To systematically investigate the human metabolic profiles over time, we performed a comprehensive integrative multi-omics analysis of 101 participants in the Swedish SciLifeLab SCAPIS Wellness Profiling (S3WP) program over two years [28]. Whole-genome sequencing was performed at baseline for all participants. Extensive phenotyping of the individuals was conducted every three months in the first year and at approximately a 6-month interval in the second year, which included plasma metabolome and lipidome profiling, proteome profiling, clinical measurements, and detailed lifestyle questionnaires covering physical activity, mental health, substance intake, and other environmental factors, alongside blood sample collection, to capture seasonal fluctuations and provide a robust temporal perspective (Fig. 1a and Additional file 2: Table S1). Using a combination of GC–MS and LC–MS technologies, we identified a total of 527 metabolites in the study. These metabolites were classified into nine main classes, covering a wide range of lipids (n = 335, 63.6%), amino acids (n = 77, 14.6%), xenobiotics (n = 37, 7.0%), peptides (n = 17, 3.2%), carbohydrates (n = 20, 3.8%), cofactors and vitamins (n = 16, 3.0%), nucleotides (n = 14, 2.7%), energy (n = 7, 1.3%), and other metabolites (n = 4, 0.8%). The metabolites were further categorized into 63 subclasses based on their functions and biochemical characteristics (Fig. 1a and Additional file 2: Table S2, see Methods for more details).
To explore the dynamic patterns of these identified metabolites, we conducted a clustering analysis applying K-nearest neighbor (KNN), shared nearest neighbor (SNN) and Louvain algorithms and revealed that the 527 measured metabolites could be classified into eight major clusters (Fig. 1b-c, Additional file 1: Figure S2 and Additional file 2: Table S3). Functional enrichment analysis based on customized class and pathway terms was further performed for each cluster to identify shared pathways among groups of metabolites exhibiting similar variation patterns. (Fig. 1d and Additional file 2: Table S3). Specifically, cluster 1 showed a co-regulation of pathways involved in fatty acid metabolism, including steroids and their derivatives, as well as monoglycerides. Cluster 2 mainly included metabolites related to the metabolism of dietary components such as caffeine and benzoates. Cluster 3 included pathways central to the urea cycle, specifically the arginine and proline metabolism. It also involved the metabolism of amino acids like methionine, cysteine, glycine, serine, and theronine, which are crucial for nitrogen balance and protein turnover [63]. Cluster 4 exhibited similar changed patterns to Cluster 3, featured by metabolites like dipeptides as well as the metabolisms of arginine, aspartate, and gamma-glutamyl amino acids, which are mainly involved in protein synthesis and amino acid recycling [64]. Cluster 5 mainly consisted of metabolites essential for the cell membranes composition and signaling, including sphingomyelin, ceramide, and phosphatidylcholine [65]. Cluster 6 showed the interconnected pathways of glycolysis, gluconeogenesis, and sugar metabolisms, including the metabolism of fructose, mannose, and galactose. Cluster 7 included metabolites related to energy storage and mobilization, such as triglycerides and diglycerides [66]. Cluster 8 comprises lyso-phosphatidylcholine, lyso-phosphatidylethanolamine, and phosphatidylcholine, key components involved in cell membrane remodeling [67].
Individual and seasonal variations in plasma metabolome profilesWe assessed the variability in the concentration of each metabolite over time by analyzing both inter-individual and intra-individual variations, calculated using the coefficient of variance (CV; Fig. 2a and Additional file 2: Table S4). Notably, our analysis revealed that the variability between individuals for each measured metabolite was greater than the variability observed within the same individual, with ratios of inter-individual to intra-individual CV ranging from 1.09 to 8.86. Among these metabolites, pyrrole-2-carboxylic acid and picolinic acid showed the most significant differences between individuals (Fig. 2a). This suggests that, despite the fluctuations in each person’s metabolic profile over time due to various environmental factors, each individual maintained a distinct metabolomic signature.
