Metabolic subtypes and biomarkers in preterm and term neonates via targeted screening

Abstract

Preterm infants exhibit metabolic immaturity, yet metabolic heterogeneity within this population remains underexplored. We performed targeted metabolomics on dried blood spots from 448 preterm (32–36 weeks) and 351 term neonates (37–40 weeks of gestation) using tandem mass spectrometry. Compared with term infants, preterm neonates showed significantly elevated tyrosine, leucine/isoleucine, arginine, and hydroxyoctadecenoylcarnitine (C18:1-OH), along with reduced glutamate (false discovery rate < 0.05). Multivariate analyses, including principal component analysis and partial least squares-discriminant analysis, identified three distinct metabolic clusters associated with gestational maturity and redox-related pathway signals. Pathway enrichment analysis highlighted disruptions in the urea cycle, ammonia recycling, purine metabolism, and mitochondrial fatty acid oxidation. Notably, C18:1-OH emerged as a key discriminatory metabolite and a potential biomarker of mitochondrial immaturity and altered fatty acid oxidation in preterm neonates. These findings support the presence of metabolically distinct subtypes within preterm infants and suggest that metabolomic profiling may contribute to precision neonatal risk stratification, although longitudinal validation is required.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-50955-8.

Introduction

Preterm birth, defined as delivery before 37 completed weeks of gestation, accounts for approximately 10% of live births worldwide and remains a leading cause of neonatal morbidity and mortality^1,2. Preterm infants face a unique set of physiological challenges due to the immaturity of multiple organ systems, particularly the liver, kidneys, and metabolic pathways essential for maintaining homeostasis and supporting rapid growth and development.

Metabolic adaptation during the neonatal period is critical for survival and long-term health, yet preterm infants often exhibit dysregulated metabolism characterized by altered amino acid profiles, impaired fatty acid oxidation, and oxidative stress^3,4. These metabolic perturbations have been linked to adverse outcomes including neurodevelopmental delay, bronchopulmonary dysplasia, and metabolic syndrome later in life^5,6.

The advent of high-throughput metabolomics techniques, particularly liquid chromatography-tandem mass spectrometry (LC-MS/MS), has revolutionized our ability to profile complex metabolic changes in biological samples with high sensitivity and specificity^7,8. Antioxidant-related metabolites, including glutathione precursors, are often reported as reduced or insufficient in preterm infants, reflecting immature enzymatic antioxidant systems and increased susceptibility to oxidative stress⁹.

While previous studies primarily focus on distinguishing preterm from term infants based on global metabolic signatures, recent evidence suggests significant heterogeneity within the preterm population itself, indicative of distinct metabolic subphenotypes that may have prognostic and therapeutic implications^10,11. Identifying these metabolic subgroups requires integrative analytical approaches combining multivariate statistics like principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) with pathway enrichment methods to unravel complex biochemical networks^12,13.

We hypothesized that targeted metabolomics of newborn screening data could reveal distinct metabolic subtypes in preterm neonates and nominate novel biomarkers for clinical risk stratification.

Methods

Study population

This observational, cross-sectional study was conducted to compare the metabolic profiles of preterm and term neonates using dried blood spot (DBS) samples collected through the national newborn screening program. Participants were selected from neonates referred to the Metabolic Laboratory of the Growth and Development Research Center (GDRC), Tehran, Iran, between March 1, 2022, and March 20, 2024.

Based on gestational age, infants were categorized into two groups: 448 preterm neonates (233 boys, 215 girls; 32–36 weeks) and 351 sex-matched full-term neonates (37–40 weeks). Feeding type (breast milk/formula) was recorded at the time of sampling. Clinical variables including gestational age (GA), birth weight (BW), birth weight z-score (BW_Z), feeding type (breast milk, formula, or parenteral nutrition), ventilation status (yes/no), and postnatal steroid exposure were extracted from the screening database when available and considered as potential confounders in sensitivity analyses.

Only live-born infants with complete demographic data and confirmed gestational age were included. Exclusion criteria included major congenital anomalies, previously diagnosed inborn errors of metabolism, incomplete clinical data, maternal gestational diabetes, preeclampsia, or antenatal corticosteroid exposure.

Ethics approval and consent

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of the Growth and Development Research Center, Tehran University of Medical Sciences, Tehran, Iran (protocol code IR.TUMS.CHMC.REC.1404.105). Written informed consent was obtained from a parent and/or legal guardian of all neonates included in the study prior to the use of their anonymized dried blood spot samples for research purposes.

