A circulating odd‐chain fatty acid biomarker signature discriminates pediatric allergic rhinitis with asthma from related allergic airway phenotypes

Abstract

Background

Reliable biomarkers for discriminating allergic rhinitis with asthma (ARA) from related allergic airway phenotypes remain lacking in pediatric populations. We aimed to identify and validate circulating metabolic biomarkers of ARA.

Methods

Serum samples from 380 participants, including healthy controls (HC), allergic rhinitis (AR), asthma (AS), and ARA, underwent untargeted carboxylomics profiling using SPCSDCarboxyl. Candidate biomarkers were evaluated in mouse models and validated in an independent cohort (n = 125) using targeted carboxylomics. Biomarker performance was assessed using multivariate and machine‐learning approaches.

Results

ARA was associated with altered circulating odd‐chain fatty acids (OCFAs), with valeric, heptanoic, nonanoic, undecanoic, and isovaleric acids showing consistent increased across cohorts. Heptanoic, nonanoic, and undecanoic acids demonstrated excellent discrimination of ARA from HC, with areas under curve (AUC) of 0.96, 0.98, and 0.94, respectively. However, these three OCFAs were broadly elevated across all allergic airway phenotypes and lacked ARA specificity. Least absolute shrinkage and selection operator (LASSO) regression identified valeric and undecanoic acids as complementary discriminatory features for ARA versus AR and AS. Among valeric acid showed the most phenotype‐discriminatory performance, with an AUC of 0.72 for ARA versus AR and 0.64 for ARA versus AS. OCFA levels correlated significantly with inflammatory indices and pulmonary function, and comparable alterations were confirmed in mouse models.

Conclusions

These findings identify a pediatric ARA‐specific OCFA signature of a distinct immunometabolic endotype involving propionyl‐CoA metabolism and fatty acid oxidation. Valeric acid uniquely discriminated ARA from isolated AR, supporting its potential as a serum reference marker; combination with undecanoic acid or clinical parameters may further improve diagnostic performance.

Carboxylomics profiling revealed altered odd‐chain fatty acids in pediatric allergic airway diseases, with valeric acid showing the most relevant discriminatory performance for allergic rhinitis with asthma. Its association with the ARA phenotype was independently supported by targeted metabolomics and mouse models.

Abbreviations

AAD

allergic airway disease

ANOVA

analysis of variance

AR

allergic rhinitis

ARA

allergic rhinitis with asthma

AS

asthma

AUC

area under the receiver operating characteristic curve

BALF

bronchoalveolar lavage fluid

BCAA

branched‐chain amino acid

DIAAA

5‐(diisopropylamino)amylamine

FDR

false discovery rate

FeNO

fractional exhaled nitric oxide

FEV₁

forced expiratory volume in 1 s

FVC

forced vital capacity

GINA

Global Initiative for Asthma

HC

healthy control

KEGG

Kyoto Encyclopedia of Genes and Genomes

LASSO

least absolute shrinkage and selection operator

OCFA

odd‐chain fatty acid

OPLS‐DA

orthogonal partial least squares discriminant analysis

OVA

ovalbumin

PCA

principal component analysis

PFT

pulmonary function test

PLS‐DA

partial least squares discriminant analysis

ROC

receiver operating characteristic

sIgE

allergen‐specific immunoglobulin E

SPCSDCarboxyl

Sensitive Profiling for ChemoSelective Derivatization Carboxylomics

TCA

tricarboxylic acid

TIgE

total immunoglobulin E

UHPLC

ultra‐high‐performance liquid chromatography

VIP

variable importance in projection

Key message.

Comprehensive serum carboxylomics profiling revealed distinct metabolic alterations associated with allergic airway disease phenotypes in children. Valeric acid emerged as a potential metabolic marker for identifying allergic rhinitis with asthma beyond conventional immune‐based indicators, providing new insights into phenotype‐specific metabolic signatures.

1. INTRODUCTION

Allergic airway diseases (AADs), including allergic rhinitis (AR), allergic asthma (AS), and the combined phenotype allergic rhinitis with asthma (ARA), represent a clinically heterogeneous spectrum shaped by genetic susceptibility, environmental exposures, and immune dysregulation. ¹ AR affects more than 400 million people worldwide, ² and approximately 40%–50% of these individuals report asthma symptoms. ³ AS prevalence increases with age: about 1.9% in children aged 0–4 years and about 7.7% in school‐age children and adolescents, ⁴ , ⁵ and about 6%–7% in adults. ⁶ Comorbidity is common, with up to 37% of patients with AS also having AR. ⁷ Disease prognosis varies markedly by age at onset, with substantially higher remission rates in childhood‐onset asthma than in adult‐onset disease. ⁸ , ⁹

Despite their high prevalence and clinical impact, differentiation among AR, AS, and ARA remains challenging in routine practice. Overlapping respiratory and nasal symptoms, such as dyspnea, chest tightness, and nasal congestion, frequently obscure clinical distinction. ¹⁰ Early and phenotype‐specific diagnosis is critical, as early‐life sensitization and wheezing strongly predict subsequent asthma development. ¹¹ However, spirometry and fractional exhaled nitric oxide (FeNO) measurement require specialist equipment and active patient cooperation, restricting their use in primary care and in youngest children, who represent the age group in which early identification of asthma comorbidity carries the greatest preventive value. Serum‐based metabolic profiling circumvents both constraints, requiring only a standard blood draw independent of patient cooperation or specialist infrastructure.

Established diagnostic tools, including serum IgE, blood eosinophils counts, FeNO, skin prick testing, and spirometry, assess type‐2 immune sensitisation and airway inflammation but were not designed to resolve disease endotypes within the allergic airway spectrum. In clinical practice, serum total IgE (TIgE) and eosinophil counts are elevated across AR, AS, and ARA alike and therefore cannot distinguish children with ARA from those with a single phenotype; FeNO and spirometry require patient cooperation that young children frequently cannot provide, ¹¹ creating a diagnostic gap precisely in the age group where early identification of asthma comorbidity carries the greatest preventive value. No currently approved single biomarker reliably differentiates AR children at risk of comorbid asthma from those with isolated disease, underscoring a critical unmet need for approaches that capture the systemic metabolic divergence between phenotypes rather than the presence or magnitude of shared type‐2 inflammation.

Metabolomics offers a systems‐level approach to interrogate small‐molecule pathways that integrate environmental exposures, immune responses, and clinical phenotypes. ¹² Previous studies have reported metabolic alterations in energy, lipid, amino acid, and oxidative‐stress pathways in AR and AS ¹³ , ¹⁴ , ¹⁵ ; however, direct head‐to‐head metabolomic comparisons among AR, AS, and ARA—particularly in pediatric populations—remain limited. Clinical translation has been further constrained by technical challenges, ¹⁶ including insufficient coverage of carboxylic acids, which play central roles in metabolism and immune regulation.