Fig. 2Inter- and intra-individual variability of plasma metabolites and seasonal influences. a The inter-individual and intra-individual variations of plasma metabolite levels, calculated as the coefficient of variation (CV) for each metabolite within each visit and across all participants, color-coded by metabolite classes. b Seasonal variation analysis of plasma metabolites using amplitude analysis, color-coded by metabolic classes. Y-axis showing the adjusted p-values with multiple test corrections using Benjamini and Hochberg method. c Heatmap showing the scaled levels of 121 metabolites with significant seasonal variations across 12 months. d-g Plasma metabolite levels throughout the year for metabolites in Cluster M1-M4. h Pathway enrichment analysis of metabolites within each seasonal cluster. Significantly enriched pathways were defined with adjusted P-values < 0.05 (Fisher’s exact test with multiple test corrections using using Benjamini and Hochberg method). i Vitamin D concentration levels across 12 months during the study period. j Succinylcarnitine levels over 12 months as an example of metabolites in Cluster M3. Each line represents an individual; red lines indicate females and blue lines indicate males. Regression lines are added using trigonometric functions
Interestingly, some metabolites showed seasonal patterns which partially contributed to the intra-individual variations. To explore the seasonal influences on metabolite variability and identify potential seasonal patterns, we performed an association analysis between metabolite levels and the month of sampling, with sex, age and BMI considered as covariates. A total of 121 metabolites showed clear associations with the month of sampling, with amplitudes ranging from 0.016 to 0.541 (Fig. 2b and Additional file 2: Table S5). Hierarchical clustering of these metabolites further showed four distinct seasonal patterns: extremely low levels from January to March (Cluster M1, n = 18); relatively low levels during the winter months (Cluster M2, n = 18); relatively high levels during winter (Cluster M3, n = 49); extremely high levels in December (Cluster M4, n = 36) (Fig. 2c-g and Additional file 2: Table S6). The metabolites in Cluster M1, exhibiting the lowest levels from January to March, were enriched in carbohydrate metabolism pathways (Fig. 2h and Additional file 2: Table S7). As an example, fructose reached its lowest levels in January and February (Additional file 1: Figure S3a). This could reflect a post-holiday shift to baseline dietary habits or a reduction in the consumption of sugars. Conversely, the elevation of energy-related pathways in Cluster M4 during December may indicate an increased caloric intake during the festive period or an adaptive physiological response to colder temperatures. Metabolites in Cluster M2, with relatively low levels in winter, were enriched in pathways associated with NAD+ signaling, nicotinate and nicotinamide metabolism, purine metabolism, and phosphatidylethanolamine biosynthesis (Fig. 2h). This could reflect a reduced requirement for or availability of these pathways’ end products during the winter, possibly due to changes in diet, reduced exposure to sunlight affecting vitamin D synthesis, which in turn affects NAD+ synthesis [68]. Interestingly, the seasonal pattern of vitamin D closely correlated with the metabolites in Cluster M2 (Pearson P = 0.04827, Fig. 2i). In contrast, Cluster M3 metabolites, which were relatively high during winter, were associated with amino acid metabolism, specifically glycine and serine metabolism, as well as pathways involved in ammonia recycling and the metabolism of methionine and homocysteine (Fig. 2h). For instance, the highest level of cysteine was observed during winter (Additional file 1: Figure S3b). These increased levels of amino acid metabolism could suggest a shift in fuel utilization towards amino acid catabolism for energy production and could also be associated with lower levels of physical activity as reported by Pedersen EB, et al [69]. Interestingly, the energy metabolism-related fatty acid succinylcarnitine in Cluster M3, which is involved in carnitine synthesis and utilization pathway, also showed a significantly high inter-individual variation, indicating varying levels of energy metabolism among individuals (Fig. 2j).