Sample collection

Dried blood spot (DBS) samples were collected by heel prick 48–72 h after birth on Whatman 903 filter paper, air-dried for 3 h at room temperature, and transported to the laboratory for analysis.

Targeted metabolic analysis by MS/MS

Targeted metabolic profiling of amino acids and acylcarnitines was performed using a Shimadzu LCMS-8045 triple quadrupole liquid chromatography–tandem mass spectrometry (LC–MS/MS) system (Kyoto, Japan) operating in electrospray ionization (ESI) positive mode under multiple reaction monitoring (MRM) conditions. A validated commercial kit for newborn screening, MassChrom^® Newborn Screening Kit (57000 F, non-derivatized; Chromsystems Instruments and Chemicals, Germany), was used according to the manufacturer’s protocol. Data acquisition, peak integration, and quantitative analysis were conducted using LabSolutions software (version 5.91, Shimadzu Corporation, https://www.shimadzu.com/an/products/software-informatics/labsolutions-series/index.html). Quantification was based on stable isotope-labeled internal standards supplied with the kit, enabling correction for matrix effects and instrumental variability. Calibration curves were generated using kit-provided standards, and analytical performance was monitored using both kit-supplied quality control materials and routine laboratory internal quality assurance procedures. Analytes corresponded to those routinely included in standard newborn screening panels.

Potential confounders, including feeding type (breast milk, formula, or parenteral nutrition) and clinical interventions (e.g., corticosteroid use), were recorded. Breast milk feeding referred exclusively to the infant’s own mother’s milk. Sensitivity analyses showed no significant impact of feeding type on key metabolites in preterm versus term comparisons (p > 0.05 for interaction terms).

Statistical and metabolomic data analysis

Data were analyzed using R software (version 4.2.1, https://www.r-project.org/) and MetaboAnalyst (version 6.0, https://www.metaboanalyst.ca/). Metabolite concentrations were log-transformed and Pareto-scaled before analysis. Univariate comparisons (t-test, Mann–Whitney U) were adjusted using the false discovery rate (FDR) method (q < 0.05).

Multivariate analyses, including principal component analysis (PCA) and partial least squares–discriminant analysis (PLS-DA), were performed to identify global metabolic patterns. PLS-DA models were validated using 5-fold cross-validation (Q² = 0.55, p < 0.001). Volcano plots, hierarchical heatmaps (Euclidean distance, Ward’s linkage), and pathway enrichment analyses using the Kyoto Encyclopedia of Genes and Genomes (KEGG) and the Small Molecule Pathway Database (SMPDB) were used to highlight key metabolites and pathways¹⁴.

To further validate differential metabolite selection beyond conventional univariate testing, Significance Analysis of Microarrays (SAM) and Empirical Bayes Analysis of Microarrays (EBAM) were applied as complementary feature-selection approaches. SAM is a permutation-based method that controls the false discovery rate by estimating a modified t-statistic with an adjustable delta threshold (delta = 0.7). EBAM applies an empirical Bayes framework to stabilize variance estimation and calculate posterior probabilities and local FDR values. Metabolites with FDR-adjusted p < 0.01 were considered significant in both methods. Overlap between SAM- and EBAM-identified metabolites was assessed to confirm robustness of the findings.

To assess the impact of clinical confounders, multivariable linear regression models were performed for key discriminatory metabolites, including hydroxyoctadecenoylcarnitine (C18:1-OH). Covariates included gestational age (weeks), birth weight z-score (BW_Z), feeding type (reference: breast milk), ventilation status (reference: no ventilation), and steroid exposure. Regression analyses were conducted in R, and for multivariable regression analyses, statistical significance was defined as p < 0.05.

All statistical tests were two-tailed, and an FDR-adjusted p-value < 0.05 was considered statistically significant.

Results

Metabolic profiling, conducted via Electrospray ionization–tandem mass spectrometry (ESI-MS/MS), identified significant metabolic differences between preterm and term infants, analyzed through integrated univariate, multivariate, and pathway enrichment approaches.

