionization modes, indicating distinct group-specific metabolic patterns (SI Fig. 2).

Fig. 1.

Fig. 1

Open in a new tab

Study design: Workflow illustrating the urine metabolite analysis conducted in COVID-19 patients, patients with liver disease, and healthy controls. The study includes sample collection, metabolite extraction, data acquisition using metabolomics platforms, statistical analysis, biological pathway and network analysis followed by biomarker discovery

Fig. 2.

Fig. 2

Open in a new tab

Urinary metabolite profiling in COVID-19 patients compared to healthy controls: A Volcano plot showing differentially expressed urinary metabolites based on fold change and statistical significance (FDR p-value); B Principal Component Analysis (PCA) plot demonstrating separation between COVID-19 patients and healthy controls based on global metabolite profiles; C, D Bar plots of significantly under- and over-represented biological pathways enriched in COVID-19 patients; and E, F Network visualizations of under- and over-represented metabolic pathways. Single node represents a metabolite (purple dots), while (Pink square nodes) edge represents biological relationships or interactions between nodes (e.g., shared pathways, biochemical reactions, or correlation-based associations). A single metabolite (Star-shaped nodes) connected with multiple metabolites represent key hub or connector metabolites

COVID-19 patients exhibit systemic metabolic reprogramming reflecting immune response and energy adaptation

Univariate analysis of urine metabolite profiles from 102 COVID-19 patients and 101 healthy controls identified widespread metabolic alterations. Of the 445 detected metabolites, 224 were significantly upregulated and 143 downregulated in COVID-19 patients, while 78 showed no significant difference (FDR p < 0.1; Fig. 2A). Principal component analysis (PCA) demonstrated clear separation between COVID-19 patients and healthy controls, with PC1 and PC2 explaining 17.0% and 9.8% of the total variance, respectively. This group separation was statistically supported by PERMANOVA (R² = 0.50, p = 0.001), indicating that disease status accounts for a substantial proportion of metabolic variance (Fig. 2B). Dual hierarchical clustering in both positive and negative ionization modes further confirmed distinct metabolic profiles between the two groups (SI Fig. 1).

Pathway enrichment analysis revealed coordinated suppression of several host metabolic pathways in COVID-19 patients, including vitamin B6 metabolism (p = 5 × 10− 02), purine metabolism (p = 1 × 10− 01), the pentose phosphate pathway, pantothenate and CoA biosynthesis, and sphingolipid metabolism (Fig. 2C). Reduced vitamin B6 metabolism has been previously associated with immune dysregulation, inflammation, and hepatic stress in COVID-19 (Masoodi et al., 2022; Imam et al., 2023). The observed downregulation of purine metabolism and the uric acid network contrasts with some earlier reports but is consistent with altered nucleotide turnover and energy imbalance during systemic infection (Li et al., 2022; Valdés et al., 2022). Suppression of the pentose phosphate pathway and CoA biosynthesis suggests impaired redox balance and energy metabolism, while reduced sphingolipid metabolism may reflect altered lipid-mediated signaling and inflammatory regulation (Khan et al., 2021; Paneque et al., 2023). Decreased S-adenosylhomocysteine further indicates disruption of methylation homeostasis, which has been linked to vascular inflammation (Azzini et al., 2020). In contrast, several pathways were significantly enriched in COVID-19 patients, most notably biotin metabolism (p = 2 × 10− 03) (Fig. 2D). Biotin is a key cofactor in fatty acid oxidation, amino acid catabolism, and gluconeogenesis, and its enrichment may represent a compensatory metabolic response to increased energetic demands during infection (Hernández-Flores et al., 2022). Elevated ubiquinone and terpenoid-quinone biosynthesis further point to mitochondrial adaptation, potentially supporting antioxidant defense through increased CoQ10 availability (Guarnieri et al., 2024).

Network enrichment analysis of downregulated metabolites (fold change < 2) revealed tightly connected clusters centered on uric acid, xanthine, S-adenosylhomocysteine, pseudouridine, and sphinganine, corresponding primarily to purine metabolism, nucleotide turnover, and methylation-related processes (Fig. 2E). Conversely, upregulated metabolites (fold change > 2) formed interconnected networks involving amino acids, energy-related metabolites, and cofactors, including L-methionine, L-histidine, L-tyrosine, pyroglutamic acid, biotin, creatine, and propionylcarnitine (Fig. 2F). These contrasting network topologies indicate a global metabolic shift away from nucleotide and purine pathways toward amino acid, cofactor, and energy-supporting processes, consistent with adaptive host responses to infection (Shen & Wang, 2021).

To focus on biologically meaningful host-derived alterations, metabolites suspected to originate from dietary, environmental, or chemical sources were excluded prior to model construction. Despite this filtering, the remaining endogenous metabolite set (n = 129) retained strong discriminatory power. Random Forest (RF) classification achieved excellent performance across feature subsets (AUC range: 0.995–0.999; Fig. 3A). An optimized model based on the top 15 metabolites yielded an AUC of 0.998 (95% CI: 0.982–1.000.982.000), demonstrating that a compact panel of endogenous metabolites robustly distinguishes COVID-19 patients from healthy controls (Fig. 3B, C). The most informative metabolites mapped predominantly to purine and one-carbon metabolism (e.g., succinyladenosine, S-adenosylhomocysteine), amino acid and peptide metabolism (e.g., N, N-dimethylarginine, acetyl-L-carnosine), and cofactor- and hormone-related pathways (e.g., biocytin, 16α-hydroxyestrone) (Steinberg & Cedergren, 1995; Yi et al., 2000; McCann et al., 2007; Obi et al., 2011; Fan et al., 2025). Collectively, these metabolites reflect coordinated disruptions in nucleotide turnover, methylation balance, amino acid metabolism, and energy regulation. Importantly, pathway-level convergence rather than individual metabolite effects underpins the strong classification performance, highlighting systemic metabolic reprogramming as a defining feature of COVID-19.