We applied a high‐sensitivity 5‐(diisopropylamino) amylamine (DIAAA) derivatization combined with UHPLC‐Q‐TOF/MS to profile carboxylic acids. This approach, termed Sensitive Profiling for ChemoSelective Derivatization Carboxylomics (SPCSDCarboxyl), analyzed serum samples from children with AR, AS, and ARA, as well as healthy controls (HC). ¹⁷ This study systematically characterized shared and phenotype‐specific carboxylate metabolite signatures, with particular focus on ARA, and evaluated their potential as minimally invasive biomarkers for pediatric allergic airway disease.

2. METHODS

2.1. Study design and participants

Serum samples from 150 children with AADs and 24 HC were collected as the training cohort from the First Affiliated Hospital of Guangzhou Medical University (FAHGMU) in 2022. Additionally, 192 children with AADs and 14 HC were recruited from the Bio‐bank of the National Clinical Research Center for Respiratory Disease as the replication cohort. An independent validation cohort of 125 participants was obtained from FAHGMU in 2026. For the independent targeted validation cohort, sample size was informed by pilot observations from our group and published carboxylomics studies in pediatric allergic airway disease, ¹³ , ¹⁷ , ¹⁸ targeting a minimum of 25 subjects per group. Post‐hoc power analysis indicated that this sample size provides approximately 80% power to detect a biomarker AUC ≥0.75 at a two‐sided α = .05, consistent with the AUC values subsequently observed for the primary OCFA biomarkers.

Patients enrolled in our study completed questionnaires on demographics and medical history. Diagnoses were based on Allergic Rhinitis and its Impact on Asthma (ARIA) guidelines for AR and ARA, and Global Initiative for Asthma (GINA) guidelines for AS. ¹⁹ , ²⁰ The detailed criteria were as follows: (1) patients aged 1–18 years; (2) sensitization to house dust mite (HDM), including Dermatophagoides pteronyssinus (Der p) and/or Dermatophagoides farinae (Der f), confirmed by serum‐specific immunoglobulin E (sIgE) testing (sIgE ≥0.35 kU/L). To minimize potential metabolic confounding from other allergic conditions, patients with sensitization to allergens other than HDM, food allergy, or atopic dermatitis were excluded. Therefore, all included participants were classified as HDM‐monosensitized children.

2.2. Specific and total IgE measurements

Blood samples were collected during the participants' first clinic visit. Serum was prepared within 2 h and stored at −80°C until analysis. Serum levels of TIgE and sIgE against Der p and Der f were measured using the fluorescence enzyme immunoassay (FEIA) method (ImmunoCAP® assay, Thermo Fisher Scientific Inc., USA). All samples were tested at the Respiratory Immunology Center of the Guangzhou Institute of Respiratory Health using standardized methods in accordance with the manufacturer's instructions.

2.3. Pulmonary function testing (PFT)

PFT was performed according to American Thoracic Society and European Respiratory Society (ATS/ERS) guidelines using the MasterScreen system (Jaeger, CareFusion GmbH, Hoechberg, Germany). ²¹ Key spirometric indices, including forced expiratory volume in 1 s % predicted (FEV₁% pred), forced vital capacity % predicted (FVC% pred), and the FEV₁/FVC ratio, were assessed.

2.4. Inflammatory parameters

Venous blood samples were collected for white blood cell analysis. Complete blood counts with automated differential leukocyte analysis were performed using a Beckman Coulter LH 755 analyzer according to standard laboratory procedures. ²²

2.5. Sample preparation and metabolomic analysis

Serum samples were centrifuged at 800 × g for 10 min, and the supernatants were stored at −80°C until analysis. For metabolomic profiling, serum aliquots were processed using SPCSDCarboxyl to enhance the detection sensitivity and chromatographic separation of carboxylic acids, particularly for structural isomers (Methods S1 and S2). Stable isotope–labeled internal standards were added prior to extraction to correct for variability during sample preparation and instrumental analysis. Protein precipitation was performed with pre‐chilled methanol, followed by centrifugation and nitrogen drying.

Derivatization with DIAAA was applied to improve the ionization efficiency of carboxylic acids. ¹⁷ , ¹⁸ Derivatized samples were analyzed using UHPLC–Q‐TOF/MS for untargeted profiling, ¹⁷ , ¹⁸ and UHPLC–MS/MS for targeted quantification, ²³ as previously described (Methods S3 and S4). For targeted UHPLC–MS/MS analysis, OCFAs and L‐alanine were quantified using optimized multiple reaction monitoring (MRM) transitions generated from authentic standards, calibration curves, and stable isotope‐labeled internal standards. Detailed analytical parameters and MRM transitions are provided in Methods S4. Sample injection order was randomized to minimize batch effects. Pooled quality control (QC) samples were prepared by combining equal aliquots from all study samples and were analyzed periodically throughout the analytical sequence to assess retention time stability, signal variation, and overall analytical reproducibility. Data normalization and quality control procedures are described in Methods S5.

2.6. Animal models

Mouse models of AADs were established using 6‐week‐old female C57BL/6 mice sensitized with ovalbumin (OVA)/alum. Mice were randomly assigned to wild‐type (WT), AR, AS, and ARA (n = 5 per group). Sensitization was performed via intraperitoneal (i.p.) injection on days 1, 7, and 14. Airway challenges were conducted from days 28 to 35: the AS group received intratracheal (i.t.) OVA, the AR group received intranasal (i.n.) OVA, and the ARA group underwent alternating i.t. and i.n. OVA exposure. On day 36, mice were euthanized, and blood, bronchoalveolar lavage fluid (BALF), nasal mucosa, and lung tissues were collected. Blood samples were processed to obtain serum (800 × g, 10 min, 4°C), BALF was collected by PBS lavage and centrifugation (400 × g, 5 min, 4°C), and nasal mucosa and lung tissues were collected for histopathological and subsequent analyses, respectively. Nasal mucosal and lung tissues histology, methacholine‐induced airway responsiveness, and BALF differential cell counts were assessed to characterize the airway phenotypes.