We further analyzed the seasonal variations of plasma proteins and found that some plasma proteins also exhibited seasonal patterns, with two opposite seasonal clusters observed: Cluster P1, with low levels during the colder season (Cluster P1, n = 51), and Cluster P2, with relatively low expression levels during the warmer months (Cluster P2, n = 312) (Additional file 1: Figure S3c-e and Additional file 2: Table S8). As an example, 46 proteins in Cluster P2 were found to closely correlate with the seasonal patterns of metabolites in Cluster M3. These proteins primarily function in amino acid, glycan, and fatty acid metabolism. Specifically, angiopoietin-like protein 4 (ANGPTL4), which regulates lipoprotein lipase and has been found to be influenced by dietary fatty acids in both human muscle [70] and mice heart [71], showed associations with a group of fatty acids, such as cis-4-decenoylcarnitine and laurylcarnitine (Additional file 1: Figure S3f-g). Along with these fatty acids, ANGPTL4 exhibited relatively lower expression levels from June to September (Additional file 1: Figure S3h). The associations between proteins and metabolites will be described in more depth below.
Sex- and BMI-related divergences in plasma metabolite levelsThe associations between plasma metabolite concentrations, lifestyle factors and clinical measurements were analyzed using canonical correspondence analysis (CCA) that incorporated all 527 metabolites, 13 lifestyle variables, and 43 clinical chemistry and anthropometric variables across visits. Regression analysis of the two CCA components showed that CCA1 was mainly associated with BMI and lipid profiles, whereas CCA2 was more closely related to sex, body composition (muscle and fat percentage), hemoglobin, and urate. Notably, a clear divergence between male and female samples was observed, emphasizing sex as a significant factor influencing plasma metabolomic levels (Fig. 3a). The associations between metabolites and sex were visualized in a volcano plot (Fig. 3b and Additional file 2: Table S9). Among the 119 metabolites that showed significant sex differences, 35 were found to be elevated in females and 84 were more abundant in males (Fig. 3b). In particular, the peptide gamma-glutamylleucine exhibited higher levels in males (Fig. 3c) and has been found to be associated with elevated cardio-metabolic risks [72]. Additionally, 86 BMI-related metabolites were identified (Fig. 3d and Additional file 2: Table S10), with glutamic acid showing the strongest association (Fig. 3e), which aligns with previous reports [73]. Furthermore, the ratio of glutamic acid to other amino acids, such as lysine, ornithine, and hydroxyproline, have been reported as promising biomarkers for identifying metabolically healthy obese individuals [73]. Interestingly, lifestyle factors such as physical activity, stress, and smoking were found to correlate with metabolite levels and show collinearity with sex. In general, the stress levels in females were higher than in males, which was also significantly associated with the elevation of the stress hormone corticosteroid [74]. On the other hand, a higher incidence of smoking among males was associated with the alterations in several metabolites, such as glutamine, which plays a central role in cellular metabolism and function, highlighting the influence of lifestyle factors on metabolic health.
Fig. 3Influence of clinical measurements and lifestyle factors on metabolite levels. a Canonical correspondence analysis (CCA) showing correlations between plasma metabolite levels, clinical measurements and lifestyle variables. The upper and left bar plots show the estimated linear regression coefficients for clinical and lifestyle variables with respect to CCA1 and CCA2. b, d Volcano plots showing the impacts of sex (b) and BMI (d) on plasma metabolite levels (Kenward-Roger approximation with Benjamini and Hochberg correction). c Concentration of gamma-glutamylleucine across four study visits, shown as an example of a sex-associated metabolite. Each line represents an individual; red lines indicate females and blue lines indicate males. e Scatter plot showing the correlation between glutamic acid concentration and BMI, color-coded by BMI