Univariate analysis and volcano plot

Volcano plot analysis (Fig. 1) highlighted key metabolites with FDR-adjusted p-values. Tyrosine, glycine, and leucine+isoleucine exhibited the most pronounced differences (− log₁₀(FDR-adjusted p ≈ 30)), indicating major metabolic disruptions in preterm infants. Glutamic acid and hydroxyoctadecenoylcarnitine (C18:1-OH) followed with values around 20. Log₂ fold changes for these metabolites were: tyrosine (≈ 2.5), glycine ( ≈ − 1.5), leucine+isoleucine (≈ 2.0), glutamic acid ( ≈ − 1.2), and C18:1-OH (≈ 1.8), reflecting dysregulation in amino acid metabolism and fatty acid oxidation. These findings are consistent with altered mitochondrial fatty acid oxidation and position C18:1-OH as a candidate biomarker of mitochondrial immaturity, pending longitudinal validation.

Fig. 1.

Volcano plot illustrating differential metabolite expression between preterm and term infants, with log2 fold change (x-axis) and -log10(p-value) (y-axis). log₂FC for key metabolites include tyrosine (2.5, -log10(p) ≈ 30), glycine (− 1.5, -log10(p) ≈ 28), and leucine+isoleucine (2.0, -log10(p) ≈ 31). FDR-adjusted p < 0.05.

Univariate analysis using Significance Analysis of Microarrays (SAM) further identified C18:1-OH as one of the most significantly altered metabolites. The SAM-derived q-value was reported as 0 due to the resolution limit of permutation-based estimation and is therefore conservatively presented as FDR-adjusted p < 0.001.

The distribution of C18:1-OH across the three metabolic clusters is illustrated in Supplementary Figure S1. Post-hoc pairwise comparisons using Tukey’s HSD test with FDR correction confirmed significant differences between clusters (Supplementary Table S1).

Supporting univariate findings, SAM and Empirical Bayes Analysis of Microarrays (EBAM) were applied as complementary feature-selection approaches. SAM and EBAM identified 27 and 35 significant metabolites, respectively (Supplementary Figures S2 and S3), with strong overlap between amino acid metabolites and long-chain acylcarnitines, supporting robustness of the findings (FDR-adjusted p < 0.01). The full list of metabolites identified by SAM and EBAM is provided in Supplementary Tables S2 and S3. Notably, the significant metabolites identified by SAM and EBAM largely belonged to two major biochemical families: amino acids and acylcarnitines. Altered amino acids included tyrosine, glycine, glutamic acid, leucine/isoleucine, valine, arginine, citrulline, and ornithine, indicating disruption of amino acid metabolism and urea cycle activity. In parallel, several medium- and long-chain acylcarnitines, particularly hydroxylated species (e.g., C14-OH, C16:1, C18:1-OH, and related derivatives), were significantly altered, supporting impaired mitochondrial fatty acid β-oxidation in preterm neonates. These metabolite families were consistent with the pathway enrichment results presented in Fig. 5.

Fig. 5.

Overview of metabolic pathways identified in the metabolomic profiling of preterm and term infants. Key amino acids (e.g., glycine, serine, arginine, proline, phenylalanine, tyrosine), intermediates (e.g., glutamate, alanine, aspartate), and metabolic processes—including ammonia recycling, urea cycle, purine metabolism, glutathione synthesis, thyroid hormone synthesis, porphyrin biosynthesis, bile acid production, carnitine metabolism, and propanoate degradation—are illustrated. Additional pathways involve catecholamine synthesis, folate metabolism, nicotinamide biosynthesis, and branched-chain amino acids (valine, leucine, isoleucine), as well as beta-alanine, cysteine, lysine, and histidine. Significant alterations observed in pathway enrichment analysis are highlighted.

Multivariate analysis (PCA and PLS-DA)

Principal component analysis (PCA) (Fig. 2) clearly separated preterm from term neonates, with PC1 (26.1%) and PC2 (21.9%) accounting for 48% of the total variance. Three distinct clusters were identified, validated by a silhouette score of 0.45: Cluster 1 (mostly preterm) showed elevated tyrosine, arginine, leucine+isoleucine, and C18:1-OH, but reduced glycine and glutamate — suggestive of hepatic and mitochondrial immaturity. Cluster 2 (mainly term) displayed a mature metabolic phenotype with opposite trends. Cluster 3 represented an intermediate metabolic state, possibly reflecting transitional maturity or nutritional influences.

Fig. 2.

Principal component analysis (PCA) score plot of metabolomic profiles in preterm and term neonates. Each point represents an individual sample, colored by class (preterm vs. term). PC1 (26.1%) and PC2 (21.9%) together explained 48.0% of the total variance. Shaded ellipses indicate three metabolic clusters (Cluster 1, Cluster 2, and Cluster 3), suggesting distinct metabolic subtypes within the study population.