2.7. Data analysis

Raw LC–MS data were processed using SCIEX OS software for peak detection, alignment, identification, normalization, and metabolite identification. Metabolites were annotated using ClassyFire, and pathway visualization was performed with iPath. Multivariate statistical analysis was performed using orthogonal partial least squares discriminant analysis (OPLS‐DA) for pairwise comparisons and partial least squares discriminant analysis (PLS‐DA) for multigroup comparisons. Model performance was assessed using R²Y and Q² values, permutation testing, and cross‐validated analysis of variance (CV‐ANOVA). Differential metabolites (DMs) were identified based on a combination of variable importance in projection (VIP >1.0) from OPLS‐DA, with |fold change| > 1.2 and false discovery rate (FDR) q < .05. Univariate comparisons between groups were performed using Student's t‐test or one‐way analysis of variance (ANOVA) as appropriate. Multiple testing correction was conducted using the Benjamini–Hochberg method to control the false discovery rate (FDR). Pathway enrichment analysis was conducted using RaMP‐DB–integrated databases (KEGG, Reactome, WikiPathways) within MetaboAnalyst 6.0. Significantly enriched pathways were defined as metabolic pathways in which the identified differential metabolites were represented more frequently than expected by chance, as determined using a hypergeometric enrichment test with FDR‐adjusted q < .05. Least absolute shrinkage and selection operator (LASSO) regression was performed on the training and replication cohort for variable selection. The resulting biomarker panel was applied without modification for receiver operating characteristic (ROC) analysis in the fully independent validation cohort (n = 125). No data from the validation cohort were used in any feature selection step. ROC curve analysis was used to evaluate the diagnostic performance of candidate metabolites, with area under the curve (AUC), sensitivity, and specificity reported. Associations between metabolite levels and clinical parameters were assessed using Spearman's rank correlation coefficient. All statistical analyses were conducted using Metware Cloud and R software (version 4.6.0). A two‐sided p value <.05 was considered statistically significant unless otherwise specified. Detailed descriptions of data processing and statistical analysis are provided in Methods S5.

3. RESULTS

3.1. Demographic and clinical characteristics

A total of 380 children were enrolled for differential metabolite discovery and randomly allocated to a training cohort (n = 174) and a replication cohort (n = 206) (Table S1). The four study groups were well balanced with respect to sex, age, and venous blood cell counts, indicating minimal baseline heterogeneity and reducing the likelihood that demographic or hematologic differences confounded the observed metabolic variations. In the validation cohort (n = 125) (Table 1), the four groups were comparable in age, sex, and BMI (all p > .05). TIgE levels differed significantly among the four groups (p < .0001), primarily driven by lower levels in HC compared with AAD groups. Although AS showed the highest median TIgE level (370.2 kU/L), followed by ARA (280.9 kU/L) and AR (131.3 kU/L), pairwise comparisons among AR, AS, and ARA did not reveal significant differences. Pulmonary function parameters and peripheral blood leukocyte subsets showed no significant intergroup differences among the three AADs groups (all p > .05); these parameters were not collected for HC.

TABLE 1.

Demographic characteristics of participants enrolled in validation cohorts.

Characteristics	Validation cohort (n = 125)
	HC	AR	AS	ARA	p Value a
Sample size	40	26	27	32	ND
Gender, male (N, %)	19, 48%	18, 69%	20, 74%	22, 69%	.092
Age (years) b , medians (IQR)	9 (6–11)	7 (4–12)	10 (6–15)	10.5 (7–14)	.125
Range (years)	1–18	1–18	2–18	3–18	ND
BMI	ND	20.08 (16.90–23.96)	16.50 (15.37–20.05)	19.98 (15.37–23.01)	.281
Total IgE (kU/L)	22.39 (10.61–38.18)	131.3 (75.4–325.1) #	370.2 (94.05–538.1) #	280.9 (144.6–986.1) #	<.0001
PFT (N, %)	ND	7, 26.92%	17, 62.96%	14, 43.75%	ND
FEV1 (pred%)	ND	88.37 (85.07–105.5)	90.01 (83.19–102.5)	94.89 (87.29–105.0)	.784
FVC (pred%)	ND	91.84 (91.20–98.50)	89.61 (84.22–106.6)	96.69 (92.23–108.5)	.317
FEV1/FVC (%)	ND	83.45 (83.18–84.10)	84.34 (83.53–85.36)	84.32 (83.68–85.08)	.561
Venous blood count
Neutrophils (×109 L−1)	ND	3.41 (3.08–4.31)	3.18 (2.40–4.55)	2.73 (2.37–4.55)	.542
Eosinophils (×109 L−1)	ND	0.44 (0.31–0.57)	0.27 (0.17–0.43)	0.30 (0.19–0.38)	.077
Basophils (×109 L−1)	ND	0.04 (0.03–0.05)	0.03 (0.02–0.05)	0.03 (0.02–0.04)	.115
Lymphocytes (×109 L−1)	ND	2.38 (1.81–3.11)	2.79 (2.13–3.37)	2.77 (2.41–3.66)	.356
Monocytes (×109 L−1)	ND	0.48 (0.30–0.59)	0.30 (0.40–0.51)	0.44 (0.39–0.62)	.372

Note: Data are presented with the number (percentage) (N, %) or median (IQR). Chi‐square test was used for categorical variables and Student t‐test or Mann–Whitney test for continuous variables; ^# p < .0001 against HC. BMI was not routinely recorded for healthy volunteers in the screening protocol. PFT, n (%): number and percentage of children who successfully completed pulmonary function testing.

Abbreviations: AR, allergic rhinitis; ARA, allergic rhinitis with asthma; AS, asthma; HC, healthy control; IQR, interquartile range; N, number; ND, not determined.

^a p‐value among HC, AR, AS, and ARA.

^b Values are presented as medians (25–75 interquartile range).

3.2. Metabolic profiling and classification of carboxylic acids

To enhance detection sensitivity and coverage of carboxylic acids, SPCSDCarboxyl was applied. A total of 330 metabolites were identified and structurally annotated using ClassyFire, yielding six superclasses, 13 classes, and 34 subclasses of metabolites (Figure 1A; Table S2). Most metabolites were classified as lipids and lipid‐like molecules or organic acids and derivatives, with fatty acyls and carboxylic acids as the predominant classes.

FIGURE 1.

Identification and classification of 330 metabolites. (A) Metabolic profiling and classification of carboxylic acids. (B) Distribution of metabolites by pathway. (C) Pathway enrichment and topological mapping. IPath analysis (iPath 3.0; https://pathways.embl.de/) was used to visualize metabolic pathways. Statistical analysis: Student's t‐test on log‐transformed data with FDR correction. FDR, false discovery rate.

Differential metabolites (DMs) were identified using a combined threshold of VIP >1.0, |fold change| > 1.2 and FDR q < .05 in comparisons between AADs and HC. Eighty‐two metabolites were increased, and four were decreased (Figure 1B; Table S2). iPath analysis based on KEGG annotations was used to visualize the pathway distribution of the identified metabolites, showing predominant involvement of fatty acid–related pathways (Figure 1C).

3.3. Identification and replication of DMs in the training study

A total of 330 carboxylic acid metabolites were detected and quantified. Among these, 228 metabolites in the training cohort and 181 in the replication cohort met the predefined differential‐selection criteria of VIP >1.0, |fold change| > 1.2, and FDR‐adjusted q < .05 in at least one pairwise comparison among HC, AR, AS, and ARA. These DMs were summarized in Tables S3 and S4, respectively.