Genome-wide association analysis of the plasma metabolite profileTo investigate the genetic influence on inter-individual differences in plasma metabolite concentration, a GWAS was performed based on individual variation coefficients for 527 plasma metabolites and 6.7 million common genetic variants (minor allele frequency, MAF > 0.05) identified through whole-genome sequencing. A total of 66 significant associations between genetic variants and individual blood metabolite concentrations were identified (P < 2.17 × 10−9, conventional P of 5 × 10−8 adjusted for the number of independent metabolites [4]). Among them, 19 independent metabolite quantitative trait loci (mQTLs) (Linkage Disequilibrium, LD, r2 < 0.1, conditional P < 0.01) involving 22 metabolites were identified (Fig. 4a and Additional file 2: Table S11). Of 19 mQTLs, 4 were pleiotropic genetic variants associated with multiple metabolites. Of these metabolites, 45% were lipids (n = 10) (Fig. 4a-b). Out of the 22 genetically associated metabolites in our study, six have been previously analyzed in other GWAS studies [4, 7]. To validate the associations between these metabolites and genetic variants, a meta-GWAS analysis was conducted for these six metabolites. Most of the genetic loci (8 out of 11) identified from meta-analysis showed the same direction of effects as in our study (Additional file 2: Table S12). Among these, the association between the genetic variant (rs34673751) from Acyl-CoA dehydrogenase short chain (ACADS) and butyrylcarnitine was the most significant in the meta-analysis. The ACADS gene encodes the enzyme short-chain acyl-CoA dehydrogenase (SCAD), which is essential for mitochondrial fatty acid beta-oxidation, while butyrylcarnitine is a short-chain acylcarnitine involved in fatty acid transport and energy metabolism. Our longitudinal analysis further demonstrated that individuals carrying A allele at rs34673751 exhibited stable higher blood butyrylcarnitine levels. Moreover, heterozygous individuals for the protein variant presented intermediate levels of blood butyrylcarnitine compared to the homozygous groups (Fig. 4c-e). To experimentally validate the association between ACADS and butyrylcarnitine, we knocked down the expression of ACADS using siRNA in 293T and HeLa cell lines. Cell viability assays confirmed that ACADS knockdown did not affect cell viability (Additional file 1: Figure S4). Subsequent metabolite analysis revealed that silencing the ACADS gene increased the levels of butyrylcarnitine in both cell types (Fig. 4f). These results provide direct evidence of ACADS’s role in the regulation of butyrylcarnitine levels. Another notable example of the identified mQTLs is the association between metabolite 4-androsten-3alpha,17alpha-diol monosulfate (2), a sulfated steroid and a derivative of androstenediol, and the gene cytochrome P450 family 3 subfamily A member 7 (CYP3A7). (Fig. 4g). The highest association was found for the variant rs45446698, located upstream of the CYP3A7 gene. Individuals carrying a TT homozygote had higher and more stable levels of 4-androsten-3alpha,17alpha-diol monosulfate (2) than individuals carrying a TG heterozygote (Fig. 4h-i).
Fig. 4Genome-wide association analysis of the genetic regulation of the plasma metabolites. a Manhattan plot showing the identified mQTLs in the study. Significant loci are annotated based on the closest gene, with colors indicating the class of the corresponding metabolite. b Chord diagram of loci shared (r2 > 0.2) among metabolites in GWAS study. Line thickness is proportionate to the -Log10(P). c Manhattan plot of butyrylcarnitine showing the genome locations of all associated mQTLs. d Boxplot showing the association between plasma levels of butyrylcarnitine and the genotype of rs34673751, color-coded by the genotype of rs34673751. e Plasma levels of butyrylcarnitine across study visits; each individual is represented by a line; color-coded by the genotype of rs34673751. f Western blot showing increased butyrylcarnitine levels in the ACADS knockdown 293T and HeLa cell lines compared to the control group. *: Kruskal − Wallis test P < 0.05. g Manhattan plot for 4-androsten-3alpha,17alpha-diol monosulfate (2), showing the associated genetic loci. h Boxplot showing the association between plasma levels of 4-androsten-3alpha,17alpha-diol monosulfate (2) and the genotype of rs45446698. i Plasma levels of butyrylcarnitine across study visits; each individual is represented by a line; color-coded by the genotype of rs34673751
Quantification of genetic and non-genetic effects on plasma metabolite levelsTo quantify the influence of genetics, lifestyle, and physiological conditions on metabolite variability, we applied a linear multivariate regression model to each metabolite. This model included all 19 mQTLs, 13 lifestyle factors, 43 anthropometric and clinical chemistry parameters, and visit. In the analysis, the genetic variants were combined as “genetic component”, all the lifestyle-related factors were combined as “lifestyle component”, and all anthropometric and clinical chemistry variables were categorized into 13 clinical classes. A summary of the analysis across all 527 analyzed plasma metabolites (Fig. 5a and Additional file 2: Table S13) showed that the influence of genetics, lifestyle, and physiological conditions on plasma metabolite