Fig. 3.

Fig. 3

Open in a new tab

Metabolite signatures distinguishing COVID-19 from healthy controls. A Multivariate ROC curves showing RF model performance across increasing feature numbers. B Variable-importance ranking of the top 15 metabolites contributing to classification. C Box-and-whisker plots depicting relative abundance differences of the most discriminative metabolites between COVID-19 patients and healthy controls

Distinct metabolic signatures in liver disease reveal adaptive and pathological responses to hepatic dysfunction

Differential metabolite analysis of urine samples from 100 liver disease patients and 101 healthy controls revealed pronounced metabolic alterations. Of the 367 metabolites detected, 153 were significantly upregulated and 156 were significantly downregulated in liver disease patients (fold change ≥ 2, FDR-adjusted p < 0.1), while 130 metabolites showed no significant change (Fig. 4A). Principal component analysis (PCA) demonstrated clear separation between liver disease patients and healthy controls along PC1 and PC2, which accounted for 15.4% and 9.7% of the total variance, respectively (Fig. 4B). Liver disease samples clustered more tightly than healthy controls, indicating relatively homogeneous metabolic perturbations within the disease group. PERMANOVA analysis confirmed the statistical significance of this separation (R² = 0.51, p = 0.001). Hierarchical clustering further supported the presence of disease-specific metabolic profiles (SI-Fig. 1).

Fig. 4.

Fig. 4

Open in a new tab

Urinary metabolite profiling in patients with liver disease compared to healthy controls: A Volcano plot displaying differentially expressed urinary metabolites based on fold change and statistical significance (adjusted p-value); B Principal Component Analysis (PCA) plot illustrating the separation of metabolite profiles between liver disease patients and healthy controls; C, D Bar plots showing significantly under- and over-represented biological pathways enriched in liver disease patients; and E, F Network visualization of significantly altered metabolic pathways, highlighting under- and over-represented metabolic networks. Single node represents a metabolite (purple dots), while (Pink square nodes) edge represents biological relationships or interactions between nodes (e.g., shared pathways, biochemical reactions, or correlation-based associations). A single metabolite (Star-shaped nodes) connected with multiple metabolites represent key hub or connector metabolites

Pathway enrichment analysis identified coordinated suppression of several core hepatic metabolic pathways in liver disease patients (Fig. 4C). Purine metabolism was significantly downregulated (p = 6 × 10− 03), consistent with the liver’s central role in nucleotide turnover and detoxification (Yang et al., 2022). Primary bile acid biosynthesis was also markedly reduced (p = 6 × 10− 03), reflecting impaired hepatocyte function, enzymatic disruption, and cholestatic processes (Li & Apte, 2015). Additional downregulated pathways included vitamin B6 metabolism and pantothenate and CoA biosynthesis (both p = 1 × 10− 01), suggesting compromised cofactor availability, impaired enzymatic activity, and disrupted energy metabolism (Parra et al., 2018). In contrast, several metabolic pathways were enriched in liver disease patients (Fig. 4D). Caffeine metabolism showed strong upregulation (p = 1 × 10− 05), consistent with altered cytochrome P450 activity and compensatory xenobiotic detoxification mechanisms (Shan et al., 2022). Multiple amino acid metabolism pathways—including glycine, serine, threonine, branched-chain amino acids, alanine, aspartate, and glutamine, were also elevated, indicating increased protein turnover and metabolic adaptation to support gluconeogenesis, nitrogen balance, and stress responses. Enrichment of the TCA.

Majority of the metabolites were mapped to pathways reflecting tissue remodeling, bile acid homeostasis, mitochondrial function, and nitrogen metabolism. These included markers of extracellular matrix turnover (e.g., isodesmosine), polyamine metabolism and translational stress (hypusine), bile acid and sphingolipid metabolism, and fatty acid β-oxidation (Ma et al., 2011; Park et al., 1997; Wang et al., 2024; Castelli et al., 2023; Hecker et al., 2014). Together, these metabolites reflect coordinated adaptations to hepatic dysfunction, encompassing disrupted bile acid synthesis, altered mitochondrial energetics, enhanced lipid turnover, and impaired nitrogen handling (Muth et al., 2003; Woolbright et al., 2014).

cycle, sphingolipid metabolism, and glyoxylate metabolism further suggests compensatory remodeling of energy-producing pathways to maintain ATP generation under hepatic stress (Satapati et al., 2012; Marí & Fernández-Checa, 2007). Network topology analysis of downregulated metabolites (fold change < 2) revealed tightly connected modules centered on uric acid, xanthine, cyclic GMP, and gluconic acid (Fig. 4E). These hubs link purine catabolism, nucleotide signaling, and carbohydrate metabolism, highlighting coordinated disruption across interconnected hepatic pathways (Kushiyama et al., 2016). Conversely, upregulated metabolites (fold change > 2) formed distinct clusters centered on citric acid, pioglitazone, and paraxanthine, corresponding to enhanced energy metabolism, lipid signaling, and xenobiotic processing (Fig. 4F). Citric acid emerged as a key connector metabolite, linking the TCA cycle with amino acid and lipid metabolism (Sahlin et al., 1990; Betteridge, 2007; Campagna & Vignini, 2025).

Collectively, these contrasting network architectures indicate extensive metabolic reprogramming in liver disease, characterized by suppression of nucleotide and bile acid pathways alongside activation of energy, lipid, amino acid, and xenobiotic metabolism. This pattern aligns with metabolic adaptations reported in hepatic fibrosis and hepatocellular carcinoma, where chronic liver injury drives compensatory mechanisms to sustain energy production and redox balance (Delgado et al., 2021; Park & Hall, 2025). To restrict analyses to biologically relevant host-derived signals, metabolites likely originating from environmental exposure, diet, or supplements were excluded prior to multivariate modeling. Despite this filtering, the remaining endogenous metabolite set (n = 129) retained strong discriminatory capacity. Random Forest classification achieved high performance across feature subsets (AUC range: 0.993–0.998; Fig. 5A). An optimized model using the top 15 metabolites achieved an AUC of 0.998 (95% CI: 0.992–1.000.992.000), underscoring the robustness of liver disease–associated metabolic signatures (Fig. 5B, C).