In the training cohort, principal component analysis (PCA) revealed a discernible trend of global metabolic separation between AADs and HC, with partial overlap among allergic phenotypes (Figure S1). Pairwise OPLS‐DA models demonstrated clear discrimination between each AAD phenotype and HC (Figure 2A–C), showing strong explanatory and predictive performance (R²Y = 0.937–0.964; Q² = 0.795–0.856). For the three‐group comparison of AR, AS, and ARA, PLS‐DA showed moderate discrimination (R²Y = 0.770; Q² = 0.588) (Figure 2D). Similar patterns were observed in the replication cohort, although the predictive performance was weaker, with pairwise OPLS‐DA models yielding R²Y = 0.797–0.950 and Q² = 0.569–0.791, and the three‐group PLS‐DA model yielding R²Y = 0.499 and Q² = 0.257 (Figure S2). All models showed significant permutation test results (all p < .001) and CV‐ANOVA results (p < .001) (Table S5).

FIGURE 2.

Multivariate analysis of metabolic profiles. (A) OPLS‐DA comparing AR to HC. (B) OPLS‐DA comparing AS to HC. (C) OPLS‐DA comparing ARA to HC. (D) PLS‐DA including AR, AS, and ARA. (E) Venn analysis identifying DMs among AR, AS, and ARA. (F) KEGG and Reactome pathways were significantly associated with DMs in AR, AS, and ARA versus HC. Enrichment analysis was performed using the RaMP‐DB metabolite set library (3694 metabolites integrated from KEGG via HMDB, Reactome, and WikiPathways). Differential metabolites were identified using Welch's t‐test for two‐group comparisons or Kruskal–Wallis test (k‐test) for multiple groups, with significance set at p < .05 and FDR correction. VIP > 1.0 and |fold change| > 1.2 were used to select metabolites. HC, healthy control; AR, allergic rhinitis; AS, asthma; ARA, allergic rhinitis with asthma; OPLS‐DA, orthogonal partial least squares discriminant analysis; PLS‐DA, partial least squares discriminant analysis; DMs, differential metabolites; VIP, variable importance in projection.

Venn diagram analysis identified 21 metabolites consistently increased across AR, AS, and ARA in both cohorts (Figure 2E; Table S6), including 11 metabolites not previously reported in allergic diseases (Table 2). Pathway enrichment analysis indicated significant involvement of omega‐6 fatty acid metabolism in allergic airway diseases compared with HC (Figure 2F), with arachidonic acid and 13,14‐dihydro‐prostaglandin F1α identified as common DMs. Undecanoic acid and dodecenedioic acid were enriched in free fatty acid receptor‐related signaling and α‐linolenic acid metabolism. Seventeen DMs could not be mapped to annotated pathways in KEGG, Reactome, or WikiPathways (Table 2).

TABLE 2.

DMs shared among multimorbidity for AR, AS, or ARA vs. HC in training and replication cohorts.

Metabolites	HMDB_ID	KEGG_ID	Type	Metabolism pathway/lipidmaps	Pre‐reported in allergy	Pre‐reported in inflammation
Arachidonic acid	HMDB0001043	C00219	up	Omega‐6‐fatty acids in senescence	Yes	Yes
13,14‐dihydro Prostaglandin F1α	HMDB0005076	NA	up	Omega‐6‐fatty acids in senescence	Yes	Yes
Undecanoic acid	HMDB0000947	C17715	up	Free fatty acid receptors	Yes	Yes
Dodecenedioic acid	HMDB0000933	C16308	up	alpha‐Linolenic acid metabolism	Yes	Yes
Hydroxynonanoic acid	HMDB0031513	NA	up	NA	Yes	Yes
Ala Lys Gln Thr	NA	NA	up	NA	Yes	Yes
Arg Ala Ala	NA	NA	up	NA	Yes	Yes
Asp Ile Lys Lys	NA	NA	up	NA	Yes	Yes
Glu Ile Arg‐1	NA	NA	up	NA	Yes	Yes
Glu Ile Arg isomer	NA	NA	up	NA	Yes	Yes
Decylpropane‐tricarboxylic acid	HMDB0031193	C19806	up	NA	No	No
3‐Acetyl‐6‐oxo‐heptanoic acid	HMDB0340970	NA	up	NA	No	No
Hydroxydecanedioic acid isomer‐2	HMDB0000350	NA	up	NA	No	No
Hydroxydodecanedioic acid	HMDB0000413	NA	up	NA	No	No
Hydroxydodecanedioic acid isomer	HMDB0241289	NA	up	NA	No	No
Hydroxy‐hexadecandioic acid	NA	NA	up	NA	No	No
Hydroxyundecanedioic acid	NA	NA	up	NA	No	No
Methyl‐trihydroxyoctadecenoate	HMDB0302290	NA	up	NA	No	No
Trihydroxy‐octadecenoic acid	HMDB0030936	NA	up	NA	No	No
Undecanedicarboxylic acid	HMDB0002327	NA	up	NA	No	No
Undecanedicarboxylic acid isomer‐2	NA	NA	up	NA	No	No

Note: Classification was based on targeted PubMed searches combining each metabolite name with allergy‐, asthma‐, atopy‐, and inflammation‐related terms. HMDB_ID: Human Metabolome Database Identifier; KEGG_ID: Kyoto Encyclopedia of Genes and Genomes Identifier; Pre‐reported, previously reported in allergy or inflammatory conditions (PubMed literature search); NA: not applicable. The differential metabolite analysis was conducted using Welch's t‐test for comparisons between two groups and the Kruskal–Wallis test (k‐test) for multiple groups, with a significance threshold set at p < .05 and false discovery rate (FDR) correction applied. Metabolites were selected based on a VIP >1.0, |fold change| > 1.2 and FDR q < .05, indicating meaningful biological variation.

Group‐specific DMs were observed, with 4, 4, and 15 uniquely altered metabolites identified in AR, AS, and ARA, respectively (Figure 2E; Table S6). Pathway analysis revealed distinct enrichment patterns across groups (Figure S3; Table S7). Tryptophan metabolism was the most significantly enriched pathway in AR alone, with 4‐(2‐aminophenyl)‐2,4‐dioxobutanoic acid was identified as key DMs within this pathway. In AS alone, benzoate degradation and arginine/proline metabolism were enriched, with benzoic acid and 4‐Hydroxyproline as key DMs. In ARA, free fatty acid receptor–related pathways predominated, featuring nonanoic acid (p = 6.84 × 10⁻²⁷, log₂FC = 4.08) and valeric acid (p = 3.83 × 10⁻⁶, log₂FC = 1.08) as key DMs. Additionally, 4‐(2‐aminophenyl)‐2,4‐dioxobutanoic acid and hydroxy‐kynurenine were increased as key DMs when comparing AS to AR.