Recursive SVM classification (Supplementary Figure S4) showed that reducing the number of variables increased classification error rates from 14.9% (26 variables) to 28.8% (5 variables), indicating that discrimination is supported by a broader metabolite signature.

Partial least squares discriminant analysis (PLS-DA) (Fig. 3), validated by 5-fold cross-validation (Q² = 0.55, p < 0.001), improved separation between groups. Variable importance in projection (VIP) scores and random forest analysis (Supplementary Figure S5) supported model robustness (error rate < 0.1).

Fig. 3.

Principal component analysis (PCA) scores plot of targeted metabolomic profiles in preterm and term neonates. Each point represents an individual infant (preterm, pink; term, green). The first two principal components (PC1 = 26.1% and PC2 = 21.9%) explain 48.0% of the total variance. Shaded ellipses indicate the distribution of samples for each group in the PCA space, illustrating partial separation between preterm and term metabolic profiles.

Hierarchical clustering and heatmap

Heatmap analysis (Fig. 4; Supplementary Figure S6) offered a detailed view of metabolite concentrations across individual samples and clusters, with a color gradient from green (low concentration) to red (high concentration). Key metabolites, including arginine, glutamate, tyrosine, citrulline, and carnitine derivatives, displayed differential patterns. Arginine was significantly elevated in preterm infants (FDR-adjusted p = 1.22 × 10⁻¹⁰), while glutamate and glycine were notably reduced (FDR-adjusted p < 0.05). Variable trends in tyrosine and C18:1-OH suggest disruptions in thyroid hormone synthesis and fatty acid oxidation, respectively. Hierarchical clustering (Euclidean distance, Ward’s linkage) delineated two major groups, clearly separating term and preterm profiles with distinct upregulation and downregulation patterns.

Fig. 4.

Heatmap of metabolomic profiles. Heatmap generated from ESI-MS/MS data, showing relative concentrations of key metabolites (e.g., arginine, glutamate, tyrosine, citrulline, carnitine species) across preterm and term infants. Rows represent metabolites and columns represent individual samples or grouped categories. Color gradients from green (low) to red (high) highlight elevated arginine in preterm infants (p = 1.22 × 10⁻¹⁰) and reduced glutamate and glycine (p < 0.05). Tyrosine and C18:1-OH show variable red trends in preterm samples, suggesting potential disruptions in thyroid hormone synthesis and fatty acid oxidation. Hierarchical clustering separates preterm and term profiles, revealing distinct upregulation and downregulation patterns.

Adjustment for clinical confounders

To determine whether the elevation of C18:1-OH was confounded by clinical variables, multivariable linear regression analysis was performed. After adjustment for gestational age, birth weight z-score, feeding type, ventilation status, and steroid exposure, none of the covariates were significantly associated with C18:1-OH concentrations (all p > 0.05; Table 1). These findings suggest that the observed differences in C18:1-OH are not fully explained by the measured clinical covariates and may reflect intrinsic developmental metabolic immaturity.

Table 1.

Multivariable linear regression analysis of log-transformed hydroxyoctadecenoylcarnitine (C18:1-OH) concentrations in preterm and term neonates.

Predictor	β (Estimate)	SE	t	p value
Intercept	0.00971	0.00307	3.17	0.0016
Gestational age (weeks)	−0.000044	0.000094	−0.47	0.638
Birth weight z-score	−0.000183	0.000180	−1.02	0.309
Feeding: formula	−0.000157	0.000366	−0.43	0.669
Feeding: parenteral nutrition	0.003904	0.002454	1.59	0.112
Ventilation: yes	0.000503	0.000466	1.08	0.281
Postnatal steroid exposure: yes	−0.000051	0.000634	−0.08	0.936

Regression coefficients (β) were adjusted for gestational age (continuous), birth weight z-score (continuous), feeding type, ventilation status, and postnatal steroid exposure (categorical variables). Reference categories were breast milk feeding, no ventilation, and no steroid exposure. Statistical significance was defined as p < 0.05.