Fig. 5.

Fig. 5

Open in a new tab

Metabolite signatures distinguishing Liver patients from healthy controls: A Multivariate ROC curves showing RF model performance across increasing feature numbers. B Variable-importance ranking of the top 15 metabolites contributing to classification. C Box-and-whisker plots depicting relative abundance differences of the most discriminative metabolites between Liver patients and healthy controls

Common and unique metabolic pathways in COVID-19 and patients with liver disease

To identify metabolic features distinguishing COVID-19 patients, liver disease patients, and healthy controls, a multiclass Random Forest (RF) model was constructed using 129 endogenous metabolites after exclusion of diet- and environment-derived compounds. The model achieved an overall classification accuracy of 91.9%, with strong class-specific performance (COVID-19: AUC = 1.0; Liver disease: AUC = 0.99; Healthy controls: AUC = 0.98) (Fig. 6A). Feature importance was assessed using normalized mean SHAP (SHapley Additive exPlanations) values, which quantify each metabolite’s contribution to disease classification (Fig. 6B, C).

Fig. 6.

Fig. 6

Open in a new tab

Identification of unique metabolites associated with COVID-19 and liver disease: A Receiver Operating Characteristic (ROC) curve showing the performance of the Random Forest classifier based on urinary metabolite profiles from healthy controls, COVID-19 patients, and patients with liver disease; B Bar plot showing the top 20 most significantly altered metabolites based on normalize SHAP score; C Box- and-whisker plots displaying normalized concentrations of each metabolite across the three groups, highlighting differential abundance patterns

The multiclass model reinforced metabolites previously identified in pairwise comparisons while also revealing additional features contributing to class separation. Metabolites most strongly associated with COVID-19 included biocytin, N,N-dimethylarginine, S-adenosylhomocysteine, and succinyladenosine, whereas liver disease classification was driven primarily by bile-acid–related metabolites, vitamin D3 derivatives, and dicarboxylic acids such as 3-hydroxysebacic and 3-hydroxysuberic acids. Notably, the multiclass framework highlighted metabolites not prioritized in binary models, including N-acetylvaline and the dipeptide Asn–Trp. N-acetylvaline contributed strongly to COVID-19 classification and mapped to branched-chain amino acid (BCAA) metabolism, whereas Asn–Trp was prioritized in liver disease and mapped to peptide turnover and amino acid catabolism, consistent with altered tryptophan metabolism in chronic liver dysfunction (Atila et al., 2021; Reshetova et al., 2024).

Pathway-level mapping of discriminatory metabolites revealed both shared and disease-specific metabolic perturbations (SI Table-3 & 4). COVID-19 associated metabolites were predominantly enriched in branched-chain amino acid metabolism, nucleotide turnover, methylation reactions, nitric oxide signaling, steroid hormone biosynthesis, and cofactor (biotin) metabolism, reflecting systemic metabolic imbalance, immune activation, and altered epigenetic regulation characteristic of viral infection (McCann et al., 2007; Obi et al., 2011; Ascencio-Anastacio et al., 2025; Wells & Holian, 2007; Wu et al., 2022; Caruso et al., 2017). In contrast, metabolites distinguishing liver disease mapped primarily to fatty acid β-oxidation, bile acid metabolism, catecholamine and tyrosine metabolism, and peptide turnover pathways, consistent with impaired hepatic energy metabolism, disrupted bile acid homeostasis, and altered neurotransmitter processing (Castelli et al., 2023; Hecker et al., 2014; Muth et al., 2003; Woolbright et al., 2014).

Despite these distinctions, both disease groups shared perturbations in peptide and amino acid metabolism as well as steroid-related pathways, suggesting common metabolic stress responses such as dysregulated protein turnover and hormonal signaling imbalance. These shared features likely reflect generalized systemic adaptation to disease-associated stress, whereas the remaining pathway alterations were disease-specific. COVID-19 exhibited pronounced dysregulation of nucleotide metabolism and BCAA pathways, potentially linked to viral replication demands and immune activation (Fan et al., 2025; Ozturk et al., 2022; Sánchez et al., 2023), while liver disease showed dominant alterations in lipid oxidation, bile acid processing, and catecholamine metabolism, reflecting organ-specific metabolic failure. Together, these findings demonstrate that multiclass metabolomic modeling can resolve both overlapping and distinct metabolic signatures across disease states, providing insight into disease-specific mechanisms and informing the development of targeted biomarkers and therapeutic strategies.