3.4. Metabolic distinctions and phenotype‐specific biomarkers in ARA

Venn analysis of pairwise comparisons identified two core differential metabolites, valeric acid and erythronic acid, which were significantly altered in ARA compared with HC, AR, and AS (Figure 3A). This pattern was reproduced in both the training and replication cohorts (Table S7).

FIGURE 3.

Identification of ARA‐specific DMs and related pathways. (A), Venn diagram showing the overlap of DMs identified by pairwise comparisons: ARA vs. HC, ARA vs. AR, and ARA vs. AS. Two ARA‐specific DMs, valeric acid and erythronic acid, were commonly altered across all three comparisons. (B), Enrichment analysis of the top 25 metabolic pathways using the RaMP‐DB library, highlighting key pathways including amino acid metabolism, olfactory signaling, and BCAA (leucine, isoleucine, and valine) metabolism. (C–H), Violin plots showing serum relative levels of valeric acid (C), erythronic acid (D), heptanoic acid (E), nonanoic acid (F), undecanoic acid (G), and isovaleric acid (H) across HC, AR, AS, and ARA groups in the training cohort. Statistical analysis was performed using unpaired Welch's t‐tests. ****p < .0001, ***p < .001, **p < .01. HC, healthy control; AR, allergic rhinitis; AS, asthma; ARA, allergic rhinitis with asthma; DMs, differential metabolites; BCAA, branched‐chain amino acid.

Metabolite set enrichment analysis revealed multiple significantly enriched (Figure 3B; Table S8), with prominent involvement of fatty acid–related processes and amino acid metabolism, including pathways involving ketoleucine, 4‐hydroxyproline, and indole‐3‐acetic acid. Pathways associated with signaling and sensory‐related processes, including olfactory signaling (involving isovaleric acid and heptanoic acid) and GPCR‐related signaling (involving valeric acid), were among the most significantly enriched pathways. Additional significantly enriched pathways leucine, isoleucine, and valine metabolism (involving isovaleric acid and ketoleucine), branched‐chain amino acid (BCAA) catabolism (involving ketoleucine), and free fatty acid receptor‐related pathways (involving valeric acid). Erythronic acid was not mapped to any significantly enriched pathway identified in Table S8.

Further targeted analysis confirmed that valeric acid, heptanoic acid, nonanoic acid, undecanoic acid, and isovaleric acid were significantly increased in the ARA group compared with the HC group, with levels progressively increasing from HC to AR/AS and peaking in ARA (Figure 3C–H).

3.5. Airway pathology and serum OCFAs in AAD models

Phenotypic assessment confirmed distinct upper‐ and lower‐airway features among the murine AAD models (Figure S4). AR and ARA mice showed evident nasal mucosal histopathological changes, whereas AS and ARA mice exhibited greater methacholine‐induced airway responsiveness and increased BALF eosinophil and neutrophil counts. Pathological examination of lung tissue stained with H&E showed that the wild‐type (WT) group (Figure 4A) had normal airway structure with minimal inflammation, while the AR (Figure 4B) and AS (Figure 4C) groups displayed mild to moderate inflammation. The ARA group (Figure 4D) exhibited the most severe changes, with significant airway alterations and increased inflammation. Metabolomics analysis revealed differences in serum samples among the groups (Figure 4E). Serum OCFA levels of valeric, heptanoic, nonanoic, and undecanoic acid were singificantly higher in the ARA mice than in WT mice (p < .05). However, no significant differences in these OCFA were observed in lung tissue or BALF across groups.

FIGURE 4.

Airway histopathology and OCFA levels in mouse models of allergic airway disease. (A–D) Representative hematoxylin and eosin (HE)‐stained lung sections from WT, AR, AS, and ARA mice. (E) Violin plots showing relative levels of OCFAs (valeric acid, heptanoic acid, nonanoic acid, and undecanoic acid) in serum, lung tissue, and BALF across the four groups. OCFA, odd‐chain fatty acid; WT, wild‐type; AR, allergic rhinitis; AS, allergic asthma; ARA, allergic rhinitis with asthma; BALF, bronchoalveolar lavage fluid.

3.6. Candidate metabolites and diagnostic performance in the validation cohort of ARA group

To validate the candidate biomarkers identified in the training and replication cohort, serum concentrations of five OCFAs and L‐alanine were quantified in an independent validation cohort using targeted UHPLC–MS/MS. Consistent with the discovery dataset, serum concentrations of heptanoic acid, nonanoic acid, undecanoic acid, and isovaleric acid were significantly elevated in patients with allergic airway diseases compared with HC, whereas valeric acid was most abundant in the ARA group and exhibited greater discriminatory potential for the comorbid phenotype (Figure 5A–E). In contrast, L‐alanine levels were significantly reduced across all disease groups relative to HC (Figure 5F).

FIGURE 5.

Validation of candidate serum metabolites and diagnostic model construction in pediatric ARA. (A–F) Serum concentrations of C5:0, C7:0, C9:0, C11:0, iC5:0, and L‐alanine in HC, AR, AS, and ARA. Data are presented as mean ± SD. Unpaired Welch's t‐tests were used for group comparisons. Statistical significance is indicated as *p < .05, **p < .01, ***p < .001, and ****p < .0001. (G) Heatmap of Spearman correlation coefficients between candidate metabolites and clinical characteristics, including inflammatory cell counts, serum IgE, pulmonary function parameters, age, and body mass index. (H) LASSO coefficient profiles for feature selection to discriminate ARA from AR. (I) LASSO coefficient profiles for feature selection to discriminate ARA from AS. (J) ROC curves for individual candidate metabolites for distinguishing ARA from HC. (K) ROC curves of LASSO‐selected metabolites and their combination for discriminating ARA from AR. (L) ROC curves of LASSO‐selected metabolites and their combination for discriminating ARA from AS. AUC, area under the curve; HC, healthy controls; AR, allergic rhinitis; AS, asthma; ARA, allergic rhinitis with asthma; BMI, body mass index; Baso, basophil count; C5:0, valeric acid; C7:0, heptanoic acid; C9:0, nonanoic acid; C11:0, undecanoic acid; iC5:0, isovaleric acid; ROC, receiver operating characteristic; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; Lym, lymphocyte count; Mono, monocyte count; Neu, neutrophil count; Eos, eosinophil count.

Correlation analysis revealed significant associations between the candidate metabolites and clinical characteristics. Specifically, valeric acid was positively correlated with forced expiratory volume in 1 s (FEV1) (r = .42, p < .05) and negatively correlated with basophil counts (r = −.45, p < .01), whereas heptanoic acid was positively associated with neutrophil counts (r = .45, p < .01) and monocyte counts (r = .38, p < .05) (Figure 5G), suggesting potential links between OCFA metabolism, pulmonary function, and inflammatory responses. In addition to the OCFA candidates, L‐alanine was included in targeted validation as a complementary metabolite related to amino acid metabolism. Correlation analysis showed that L‐alanine levels were inversely associated with several OCFAs, including C7:0, C9:0, and C11:0, suggesting coordinated metabolic alterations involving amino acid and odd‐chain fatty acid metabolism (Figure S5).

