Metabolic subtypes in hypertriglyceridemia and associations with diseases: insights from population-based metabolome atlas

Hypertriglyceridemia (HTG) is a metabolic disorder characterized by elevated triglyceride levels and is often associated with various cardiovascular and endocrine comorbidities [1,2,3]. While prior research has investigated the metabolic heterogeneity of HTG, the clinical implications of these metabolic subtypes and their specific relationships with comorbidities remain inadequately understood. In the present study, we employed untargeted metabolomics in conjunction with clustering analysis to delineate distinct metabolic subtypes of HTG [4, 5]. More importantly, we systematically evaluated their associations with obesity, hypothyroidism, carotid plaque, and hypertension, and further explored the clinical utility of key metabolic biomarkers derived from clustering and correlation analyses. These findings offer novel insights into the metabolic mechanisms underlying HTG-associated comorbidities, providing a robust foundation for precision diagnostics and targeted therapeutic strategies.

We performed untargeted metabolomics and lipidomics profiling on individuals diagnosed with hypertriglyceridemia (HTG) to examine metabolic heterogeneity. The objectives of this study were threefold: (1) to identify metabolic subtypes through the application of unsupervised clustering algorithms, (2) to explore the associations between metabolic clusters and clinical comorbidities, and (3) to identify key metabolic biomarkers using machine learning models. This methodology integrates metabolic profiles with clinical phenotypes, providing insights into disease-specific metabolic pathways and facilitating the development of personalized management strategies.

A cohort of 61 patients diagnosed with hypertriglyceridemia was enrolled in the study. The study protocol was approved by the institutional ethics committee (approval number: SWYX: NO. 2020-264), and all participants provided written informed consent. Comprehensive documentation of their clinical characteristics was conducted, encompassing body mass index (BMI), markers of thyroid function, blood pressure measurements, and imaging data pertinent to carotid plaque evaluation. Through untargeted metabolomic and lipidomic profiling, a total of 390 metabolites were identified, spanning a wide array of classes such as fatty acyls, glycerolipids, glycerophospholipids, organic compounds, and sphingolipids. These metabolites were systematically mapped onto critical metabolic pathways, including those involved in lipid metabolism, energy production, and inflammatory signaling (Fig. 1).

Fig. 1

Study workflow of metabolic subtyping in hypertriglyceridemia (HTG). The schematic outlines the study design, including untargeted metabolomics and lipidomics profiling, unsupervised clustering analysis, clinical correlation assessment, and biomarker identification

To investigate the metabolic heterogeneity among patients with HTG, we employed the k-means clustering algorithm on a metabolomic dataset, resulting in the identification of three distinct metabolic clusters, as evidenced by clear separations in Principal Component Analysis (PCA) (Fig. 2A). Cluster 1 was characterized by elevated levels of fatty acyls and glycerolipids, suggesting disruptions in fatty acid metabolism and glycerolipid biosynthesis pathways. Cluster 2 exhibited significant enrichment in glycerophospholipids and organic compounds, indicating potential alterations in cell membrane turnover and phospholipid signaling pathways. Conversely, Cluster 3 displayed higher levels of sphingolipids, which are closely associated with inflammatory responses and the regulation of oxidative stress. The proportional distribution of these metabolic classes across the clusters is illustrated in Fig. 2B, while a heatmap visualization (Fig. 2C) highlights distinct metabolite abundance profiles within each cluster. The findings indicate that HTG is not a uniform condition; instead, it comprises distinct metabolic subtypes, each characterized by specific metabolic dysregulations. These variations may account for the diverse clinical manifestations and differential responses to therapeutic interventions associated with HTG.