Limitations

This study has a few limitations that we would like to acknowledge. Although basic demographic variables (age, sex, and BMI) were available, detailed clinical metadata, including medication history, dietary intake, disease severity, comorbidities, and environmental exposures, were not systematically recorded. Samples were obtained from the ILBS biobank during the peak of the COVID-19 pandemic, which limited the feasibility of retrieving additional clinical information retrospectively. Moreover, given the substantial dietary diversity in India, the absence of dietary and medication data is particularly relevant for urine metabolomics, as these factors can strongly influence xenobiotic- and diet-derived metabolites. To mitigate this, we excluded metabolites with a high likelihood of exogenous origin and focused our biological interpretation on well-annotated endogenous pathways; however, some residual confounding cannot be completely ruled out. Second, urinary metabolite concentrations were not normalized to creatinine or specific gravity, which may introduce variability due to differences in urine dilution. While consistent sample handling and analytical procedures were applied across all groups to minimize systematic bias, future studies incorporating urine dilution correction are expected to improve quantitative robustness and cross-study comparability. Third, the cross-sectional study design limits causal inference and does not allow assessment of temporal metabolic changes during disease progression or recovery. Longitudinal studies will therefore be essential to distinguish transient disease-associated alterations from stable metabolic signatures. Fourth, the findings are derived from an untargeted discovery cohort without external validation or targeted quantitative confirmation. Although internal cross-validation and machine-learning–based feature selection were used to enhance robustness, independent cohort validation and targeted assays will be necessary before clinical translation. Finally, this study predominantly relies on MSI level 2–3 metabolite annotations, reflecting the inherent challenges of large-scale untargeted urine metabolomics. To avoid overinterpretation, we moderated metabolite-level discussion, emphasized pathway-level trends, and clearly distinguished putative from more confidently annotated features. Expanding spectral libraries and integrating targeted validation approaches may further improve annotation confidence in future work. Despite these limitations, we believe that this study provides a comprehensive comparative analysis of urinary metabolic alterations in COVID-19 and liver disease, highlights both shared and disease-specific metabolic pathways, and establishes a strong foundation for future hypothesis-driven and validation-focused metabolomics investigations.

Conclusions

Metabolic reprogramming is a hallmark of both infectious and chronic diseases, reflecting the host’s adaptive responses to stress, energy demands, and organ-specific dysfunction. In this study, urine metabolite profiles of patients infected with SARS-CoV-2 (COVID-19) and patients with liver disease were compared to those of healthy controls to characterize disease-associated metabolic changes. Using Random Forest–based classification for COVID-19 versus healthy, liver disease versus healthy, and a multi-class model, we identified both shared and distinct metabolic pathways and signatures. In both COVID-19 and liver disease, disruptions in amino acid, peptide, and steroid-related pathways were observed, indicating common stress responses and systemic adaptation. Unique metabolites, such as N-Acetylvaline, Succinyladenosine, and S-adenosylhomocysteine in COVID-19, highlighted altered nucleotide and branched-chain amino acid metabolism, reflecting heightened immune activity and energy demands. In liver disease, metabolites including 3-Hydroxysebacic acid, Asn-Trp, and bile acid derivatives indicated perturbations in fatty acid oxidation, bile acid metabolism, and nitrogen handling, consistent with organ-specific metabolic stress and mitochondrial adaptation. These findings enhance our understanding of disease-specific metabolic remodeling and point to potential biomarkers for diagnosis and therapeutic targeting. Future studies should investigate the functional roles of these metabolites and validate their clinical utility, paving the way for precision medicine in both infectious and hepatic disorders.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 (6.8MB, xls)
Supplementary Material 2 (7.1MB, docx)

Acknowledgements

We would like to thank the ILBS Biobank for providing the samples and Jamia Millia Islamia for supporting the project. FK gratefully acknowledges financial support from ICMR (IIRP-2023-0223) and extends thanks to Dr. Rohit Kumar from the Department of Pulmonary, Critical Care, and Sleep Medicine, VMMC, Safdarjung Hospital, for his valuable support. SS acknowledges financial support from ICMR (ISRM/12(45)/2021-6413).

Author contributions

GJ: Contributed in new methods, analyzed data and wrote the paper; FK: Wrote the paper and performed research; SS: Sample collection, data recording and generation; AR: Contributed in data analysis; AM: Contributed in New method and analyzed data NM: Contributed in new methods, performed research; GDJ: Performed research and analyzed data; CB: Study Design and sample collection; MCJ: Study design, analyzed data, wrote paper. All authors reviewed the manuscript.

Funding

MCJ acknowledges support from ICMR BMI/12(45)/2021 and University Grants Commission (UGC), F.4–5/236 FRP (2015) BSR.

Data availability

Data can be shared upon reasonable request to the corresponding authors.

Declarations

Competing interests

The authors declare no competing interests.

Conflict of interest

All authors declare no conflict of interest.

Ethical approval

Institutional Ethical approval for the study was taken from Jamia Millia Islamia (IEC registration number: 1/10/290/JMI/IEC/2020) as well as from ILBS (No. F.37/(1)/9/ILBS/DOA/2020/20217/776).

Footnotes

Publisher’s note

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

Contributor Information

Garima Juyal, Email: garima.juyal@bennett.edu.in.

Mohan Chandra Joshi, Email: mjoshi@jmi.ac.in.