Feature selection using LASSO regression identified valeric acid and undecanoic acid as the dominant OCFA variables distinguishing ARA from AR and AS (Figure 5H,I). Receiver operating characteristic analysis demonstrated excellent performance of heptanoic acid, nonanoic acid, and undecanoic acid for discriminating ARA from HC, with area under the curve (AUC) values of 0.96 (95% confidence interval [CI]: 0.91–1.00), 0.98 (95% CI: 0.93–1.00), and 0.94 (95% CI: 0.88–0.99), respectively. These metabolites achieved sensitivities of 92.5%, 97.5%, and 87.5%, together with specificities of 100.0%, 100.0%, and 96.9%, respectively (Figure 5J; Table 3). Notably, valeric acid exhibited the strongest ability to distinguish ARA from isolated AR (AUC = 0.72, 95% CI: 0.59–0.85), with a sensitivity of 65.4% and specificity of 78.1%. Combining valeric acid with undecanoic acid modestly improved the classification performance (AUC = 0.74, 95% CI: 0.61–0.87) (Figure 5K; Table 3). Valeric acid achieved AUC = 0.64 for ARA versus AS, whereas undecanoic acid AUC = 0.75; their combination AUC = 0.74 (Figure 5L; Table 3). Multivariable logistic regression adjusting for IgE showed that valeric acid and undecanoic acid remained significantly associated with ARA from AR, whereas undecanoic acid remained significant for ARA versus AS; valeric acid was not significant in the latter comparison (Figure S6).

TABLE 3.

Diagnostic performance of serum OCFAs for differentiating ARA from HC and single‐disease phenotypes.

Compare groups	Compounds	AUC	95% CI	Sensitivity (%)	Specificity (%)
ARA vs. HC	Valeric acid	0.72	0.60–0.84	75.00	75.00
	Heptanoic acid	0.96	0.91–1.00	92.50	100.00
	Nonanoic acid	0.98	0.93–1.00	97.50	100.00
	Undecanoic acid	0.94	0.88–0.99	87.50	96.88
	Isovaleric acid	0.91	0.85–0.98	80.00	90.63
ARA vs. AR	Valeric acid	0.72	0.59–0.85	65.38	78.13
	Undecanoic acid	0.64	0.49–0.78	38.46	93.75
	Combined	0.74	0.61–0.87	69.23	78.13
ARA vs. AS	Valeric acid	0.64	0.49–0.78	55.56	75.00
	Undecanoic acid	0.75	0.62–0.87	70.37	68.75
	Combined	0.74	0.62–0.87	74.07	65.63

Abbreviations: AR, allergic rhinitis; ARA, allergic rhinitis with asthma; AS, asthma; AUC, area under the receiver operating characteristic curve; CI, confidence interval; HC, healthy control.

Collectively, these findings establish an OCFA‐associated biomarker signature for pediatric ARA. While heptanoic acid, nonanoic acid, and undecanoic acid demonstrated excellent performance in distinguishing ARA from HC, valeric acid uniquely discriminated ARA from isolated AR, highlighting its potential utility for identifying children at risk of asthma comorbidity.

3.7. Quantitative OCFA data support a propionyl‐CoA–centred metabolic axis in ARA

To integrate individual OCFA findings into a cohesive metabolic framework, we analyzed the co‐elevation patterns and phenotype specificity of identified fatty acids across discovery cohorts. In the training cohort, all four straight‐chain OCFAs showed elevation in ARA compared to HC. Valeric acid (C5:0) had a log₂ fold‐change of 1.08 (p = 3.83 × 10⁻⁶), and nonanoic acid (C9:0) had a log₂ fold‐change of 4.08 (p = 6.84 × 10⁻²⁷). Heptanoic acid (C7:0) and undecanoic acid (C11:0) also exhibited similar co‐directional increases (Figure 3C–H; Table S3). The consistent elevation of all four straight‐chain OCFAs from C5 to C11 suggests a sequential β‐oxidation‐driven interconversion within a shared propionyl‐CoA–primed biosynthetic pathway, rather than isolated dysregulation of individual fatty acids.

Among the metabolites analyzed, valeric acid (C5:0) showed the most significant phenotype‐specific increase in ARA compared to AR and AS, despite having the smallest absolute fold‐change relative to HC (log₂FC = 1.08). This pattern of modest overall elevation, yet selective enrichment in ARA over single‐disease phenotypes, highlights its unique ability to distinguish ARA from AR (AUC = 0.72, Table 3). In contrast, other OCFAs, while achieving high AUC for distinguishing ARA from HC, are broadly elevated across AR and AS.

Isovaleric acid (iC5:0) is a branched‐chain C5 fatty acid, not a straight‐chain OCFA. However, its consistent co‐elevation with the straight‐chain OCFA signature in both discovery cohorts (Figure 3H) and its biosynthesis from leucine catabolism via propionyl‐CoA align with the shared metabolic framework. This mechanistic coherence justifies its inclusion in the integrated ARA metabolic signature.

L‐alanine levels were significantly lower in all disease groups compared to healthy controls, with the most pronounced reduction observed in ARA (Figure 5F). As a key pyruvate‐derived gluconeogenic amino acid, its systemic decrease suggests a shift of pyruvate‐derived carbon substrates towards mitochondrial anaplerosis rather than gluconeogenesis. This pattern aligns with the proposed propionyl‐CoA axis. The combined findings of quantitative OCFA co‐elevation, phenotype‐specific valeric acid increase, and L‐alanine depletion define a propionyl‐CoA–linked metabolic axis characterizing the ARA endotype, as illustrated in Figure 6.

FIGURE 6.

OCFA–centered, propionyl‐CoA–linked metabolic reprogramming in the AADs phenotype. Propionyl‐CoA, generated from BCAA catabolism, can enter the tricarboxylic acid (TCA) cycle via an anaplerotic pathway (propionyl‐CoA → methylmalonyl‐CoA → succinyl‐CoA) or serve as a primer for fatty acid synthase (FASN) to initiate OCFA synthesis. In ARA, OCFAs (C5:0, C7:0, C9:0, and C11:0) form a linked remodeling axis, consistent with interconversion via sequential cycles of β‐oxidation. Among these, C5:0 shows the largest relative increase in ARA, indicating preferential enrichment of terminal OCFA products within the coordinated lipid‐remodeling network. AADs, Allergic airway diseases; ARA, allergic rhinitis with asthma; BCAA, branched‐chain amino acid; CoA, coenzyme A; FAO, fatty acid β‐oxidation; FASN, fatty acid synthase; PCC, propionyl‐CoA carboxylase; TCA, tricarboxylic acid; C5:0, valeric acid; C7:0, heptanoic acid; C9:0, nonanoic acid; C11:0, undecanoic acid.