Fig. 2

Identification and characterization of metabolic subtypes in HTG (A) Principal Component Analysis (PCA) plot showing the separation of three metabolic clusters identified based on the results of K-means unsupervised clustering analysis (B) Proportional distribution of major metabolite classes across the three clusters, including fatty acyls, glycerolipids, glycerophospholipids, and sphingolipids (C) Heatmap illustrating the abundance profiles of metabolites within each cluster, showing distinct metabolic patterns

To elucidate the clinical relevance of the identified metabolic clusters, we assessed their associations with significant clinical conditions, including obesity, hypothyroidism, carotid plaque, and hypertension (Fig. 3A). Statistical significance was evaluated using Fisher’s Exact Test. Cluster 2 demonstrated a robust correlation with carotid plaque (P = 0.036), indicating a metabolic profile that may predispose individuals to vascular dysfunction and plaque formation. In contrast, Cluster 3 exhibited a notable association with hypothyroidism (P = 0.018), suggesting metabolic disturbances that could be related to impairments in thyroid hormone synthesis and signaling. In contrast, no significant associations were observed between the metabolic clusters and obesity (P = 0.793) or hypertension (P = 1.000). The metabolic characterization revealed distinct profiles for each cluster; Cluster 2 was characterized by elevated concentrations of carnitine derivatives (CAR 12:0;O, CAR 14:0;O) and triglycerides (TG 50:0), metabolites that have been previously associated with lipid accumulation and vascular inflammation. Conversely, Cluster 3 displayed increased levels of bilirubin, CAR 2:0, and 1-hexadecanol, indicating potential disruptions in heme metabolism and fatty acid oxidation pathways. These findings emphasize the unique metabolic profiles associated with the observed clinical correlations, underscoring the potential for cluster-specific biomarkers to inform personalized management strategies for patients with HTG. The identification of these metabolic subtypes not only advances our comprehension of the heterogeneity within HTG but also establishes a foundation for precision medicine approaches, facilitating tailored interventions based on disease-specific metabolic signatures.

Fig. 3

Clinical Associations and biomarker identification in HTG metabolic clusters (A) Associations between the three metabolic clusters and clinical conditions, including obesity, hypothyroidism, carotid plaque, and hypertension (B, C) Receiver Operating Characteristic (ROC) curves for the top five metabolites associated with carotid plaque (B) or hypothyroidism (C), selected through Random Forest analysis

To elucidate the principal metabolites influencing the associations with these diseases, we utilized Random Forest analysis followed by Receiver Operating Characteristic (ROC) curve analysis to assess their diagnostic efficacy. In the context of carotid plaque (Fig. 3B), the highest-ranking metabolites were identified as CAR 12:0;O, CAR 14:0;O, and TG 50:0, exhibiting Area Under the Curve (AUC) values between 0.803 and 0.894, indicative of substantial diagnostic performance. Conversely, for hypothyroidism (Fig. 3C), the most informative biomarkers were Bilirubin, CAR 2:0, and 1-Hexadecanol, with AUC values ranging from 0.514 to 0.786, reflecting moderate diagnostic utility. These metabolites are primarily implicated in lipid oxidation, heme degradation, and mitochondrial energy metabolism, underscoring their mechanistic significance in the pathophysiology of hypertriglyceridemia-associated comorbidities. These findings not only identify potential biomarkers for disease evaluation but also enhance our understanding of the metabolic pathways underlying these clinical phenotypes, thereby facilitating the development of targeted therapeutic strategies.

This research highlights the efficacy of unsupervised metabolomics clustering in identifying distinct metabolic subtypes among patients with HTG, each defined by unique metabolic profiles and clinical phenotypes. Through the identification of key biomarkers, including CAR 12:0;O, TG 50:0, and Bilirubin, which are associated with conditions such as carotid plaque and hypothyroidism, our findings offer promising candidates for early disease assessment, risk stratification, and personalized therapeutic interventions. Despite the promising findings, the study is limited by a modest sample size and cross-sectional design. Future research involving larger cohorts and longitudinal studies is essential to validate these biomarkers and establish causal relationships.

In conclusion, we identified three distinct metabolic subtypes in HTG patients, each with unique metabolic signatures and associations with carotid plaque and hypothyroidism. Key metabolites, including CAR 12:0;O, TG 50:0, and Bilirubin, emerged as promising biomarkers. These findings highlight the potential of metabolomics-driven clustering for advancing precision medicine strategies in HTG management.