References

  1. Ansone, L., Rovite, V., Brīvība, M., Jagare, L., Pelcmane, L., Borisova, D., Thews, A., Leiminger, R., & Kloviņš, J. (2024). Longitudinal NMR-based metabolomics study reveals how hospitalized COVID-19 patients recover: Evidence of dyslipidemia and energy metabolism dysregulation. International Journal of Molecular Sciences,25(3), Article 1523. 10.3390/ijms25031523 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Ascencio-Anastacio, A. M., Larios-Serrato, V., Mata-Marín, J. A., Rodríguez Evaristo, M., Núñez-Armendáriz, M., Cano-Díaz, A. L., Chaparro-Sánchez, A., Salinas-Velázquez, G. E., Maldonado-Rodríguez, A., Torres, J., García-Flores, M. M., Martínez-Valencia, Z. E., Arroyo-Sánchez, B. I., Olin-Sandoval, V., Minauro, F., Gaytán-Martínez, J. E., & Pompa-Mera, E. N. (2025). Metabolomic profile of weight gain of people living with HIV treated with integrase strand transfer inhibitor regimens reveals dysregulated lipid metabolism and mitochondrial dysfunction. Metabolite,11, Article 695. 10.3390/metabo15110695 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Atila, A., Alay, H., Yaman, M. E., Akman, T. C., Cadirci, E., Bayrak, B., Celik, S., Atila, N. E., Yaganoglu, A. M., Kadioglu, Y., Halıcı, Z., Parlak, E., & Bayraktutan, Z. (2021). The serum amino acid profile in COVID-19. Amino Acids,53(10), 1569–1588. 10.1007/s00726-021-03081-w [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Azzini, E., Ruggeri, S., & Polito, A. (2020). Homocysteine: Its possible emerging role in at-risk population groups. International Journal of Molecular Sciences,21(4), Article 1421. 10.3390/ijms21041421 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Badawy, A.-B. (2023). The kynurenine pathway of tryptophan metabolism: A neglected therapeutic target of COVID-19 pathophysiology and immunotherapy. Bioscience Reports. 10.1042/BSR20230595 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bajpai, M., Maheshwari, A., Dogra, V., Kumar, S., Gupta, E., Kale, P., Saluja, V., Thomas, S. S., Trehanpati, N., Bihari, C., et al. (2022). Efficacy of convalescent plasma therapy in the patient with COVID-19: A randomised control trial (COPLA-II trial). BMJ Open,12, Article e055189. 10.1136/bmjopen-2021-055189 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Betteridge, D. J. (2007). Effects of pioglitazone on lipid and lipoprotein metabolism. Diabetes, Obesity and Metabolism,9(5), 640–647. 10.1111/j.1463-1326.2007.00715.x [DOI] [PubMed] [Google Scholar]
  8. Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. 10.1023/A:1010933404324 [Google Scholar]
  9. Buchynskyi, M., Kamyshna, I., Oksenych, V., Zavidniuk, N., & Kamyshnyi, A. (2023). The intersection of COVID-19 and metabolic-associated fatty liver disease: An overview of the current evidence. Viruses,15(5), Article 1072. 10.3390/v15051072 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Buchynskyi, M., Oksenych, V., Kamyshna, I., & Kamyshnyi, O. (2024). Exploring paxlovid efficacy in COVID-19 patients with MAFLD: Insights from a single-center prospective cohort study. Viruses,16(1), 112. 10.3390/v16010112 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Campagna, R., & Vignini, A. (2025). The role of xenobiotic caffeine on cardiovascular health: Promises and challenges. Journal of Xenobiotics,15(2), Article 51. 10.3390/jox15020051 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Caruso, G., Fresta, C. G., Martinez-Becerra, F., et al. (2017). Carnosine modulates nitric oxide in stimulated murine RAW 264.7 macrophages. Molecular and Cellular Biochemistry,431, 197–210. 10.1007/s11010-017-2991-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Castelli, S., Ciccarone, F., De Falco, P., & Ciriolo, M. R. (2023). Adaptive antioxidant response to mitochondrial fatty acid oxidation determines the proliferative outcome of cancer cells. Cancer Letters,554, Article 216010. 10.1016/j.canlet.2022.216010 [DOI] [PubMed] [Google Scholar]
  14. Chen, C., Wang, J., Pan, D., Wang, X., Xu, Y., Yan, J., Wang, L., Yang, X., Yang, M., & Liu, G. (2023). Applications of multi-omics analysis in human diseases. MedComm. 10.1002/mco2.315 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Delgado, M. E., Cárdenas, B. I., Farran, N., & Fernandez, M. (2021). Metabolic reprogramming of liver fibrosis. Cells,10(12), Article 3604. 10.3390/cells10123604 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Dillard, L. R., Wase, N., Ramakrishnan, G., et al. (2022). Leveraging metabolic modeling to identify functional metabolic alterations associated with COVID-19 disease severity. Metabolomics,18, Article 51. 10.1007/s11306-022-01904-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. European Association for the Study of the Liver. (2018). EASL clinical practice guidelines: Management of hepatocellular carcinoma. Journal of Hepatology,69(1), 182–236. 10.1016/j.jhep.2018.03.019 [DOI] [PubMed] [Google Scholar]