4. DISCUSSION

This study adopted high‐sensitivity carboxylomics technology and conducted comprehensive metabolic profiling of serum samples from two clinical centers to analyze metabolic alterations in pediatric AR, AS, and ARA. Using animal models and targeted metabolomics, we confirmed that ARA constitutes a metabolically distinct disease endotype, rather than merely an overlap between AR and AS. Relative to HC, all allergic airway phenotypes exhibited profound perturbations in lipid metabolism, with ARA uniquely characterized by coordinated metabolic remodeling involving OCFAs, BCAA metabolism, and lipid oxidation pathways. Furthermore, circulating OCFAs exhibited strong diagnostic utility for ARA, with valeric acid identified as a robust metabolite marker that distinguishes the ARA metabolic endotype from AR single‐allergy disease phenotypes.

These results indicate that alterations in circulating OCFAs were broadly shared across AADs phenotypes, whereas valeric acid as selectively elevated in ARA compared with the single‐disease groups. Rather than reflecting uniform metabolic shifts across allergic conditions, our data indicate a progressive transition from limited pathway perturbations in AR and AS to a multi‐pathway metabolic disturbance in ARA. Importantly, the magnitude and coordination of these alterations in ARA exceed those observed in either single disease alone, arguing against a simple additive effect of AR and AS and supporting the interpretation of ARA as a biologically distinct metabolic state. ²⁴ , ²⁵ , ²⁶ , ²⁷ This distinction is consistent with recent multimorbidity‐focused studies proposing that combined allergic phenotypes emerge through shared but amplified molecular programs rather than through linear disease overlap. ²⁷ , ²⁸ At the pathway level, fatty acid metabolism emerged as a shared metabolic backbone across allergic phenotypes, with consistent enrichment of omega‐6 fatty acid pathways, including arachidonic acid and its downstream mediators in AR, AS, and ARA. These findings are consistent with accumulating metabolomic evidence implicating dysregulated lipid metabolism in allergic airway inflammation. ¹³ , ¹⁵ , ¹⁸ Collectively, the data demonstrate that shared inflammatory lipid pathways alone are insufficient to stratify clinically distinct allergic endotypes and that ARA represents a metabolically reorganized phenotype within the allergic airway spectrum.

Metabolite set enrichment analysis also highlighted olfactory signaling and leucine, isoleucine, and valine metabolism pathways, driven largely by heptanoic acid, isovaleric acid, and valeric acid. While olfactory receptors and other metabolite‐sensing GPCRs are increasingly recognized as functional regulators in airway epithelial and immune cells, ²⁹ , ³⁰ , ³¹ pathway enrichment reflects coordinated statistical associations rather than direct evidence of receptor activation. Accordingly, these results should be interpreted as defining a signaling‐relevant metabolic context that warrants functional validation, consistent with recommended caution in pathway‐based metabolomic inference. ³² , ³³ The elevation of isovaleric acid and enrichment of olfactory signaling and leucine, isoleucine, and valine metabolism pathways in ARA further suggest coordinated immunometabolic remodeling. Although isovaleric acid has been reported in various inflammatory conditions, its selective amplification in ARA compared with both AR and AS supports phenotype‐specificity rather than nonspecific inflammation. ¹⁴ , ³³ , ³⁴ Moreover, the concurrent involvement of lipid and amino acid metabolism aligns with recent immunometabolic frameworks that propose that chronic allergic inflammation is accompanied by the integrated rewiring of multiple metabolic axes. ¹⁴ , ³⁴

Circulating OCFAs represent a previously underrecognized metabolic signature of allergic airway disease. Prior studies primarily linked OCFAs to cardiometabolic traits, demonstrating inverse associations with diabetes and cardiovascular risk. These fatty acids are thought to originate from dietary intake and gut microbiota‐related metabolism. ³⁵ , ³⁶ In this study, we show that medium‐chain OCFAs (C5:0–C11:0) progressively increase across AR, AS, and ARA, with the highest levels observed in ARA and strong discriminatory performance for disease stratification. Previous research has largely focused on classical short‐chain fatty acids in allergic inflammation. ³⁷ , ³⁸ In contrast, our analysis extends this framework to longer OCFAs and their downstream catabolic products. One study indicated that, compared to healthy controls, patients with asthma exhibited a lower frequency of propanoic acid exhalation and a higher frequency of nonanoic acid. ³⁹ Although OCFAs may originate from multiple sources, including dietary intake, microbial metabolism, and endogenous β‐oxidation, the consistent phenotype‐dependent gradient observed across independent cohorts suggests that the elevation in ARA is unlikely to result solely from dietary variation. ²⁹ , ³⁷ , ³⁸ These findings extend OCFA biology beyond metabolic disorders to allergic airway endotypes and suggest their potential value as systemic biomarkers of disease severity.

Conventional biomarkers, including serum TIgE, blood eosinophil counts, and FeNO, primarily reflect type‐2 immune sensitization but have limity ability to distinguish airway disease phenotypes. Although TIgE differed significantly across the four groups, substantial overlap was observed among AR, AS, and ARA. Importantly, after adjustment for TIgE, valeric acid remained independently associalted with ARA status in the ARA versus AR compained, indicating that its associalted with ARA was not fully explained by IgE‐related sensitization intensity. Valeric acid also showed moderate discrimination of ARA from isolated AR (AUC = 0.72), suggesting that may provide metabolic information beyond conventional type‐2 inflammatory markers. Furthermore, OCFA profiling requires only a standard venous blood sample and does not depend on patient cooperation, which may facilitate metabolic phenotyping in young children who cannot reliably perform spirometry or FeNO measurement.

Notably, valeric acid was positively associated with FEV1 and inversely associated with basophil counts. Although direct evidence linking valeric acid to airway function is currently limited, previous studies have demonstrated close relationships between fatty acid metabolism and pulmonary function in chronic airway diseases. ⁴⁰ Furthermore, fatty acid‐derived metabolites are increasingly recognized as important regulators of innate and adaptive immune responses. The positive correlations of heptanoic acid with neutrophil and monocyte counts observed in our cohort are therefore consistent with the emerging concept that altered lipid metabolism contributes to systemic inflammatory regulation. ⁴¹ Interestingly, the inverse relationship between valeric acid and basophil counts has not been previously reported and may represent a novel feature of the ARA‐associated OCFA signature, warranting further mechanistic investigation. Valeric acid may modulate IgE‐dependent basophil degranulation via affinity receptors (FcεRI) signaling, warranting investigation.