Metabolic pathway and enrichment analysis

Pathway enrichment analysis (Fig. 5), leveraging MetaboAnalyst’s KEGG and SMPDB databases, identified significant alterations in pathways related to the urea cycle, ammonia recycling, purine metabolism, and porphyrin biosynthesis (rather than direct bile acid biomarkers). Elevated urea cycle intermediates (arginine, ornithine) indicated hepatic immaturity and altered nitrogen metabolism in preterm neonates. Reduced glutamate and glycine levels are notable, as these amino acids serve as key substrates for glutathione synthesis, suggesting potential impairment of antioxidant defense mechanisms and increased susceptibility to oxidative stress. Although bile acid-related pathways appeared in the enrichment output, direct bile acid metabolites were not measured in the targeted newborn screening panel; therefore, these pathway-level findings should be interpreted cautiously as indirect signals rather than direct biomarker evidence.

Supplementary boxplots (Supplementary Figure S7) confirmed increased arginine and decreased glutamate levels, with inter-individual variability reflective of metabolic instability. A consolidated overview of the principal discriminatory metabolites and their enrichment-supported pathway associations is provided in Supplementary Table S4.

Discussion

Metabolic alterations in preterm neonates

This study demonstrates that preterm neonates exhibit coordinated alterations in amino acid and acylcarnitine metabolism compared to term infants, reflecting hepatic and mitochondrial immaturity. Rather than isolated metabolite shifts, the overall pattern supports a systems-level perturbation involving nitrogen handling/urea cycle activity, redox balance, and mitochondrial fatty acid β-oxidation.

The characteristic elevation of tyrosine, branched-chain amino acids (leucine and isoleucine), and arginine, together with reduced glycine and glutamate, aligns with prior neonatal metabolomic studies reporting gestational age–dependent differences in amino acids and acylcarnitines and implicating amino acid metabolism and β-oxidation–related pathways^10,15. Large-scale newborn screening metabolomics data further indicate that gestational age and birth weight z-score are major determinants of neonatal amino acid/acylcarnitine profiles and highlight pathway-level signals involving phenylalanine/tyrosine metabolism and urea cycle–related pathways¹⁰.

Conversely, reduced glycine and glutamate are notable because they serve as key substrates for glutathione synthesis, potentially limiting antioxidant capacity and increasing susceptibility to oxidative stress in preterm infants^9,16.

C18:1-OH as a candidate biomarker

Among acylcarnitines, hydroxyoctadecenoylcarnitine (C18:1-OH) emerged as a key discriminatory metabolite and a potential biomarker of impaired mitochondrial fatty acid oxidation. While long-chain (including hydroxylated) acylcarnitines have been linked to mitochondrial dysfunction and developmental metabolic immaturity, our data highlight C18:1-OH as a reproducible differentiator within preterm neonates and across metabolomic subtypes. The observed elevation is more consistent with incomplete β-oxidation/mitochondrial handling during early adaptation than with a specific inherited fatty acid oxidation disorder in this screened cohort, although confirmatory testing remains essential in clinical screening contexts^17,18.

Importantly, the association of C18:1-OH with the identified metabolic patterns remained robust after adjustment for key clinical covariates (Table 1), supporting that C18:1-OH is not merely a proxy for measured early clinical factors in our dataset.

Clinical confounders, early care, and obstetrical context

Because prematurity is frequently accompanied by differences in gestational age, birth weight z-score, feeding exposure, and early intensive care (e.g., invasive ventilation or steroid exposure), we explicitly tested whether these factors could account for variability in C18:1-OH. In multivariable regression models adjusting for gestational age, birth weight z-score, feeding type, ventilation status, and steroid exposure, none of these variables showed an independent association with C18:1-OH concentrations (Table 1). Importantly, these findings indicate that the observed elevation of C18:1-OH and the identified metabolic subtypes cannot be fully explained by measured early clinical variables alone. Rather, they suggest that intrinsic developmental metabolic immaturity may contribute independently to the observed metabolomic patterns.

Nevertheless, detailed obstetrical complications leading to prematurity—such as chorioamnionitis or intra-amniotic infection, preterm premature rupture of membranes, and hypertensive disorders of pregnancy—were not consistently available in the newborn screening registry and therefore could not be incorporated into the regression models. These obstetric phenotypes are biologically distinct and are known to differentially influence fetal inflammatory status, placental function, and metabolic programming^19–21. Accordingly, residual confounding from unmeasured antenatal or placental factors cannot be excluded and represents a limitation of the present study.

Future studies incorporating detailed obstetric phenotyping and placental characterization will be necessary to disentangle inflammatory versus placental vascular etiologies and their differential metabolic consequences.