  18. Fan, Y., Xie, X., Bai, G., Ji, W., He, Y., Zhang, L., Zhou, H., Li, L., Li, H., & Qiang, D. (2025). Nucleotide metabolic abnormalities in post-COVID-19 condition and type 2 diabetes mellitus patients and their association with endocrine dysfunction. Open Medicine (Warsaw),20(1), Article 20251221. 10.1515/med-2025-1221 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gama-Almeida, M. C., Pinto, G. D. A., Teixeira, L., Hottz, E. D., Ivens, P., Ribeiro, H., Garrett, R., Torres, A. G., Carneiro, T. I. A., Barbalho, B., Ludwig, C., Struchiner, C. J., Assunção-Miranda, I., Valente, A. P. C., Bozza, F. A., Bozza, P. T., dos Santos, G. C., & El-Bacha, T. (2023). Integrated NMR and MS analysis of the plasma metabolome reveals major changes in one-carbon, lipid, and amino acid metabolism in severe and fatal cases of COVID-19. Metabolites,13(7), Article 879. 10.3390/metabo13070879 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Guarnieri, J. W., Lie, T., Albrecht, Y. E. S., Hewin, P., Jurado, K. A., Widjaja, G. A., Zhu, Y., McManus, M. J., Kilbaugh, T. J., Keith, K., Potluri, P., Taylor, D., Angelin, A., Murdock, D. G., & Wallace, D. C. (2024). Mitochondrial antioxidants abate SARS-COV-2 pathology in mice. Proceedings of the National Academy of Sciences. 10.1073/pnas.2321972121 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hecker, M., Sommer, N., Voigtmann, H., Pak, O., Mohr, A., Wolf, M., Vadász, I., Herold, S., Weissmann, N., Morty, R. E., Seeger, W., & Mayer, K. (2014). Impact of short- and medium-chain fatty acids on mitochondrial function in severe inflammation. Journal of Parenteral and Enteral Nutrition,38(5), 587–594. 10.1177/0148607113489833 [DOI] [PubMed] [Google Scholar]
  22. Hernández-Flores, T., Pedraza-Brindis, E. J., Cárdenas-Bedoya, J., Ruíz-Carrillo, J. D., Méndez-Clemente, A. S., Martínez-Guzmán, M. A., & Iñiguez-Gutiérrez, L. (2022). Role of micronutrients and gut microbiota-derived metabolites in COVID-19 recovery. International Journal of Molecular Sciences,23(20), Article 12324. 10.3390/ijms232012324 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Imam, M. T., Almalki, Z. S., Alzahrani, A. R., Al-Ghamdi, S. S., Falemban, A. H., Alanazi, I. M., Shahzad, N., Muhammad Alrooqi, M., Jabeen, Q., & Shahid, I. (2023). COVID-19 and severity of liver diseases: Possible crosstalk and clinical implications. International Immunopharmacology,121, Article 110439. 10.1016/j.intimp.2023a.110439 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kaur, A., Azeez, G. A., Thirunagari, M., Fatima, N., Anand, A., Palvia, A. R., & Yu, A. K. (2024). Association of chronic hepatitis B with colorectal cancer and its dual impact on colorectal liver metastasis: A narrative review. Cureus,16(12), Article e76079. 10.7759/cureus.76079 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Khan, S. A., Goliwas, K. F., & Deshane, J. S. (2021). Sphingolipids in lung pathology in the coronavirus disease era: A review of sphingolipid involvement in the pathogenesis of lung damage. Frontiers in Physiology. 10.3389/fphys.2021.760638 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kocaoğlu, S., Berkan, T., Karan, A., & Başdemir, G. (1994). Study on the role of urine gamma-glutamyl transpeptidase activity during investigation of nephrotoxicity. Archivum Immunologiae Et Therapiae Experimentalis (Warszawa), 42(2), 125–127. [PubMed] [Google Scholar]
  27. Kushiyama, A., Nakatsu, Y., Matsunaga, Y., Yamamotoya, T., Mori, K., Ueda, K., Inoue, Y., Sakoda, H., Fujishiro, M., Ono, H., & Asano, T. (2016). Role of uric acid metabolism-related inflammation in the pathogenesis of metabolic syndrome components such as atherosclerosis and nonalcoholic steatohepatitis. Mediators of Inflammation,2016, Article 8603164. 10.1155/2016/8603164 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. León-Del-Río, A. (2005). Biotin-dependent regulation of gene expression in human cells. Journal of Nutritional Biochemistry, 16(7), 432–434. 10.1016/j.jnutbio.2005.03.021 [DOI] [PubMed] [Google Scholar]
  29. Lewis, J. S., Thomas, T. J., Klinge, C. M., Gallo, M. A., & Thomas, T. (2001). Regulation of cell cycle and cyclins by 16α-hydroxyestrone in MCF-7 breast cancer cells. Journal of Molecular Endocrinology,27(3), 293–307. 10.1677/jme.0.0270293 [DOI] [PubMed] [Google Scholar]
  30. Li, F., Fu, L., Liu, X., Liu, X., Liang, Y., Lv, Y., Yang, Z., Guo, A., Chen, Z., Li, W., Pan, F., & Luo, Q. (2022). Serum metabolomic abnormalities in survivors of non-severe COVID-19. Heliyon,8(9), Article e10473. 10.1016/j.heliyon.2022.e10473 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Li, T., & Apte, U. (2015). Bile Acid Metabolism and Signaling in Cholestasis, Inflammation, and Cancer (pp. 263–302). 10.1016/bs.apha.2015.04.003 [DOI] [PMC free article] [PubMed]
  32. Lin, C., Tian, Q., Guo, S., Xie, D., Cai, Y., Wang, Z., Chu, H., Qiu, S., Tang, S., & Zhang, A. (2024). Metabolomics for clinical biomarker discovery and therapeutic target identification. Molecules (Basel, Switzerland),29(10), Article 2198. 10.3390/molecules29102198 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Liu, J., Liao, X., Qian, S., Yuan, J., Wang, F., Liu, Y., Wang, Z., Wang, F. S., Liu, L., & Zhang, Z. (2020). Community transmission of severe acute respiratory syndrome Coronavirus 2, Shenzhen, China, 2020. Emerging Infectious Diseases. 10.3201/eid2606.200239 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lundberg, S. M., & Lee, S. I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems (NIPS). 10.5555/3295222.3295230 [Google Scholar]
  35. Ma, S., Turino, G. M., & Lin, Y. Y. (2011). Quantitation of desmosine and isodesmosine in urine, plasma, and sputum by LC-MS/MS as biomarkers for elastin degradation. Journal of Chromatography. B, Analytical Technologies in the Biomedical and Life Sciences,879(21), 1893–1898. 10.1016/j.jchromb.2011.05.011 [DOI] [PubMed] [Google Scholar]
  36. Marí, M., & Fernández-Checa, J. C. (2007). Sphingolipid signalling and liver diseases. Liver International, 27(4), 440–450. 10.1111/j.1478-3231.2007.01475.x [DOI] [PubMed] [Google Scholar]