The inverse association between L‐alanine depletion and several OCFA alterations suggests that ARA‐related metabolic remodeling extends beyond lipid metabolism to broader amino acid and mitochondrial metabolic pathways. Together, these coordinated changes point to a potential propionyl‐CoA–centred metabolic axis (Figure 6). Propionyl‐CoA, generated primarily from BCAA catabolism, can enter the TCA cycle via methylmalonyl‐CoA and succinyl‐CoA, ⁴² supporting mitochondrial anaplerosis, or serve as a primer for FASN‐mediated OCFA synthesis. ⁴³ This dual routing integrates amino acid catabolism, anaplerotic carbon flux, and lipid biosynthesis within a single metabolic framework. ⁴⁴ , ⁴⁵ Altered L‐alanine levels may reflect changes in pyruvate‐derived carbon utilization; however, this proposed metabolic linkage requires further mechanistic validation. ⁴⁶ The predominance of these alterations in serum suggests that ARA is characterized by systemic rather than airway‐restricted metabolic reorganization, highlighting the propionyl‐CoA axis as a hypothesis‐generating framework for future investigation.

Several limitations of this study warrant consideration. First, the study population was restricted to HDM‐monosensitized children, and two boundaries of generalizability remain untested: whether the OCFA signature is specific to the allergic airway spectrum or is also altered in non‐allergic pediatric respiratory and inflammatory conditions (e.g., non‐allergic rhinitis, post‐infectious cough); and whether the profile differs in children with polysensitization, food allergy, or atopic dermatitis remains uncertain because these populations were not included in the current cohorts. Second, several key baseline variables were not systematically collected, including fasting state at blood draw, dietary habits, gut microbiota composition, and current or recent use of inhaled or systemic corticosteroids, antihistamines, or leukotriene receptor antagonists; because circulating OCFAs are influenced by both diet and gut‐microbial metabolism, these unmeasured factors represent potential confounders that future prospective studies should incorporate. Third, the murine ARA model employed dual‐route antigen administration (intranasal and intratracheal), resulting in higher total antigen exposure than the single‐route AR and AS protocols. Although nasal histology, airway responsiveness, and BALF inflammatory profiles supported the intended airway phenotypes, the murine findings should be interpreted as supportive rather than dosimetrically equivalent validation of the human findings. Behavioral assessments, such as sneezing and nasal‐rubbing scores, were not performed and remain a limitation of the model. Fourth, the cross‐sectional design of the clinical study precludes causal inference and does not permit assessment of OCFA dynamics during disease progression or in response to allergen immunotherapy. Finally, functional validation of the propionyl‐CoA axis and valeric acid–mediated signaling was not performed; dedicated in vitro and in vivo mechanistic studies are needed to establish causality and to evaluate whether targeting this pathway holds therapeutic relevance in pediatric allergic airway disease.

5. CONCLUSION

Through high‐sensitivity carboxylomics applied across three independent pediatric cohorts and a murine model, this study establishes that ARA constitutes a metabolically distinct endotype unified by coordinated elevation of circulating OCFAs and underpinned by a propionyl‐CoA–linked metabolic axis encompassing branched‐chain amino acid catabolism and fatty acid oxidation. This reorganization is not additive of isolated AR and AS components but represents a qualitative metabolic transition, consistent with ARA as a biologically distinct immunometabolic state rather than a simple disease co‐occurrence. Valeric acid was the only OCFA specifically elevated in ARA above the single‐disease phenotypes among the identified biomarkers, uniquely discriminated ARA from isolated AR that existing clinical tests, including serum IgE, blood eosinophil counts, and FeNO were unable to reliably achieve, supporting its potential as a serum reference marker for phenotype stratification; combination with undecanoic acid or clinical parameters may further improve net diagnostic performance. Future studies should examine OCFA trajectories longitudinally alongside disease progression, validate biomarker performance in ethnically diverse cohorts, and pursue mechanistic studies of valeric acid signaling in airway epithelial and innate immune cells. Importantly, OCFA profiling requires only a standard serum sample and is independent of patient cooperation, making it applicable to the youngest children, precisely the age group in which early identification of asthma comorbidity offers the greatest window for preventive intervention. Integration of OCFA profiling with existing pediatric allergy assessment pathways could ultimately enable precision endotyping, earlier therapeutic targeting, and more individualized management of pediatric allergic airway disease.

AUTHOR CONTRIBUTIONS

Yida Zhang: Methodology; data curation; validation; writing – original draft. Min Jiang: Investigation; supervision. Xiqing Bian: Methodology; conceptualization; data curation; investigation; supervision; funding acquisition. Junxiong Chen: Methodology; data curation; validation; formal analysis. Qing Wang: Methodology; data curation; investigation; formal analysis; resources. Peiyan Zheng: Writing – original draft; writing – review and editing; funding acquisition; investigation; methodology; data curation; validation; formal analysis; project administration. Jian‐Lin Wu: Conceptualization; data curation; investigation; validation; writing – review and editing; resources; funding acquisition. Na Li: Writing – review and editing; resources; funding acquisition. Baoqing Sun: Writing – review and editing; project administration; resources; funding acquisition. Ying Wu: Methodology; data curation; investigation; validation; supervision.

FUNDING INFORMATION

This study was supported by the emergency key project of Noncommunicable Chronic Diseases‐National Science and Technology Major Project (2024ZD0533200, 2024ZD0533205), Zhongnanshan Medical Foundation of Guangdong province (ZNSXS‐20220083), Cultivation Project of the First Affiliated Hospital of Guangzhou Medical University (2022111708151837, ZH202105), The talent Development Foundation of The First Dongguan Affiliated Hospital of Guangdong Medical University (PU2024002).

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflicts of interest.

ETHICS STATEMENT

All experiments involving the training and replication cohorts adhered to the guidelines set by the Medical Ethics Committee (approval number: gyfyy‐2022‐135). For the validation cohort, the experiments complied with the guidelines established by the Medical Ethics Committee (approval number: ES‐2025‐052‐02). All participants provided written informed consent for the collection and use of their samples. Animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of The First Affiliated Hospital of Guangzhou Medical University (approval number: 20250478A).

CONSENT

Written informed consent for publication was obtained from all participants or their legal guardians, as documented in the Ethics approval and consent to participate statement above.

Supporting information

Figure S1.

Table S1.

Zheng P, Zhang Y, Bian X, et al. A circulating odd‐chain fatty acid biomarker signature discriminates pediatric allergic rhinitis with asthma from related allergic airway phenotypes. Pediatr Allergy Immunol. 2026;37:e70483. doi: 10.1111/pai.70483

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in ZhengPeiyan2026_20260814_011340_mwtab.txt at https://metabolomicsworkbench.org/data/, reference number 7988.