Feeding modality and early metabolic adaptation

Although feeding modality was recorded and included in adjusted regression models, it was not significantly associated with C18:1-OH concentrations or the identified metabolic clusters in our cohort. However, this finding should be interpreted in the context of early sampling (48–72 h), a period during which neonatal metabolic profiles are strongly shaped by gestational maturity and immediate postnatal metabolic transition^4,10. Early nutrition is known to influence neonatal metabolic adaptation, particularly lipid and amino acid metabolism^22–24.

At the same time, metabolomic differences between breastfed and formula-fed infants may be modest or not yet fully established within the first days of life, and may also vary by biospecimen especially during the immediate postnatal transition period^4,25. Moreover, feeding data available in the newborn screening registry were limited to broad categories (breast milk vs. formula) and did not capture exclusive versus mixed feeding, volume intake, caloric density, or detailed parenteral nutrition composition. Therefore, the absence of a statistically significant association within the 48–72 h sampling window should not be interpreted as evidence of a null biological effect of feeding per se. Rather, it suggests that gestational maturity and early metabolic transition may exert a stronger influence on metabolomic patterns at this early time point. Future studies incorporating longitudinal sampling and more granular nutritional characterization may better delineate feeding-related metabolic effects.

Identification of metabolic subtypes and biological interpretation

Multivariate analyses (PCA and PLS-DA) revealed three distinct metabolic clusters, indicating heterogeneity beyond simple gestational age classification. One cluster predominantly represented term neonates, possibly reflecting a metabolically mature phenotype with improved nutritional adaptation. Cluster 1 (predominantly preterm) was characterized by elevated tyrosine, arginine, leucine/isoleucine, and C18:1-OH with reduced glycine and glutamate, consistent with combined hepatic/mitochondrial immaturity, altered nitrogen handling, and vulnerability to oxidative stress^4,9,26,27. Cluster 2 (predominantly term) reflected a more mature metabolic phenotype, whereas Cluster 3 appeared to represent an intermediate/transitional state, potentially influenced by gestational maturity, sampling time, and early nutritional exposures^11,28,29. The persistence of distinct clusters despite adjustment for measured clinical covariates suggests intrinsic biochemical heterogeneity within preterm neonates that may be clinically relevant.

Pathway-level interpretation and clinical implications

Pathway enrichment analysis indicated alterations in pathways related to urea cycle/ammonia recycling, purine metabolism, and porphyrin biosynthesis, which may be consistent with metabolic immaturity and altered energy/redox physiology^30–32. Elevated arginine and related intermediates support disruption in nitrogen handling and urea cycle–linked metabolism, while decreased glycine and glutamate may compromise glutathione synthesis and antioxidant defenses^9,33,34. Although bile acid–related pathways appeared in pathway enrichment output, direct bile acid metabolites were not measured in the targeted newborn screening panel; therefore, these signals should be interpreted as indirect pathway-level findings rather than direct biomarker evidence.

Collectively, these findings suggest potential strategies for risk stratification and targeted neonatal care. Monitoring C18:1-OH and related long-chain acylcarnitines may help identify infants with greater metabolic immaturity or mitochondrial vulnerability, while nutritional approaches supporting nitrogen handling and antioxidant capacity (e.g., amino acid adequacy for glutathione synthesis) warrant evaluation. Future longitudinal and multicenter studies integrating clinical outcomes and multi-omics will be essential to validate subtype stability and clinical utility^35,36.

Conclusion

This study identified distinct metabolic subtypes in preterm neonates and highlighted hydroxyoctadecenoylcarnitine (C18:1-OH) as a candidate biomarker of mitochondrial immaturity and altered fatty acid oxidation. The findings underscore metabolic heterogeneity among preterm infants and support the concept that early-life metabolomic profiling may contribute to future risk stratification strategies.

However, given the cross-sectional design and early sampling window, longitudinal and multicenter studies integrating clinical outcomes and multi-omics approaches will be required to confirm subtype stability, clarify prognostic relevance, and determine potential clinical applicability.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, Farzaneh Abbasi; Methodology, Maryam Gholami; Data Curation, Saeideh Abdolahpour, Maryam Gholami; Writing – Original Draft Preparation, Saeideh Abdolahpour; Writing – Review & Editing, Farzaneh Abbasi, Saeideh Abdolahpour, Maryam Gholami; Project Administration, Reihaneh Mohsenipour. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. All data will be shared for non-commercial research purposes, in compliance with participant confidentiality and ethical guidelines.

Declarations

Competing interests

The authors declare no competing interests.