  37. Masoodi, M., Peschka, M., Schmiedel, S., Haddad, M., Frye, M., Maas, C., Lohse, A., Huber, S., Kirchhof, P., Nofer, J. R., & Renné, T. (2022). Disturbed lipid and amino acid metabolisms in COVID-19 patients. Journal of Molecular Medicine (Berlin, Germany),100(4), 555–568. 10.1007/s00109-022-02177-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. McCann, S. E., Wactawski-Wende, J., Kufel, K., Olson, J., Ovando, B., Kadlubar, S. N., Davis, W., Carter, L., Muti, P., Shields, P. G., & Freudenheim, J. L. (2007). Changes in 2-hydroxyestrone and 16α-hydroxyestrone metabolism with flaxseed consumption: Modification by COMT and CYP1B1 genotype. Cancer Epidemiology, Biomarkers & Prevention : a Publication of the American Association for Cancer Research, Cosponsored by the American Society of Preventive Oncology,16(2), 256–262. 10.1158/1055-9965.EPI-06-0633 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Mueller, A. L., McNamara, M. S., & Sinclair, D. A. (2020). Why does COVID-19 disproportionately affect older people? Aging,12(10), 9959–9981. 10.18632/aging.103344 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Mussabay, K., Kozhakhmetov, S., Dusmagambetov, M., Mynzhanova, A., Nurgaziyev, M., Jarmukhanov, Z., Vinogradova, E., Dusmagambetova, A., Daulbaeva, A., Chulenbayeva, L., Tauekelova, A., Bekbossynova, M., & Kushugulova, A. (2024). Gut microbiome and cytokine profiles in Post-COVID Syndrome. Viruses,16(5), Article 722. 10.3390/v16050722 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Muth, A., Mosandl, A., Wanders, R. J. A., Nowaczyk, M. J. M., Baric, I., Böhles, H., & Sewell, A. C. (2003). Stereoselective analysis of 2-hydroxysebacic acid in urine of patients with Zellweger syndrome and of premature infants fed with medium-chain triglycerides. Journal of Inherited Metabolic Disease,26(6), 583–592. 10.1023/a:102590821663 [DOI] [PubMed] [Google Scholar]
  42. Niazkar, H. R., Zibaee, B., Nasimi, A., & Bahri, N. (2020). The neurological manifestations of COVID-19: A review article. Neurological Sciences,41(7), 1667–1671. 10.1007/s10072-020-04486-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Obi, N., Vrieling, A., Heinz, J., & Chang-Claude, J. (2011). Estrogen metabolite ratio: Is the 2-hydroxyestrone to 16α-hydroxyestrone ratio predictive for breast cancer? International Journal of Women’s Health,3, 37–51. 10.2147/IJWH.S7595 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Oropeza-Valdez, J. J., Padron-Manrique, C., Vázquez-Jiménez, A., Soberon, X., & Resendis-Antonio, O. (2024). Exploring metabolic anomalies in COVID-19 and post-COVID-19: a machine learning approach with explainable artificial intelligence. Frontiers in Molecular Biosciences,11, 1429281. 10.3389/fmolb.2024.1429281 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Ozturk, A., Bayraktar, N., Bayraktar, M., Ibrahim, B., Bozok, T., & Resat, C. M. (2022). Evaluation of amino acid profile in serum of patients with COVID-19 for providing a new treatment strategy. Journal of Medical Biochemistry,41(4), 526–533. 10.5937/jomb0-37514 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Paneque, A., Fortus, H., Zheng, J., Werlen, G., & Jacinto, E. (2023). The hexosamine biosynthesis pathway: Regulation and function. Genes,14(4), Article 933. 10.3390/genes14040933 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Pang, Z., Lu, Y., Zhou, G., Hui, F., Xu, L., Viau, C., Spigelman, A. F., MacDonald, P. E., Wishart, D. S., Li, S., & Xia, J. (2024). MetaboAnalyst 6.0: Towards a unified platform for metabolomics data processing, analysis and interpretation. Nucleic Acids Research,52(W1), W398–W406. 10.1093/nar/gkae253 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Park, M. H., Lee, Y. B., & Joe, Y. A. (1997). Hypusine is essential for eukaryotic cell proliferation. Biological Signals,6(3), 115–123. 10.1159/000109117 [DOI] [PubMed] [Google Scholar]
  49. Park, S., & Hall, M. N. (2025). Metabolic reprogramming in hepatocellular carcinoma: Mechanisms and therapeutic implications. Experimental & Molecular Medicine,57(3), 515–523. 10.1038/s12276-025-01415-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Parra, M., Stahl, S., & Hellmann, H. (2018). Vitamin B6 and its role in cell metabolism and physiology. Cells,7(7), Article 84. 10.3390/cells7070084 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Priya, S., Burns, M. B., Ward, T., Mars, R. A. T., Adamowicz, B., Lock, E. F., Kashyap, P. C., Knights, D., & Blekhman, R. (2022). Identification of shared and disease-specific host gene–microbiome associations across human diseases using multi-omic integration. Nature Microbiology,7(6), 780–795. 10.1038/s41564-022-01121-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Reshetova, M., Markin, P., Appolonova, S., Yunusov, I., Zolnikova, O., Bueverova, E., Dzhakhaya, N., Zharkova, M., Poluektova, E., Maslennikov, R., & Ivashkin, V. (2024). Tryptophan metabolites in the progression of liver diseases. Biomolecules,14(11), Article 1449. 10.3390/biom14111449 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Richardson, S., Hirsch, J. S., Narasimhan, M., Crawford, J. M., McGinn, T., Davidson, K. W., Barnaby, D. P., Becker, L. B., Chelico, J. D., Cohen, S. L., Cookingham, J., Coppa, K., Diefenbach, M. A., Dominello, A. J., Duer-Hefele, J., Falzon, L., Gitlin, J., Hajizadeh, N., Harvin, T. G., & Zanos, T. P. (2020). Presenting Characteristics, Comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the new York City area. Journal of the American Medical Association, 323(20), 2052. 10.1001/jama.2020.6775 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Sahlin, K., Katz, A., & Broberg, S. (1990). Tricarboxylic acid cycle intermediates in human muscle during prolonged exercise. American Journal of Physiology, 259(5), C834–C841. 10.1152/ajpcell.1990.259.5.C834 [DOI] [PubMed] [Google Scholar]
  55. Sánchez, A., García-Pardo, G., Gómez-Bertomeu, F., López-Dupla, M., Foguet-Romero, E., Buzón, M. J., Almirante, B., Olona, M., Fernández-Veledo, S., Vidal, F., Chafino, S., Rull, A., & Peraire, J. (2023). Mitochondrial dysfunction, lipids metabolism, and amino acid biosynthesis are key pathways for COVID-19 recovery. IScience,26(10), Article 107948. 10.1016/j.isci.2023.107948 [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Satapati, S., Sunny, N. E., Kucejova, B., Fu, X., He, T. T., Méndez-Lucas, A., Shelton, J. M., Perales, J. C., Browning, J. D., & Burgess, S. C. (2012). Elevated TCA cycle function in the pathology of diet-induced hepatic insulin resistance and fatty liver. Journal of Lipid Research, 53(6), 1080–1092. 10.1194/jlr.M023382 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Shan, L., Wang, F., Zhai, D., Meng, X., Liu, J., & Lv, X. (2022). Caffeine in liver diseases: Pharmacology and toxicology. Frontiers in Pharmacology. 10.3389/fphar.2022.1030173 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Shen, T., & Wang, T. (2021). Metabolic reprogramming in COVID-19. International Journal of Molecular Sciences,22(21), Article 11475. 10.3390/ijms222111475 [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Steinberg, S., & Cedergren R. (1995). A correlation between N2-dimethylguanosine presence and alternate tRNA conformers. RNA,1(9), 886–91. [PMC free article] [PubMed] [Google Scholar]
  60. Tan, Q., Deng, S., & Xiong, L. (2025). Role of kynurenine and its derivatives in liver diseases: Recent advances and future clinical perspectives. International Journal of Molecular Sciences,26(3), 968. 10.3390/ijms26030968 [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Tsuji, A., Ikeda, Y., Yoshikawa, S., Taniguchi, K., Sawamura, H., Morikawa, S., Nakashima, M., Asai, T., & Matsuda, S. (2023). The tryptophan and kynurenine pathway involved in the development of immune-related diseases. International Journal of Molecular Sciences,24(6), Article 5742. 10.3390/ijms24065742 [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Valdés, A., Moreno, L. O., Rello, S. R., Orduña, A., Bernardo, D., & Cifuentes, A. (2022). Metabolomics study of COVID-19 patients in four different clinical stages. Scientific Reports,12(1), Article 1650. 10.1038/s41598-022-05667-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Wang, Y., Xu, H., Zhou, X., Chen, W., & Zhou, H. (2024). Dysregulated bile acid homeostasis: Unveiling its role in metabolic diseases. Medical Reviews, 4(4), 262–283. 10.1515/mr-2024-0020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Wang, X., Zhang, A., Han, Y., Wang, P., Sun, H., Song, G., Dong, T., Yuan, Y., Yuan, X., Zhang, M., Xie, N., Zhang, H., Dong, H., & Dong, W. (2012). Urine metabolomics analysis for biomarker discovery and detection of jaundice syndrome in patients with liver disease. Molecular & Cellular Proteomics, 11(8), 370–380. 10.1074/mcp.M111.016006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Wells, S. M., & Holian, A. (2007). Asymmetric dimethylarginine induces oxidative and nitrosative stress in murine lung epithelial cells. American Journal of Respiratory Cell and Molecular Biology,36(5), 520–528. 10.1165/rcmb.2006-0302SM [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Wilson, A. D., & Forse, L. B. (2023). Potential for early noninvasive COVID-19 detection using electronic-Nose technologies and disease-specific VOC metabolic biomarkers. Sensors,23(6), 2887. 10.3390/s23062887 [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Woolbright, B. L., McGill, M. R., Staggs, V. S., Winefield, R. D., Gholami, P., Olyaee, M., Sharpe, M. R., Curry, S. C., Lee, W. M., & Jaeschke, H. (2014). Glycodeoxycholic acid levels as prognostic biomarker in acetaminophen-induced acute liver failure patients. Toxicological Sciences,142(2), 436–444. 10.1093/toxsci/kfu195 [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Wu, Q., Li, J., Zhu, J., Sun, X., He, D., Li, J., Cheng, Z., Zhang, X., Xu, Y., Chen, Q., Zhu, Y., & Lai, M. (2022). Gamma-glutamyl-leucine levels are causally associated with elevated cardiometabolic risks. Frontiers in Nutrition,9, Article 936220. 10.3389/fnut.2022.936220 [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Yang, R., Yang, C., Ma, L., Zhao, Y., Guo, Z., Niu, J., Chu, Q., Ma, Y., & Li, B. (2022). Identification of purine biosynthesis as an NADH-sensing pathway to mediate energy stress. Nature Communications,13(1), Article 7031. 10.1038/s41467-022-34850-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Yi, P., Melnyk, S., Pogribna, M., Pogribny, I. P., Hine, R. J., & James, S. J. (2000). Increase in plasma homocysteine associated with parallel increases in plasma S-adenosylhomocysteine and lymphocyte DNA hypomethylation. Journal of Biological Chemistry, 275(38), 29318–29323. 10.1074/jbc.M002725200 [DOI] [PubMed] [Google Scholar]
  71. Zhou, Z., Ren, L., Zhang, L., Zhong, J., Xiao, Y., Jia, Z., Guo, L., Yang, J., Wang, C., Jiang, S., Yang, D., Zhang, G., Li, H., Chen, F., Xu, Y., Chen, M., Gao, Z., Yang, J., Dong, J., & Wang, J. (2020b). Heightened innate immune responses in the respiratory tract of COVID-19 patients. Cell Host & Microbe, 27(6), 883–890e2. 10.1016/j.chom.2020.04.017 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1 (6.8MB, xls)
Supplementary Material 2 (7.1MB, docx)

Data Availability Statement

Data can be shared upon reasonable request to the corresponding authors.

Articles from Metabolomics are provided here courtesy of Springer

RESOURCES