level varied considerably. Genetics emerged as one of the important factors; out of the 22 metabolites with at least one significant genetic association, 5 metabolites had a genetic contribution greater than 20% (Fig. 5b). The metabolite most affected by genetics was 1,3-Dimethylurate, which is formed from caffeine and can be used as an indicator of caffeine metabolism [75], with 30% of its blood level variance explained by genetic variants. Besides genetics, 469 metabolites were influenced by various clinical factors, with a total contribution greater than 20%. Consistent with the CCA analysis (Fig. 3a), body composition and lipid profiles showed the most significant associations with plasma metabolite levels, with 69 metabolites associated with each of them. In addition, 54 metabolites were associated with urate levels, 48 with kidney function, 36 with glucose homeostasis, 18 with liver function, 17 with heart function, 15 with leukocytes, 6 with the acute phase response, and 28 with other clinical parameters. The top 25 most significant metabolites associated with clinical components were highlighted in Fig. 5c. As an example, a significant association was observed between body composition and pyroglutamylvaline, aligning with previous study that found pyroglutamylvaline to be positively associated with leg muscle [76]. Additionally, multiple metabolites have been identified as being associated with immune cell and red cell populations, as well as inflammatory biomarkers like C-reactive protein (CRP) (Fig. 5c and Additional file 2: Table S13). These results indicated the intricate crosstalk between metabolism, immune response, and hematopoiesis.
Fig. 5Influence of genetic, clinical and lifestyle factors on plasma metabolite level variability. a Overview of influence of genetic, clinical and lifestyle factors on the plasma metabolite level variability. b Barplot showing the variance explanation fraction for each component across all 22 genetic-related metabolites, color-coded by the variable classes. c Barplot showing the top 25 metabolites most strongly associated with clinical components. d Barplot showing the top 25 metabolites most strongly associated with lifestyle components
Furthermore, we identified 39 metabolites associated with various lifestyle factors, with 9 showing an influence from lifestyle factors greater than 20% (Additional file 2: Table S13). These lifestyle-related metabolites included 15 associated with smoking, 8 with changes in housing, 4 with physical activity, 4 with stress, and 9 with other lifestyle factors. The top 25 most significant metabolites associated with lifestyle factors are listed in Fig. 5d. Notably, smoking had the most prominent influence on blood metabolite levels. Among the smoking associated metabolites, glutamine, the most abundant amino acid in the body and considered conditionally essential for critical illness and injury [77, 78], was negatively associated with smoking, along with factors such as 3-pyridinol, an intermediate product in nicotine degradation [79] (Additional file 2: Table S13).
Individual metabolite-protein profiles in human plasmaTo investigate the coordinated fluctuation patterns of plasma metabolites and proteins, we applied linear mixed modeling (LMM) to 527 metabolites and 794 proteins, adjusting for cofounders including subject, visit, sex, age, and BMI. The analysis revealed 5,649 significant protein-metabolite pairs, involving 479 metabolites and 625 proteins, each characterized by a correlation with a false discovery rate (FDR) adjusted P of less than 0.05 (Additional file 2: Table S14). Among these significant associations, 459 involving 121 metabolites and 257 proteins (48.93% of the overlapping metabolite-protein associations) aligned with previous findings from Benson MD, et al. [80], despite differences in molecular measurement platforms. In Fig. 6a we present the 200 most significant protein-metabolite associations to illustrate the complex interplay within the protein-metabolite network. Multiple important hormones, enzymes, receptors and cytokines, such as glucagon (GCG), lipoprotein lipase (LPL), natriuretic peptide (NPPC), and interleukin 10 (IL10), have been identified as hub proteins in the network, highlighting their broad regulatory roles in human metabolism.
Fig. 6Characterization of protein-metabolite network. a Network presenting the top 200 significant protein-metabolite pairs identified by the linear mixed model (LMM) (FDR-adjusted P < 0.05). Solid circles represent proteins in the inner ring, color-coded by protein annotation. Squares represent metabolites in the outer ring, color-coded by different metabolite-related influence factors. Pairs of related proteins and metabolites are connected by dashed lines (indicating correlations supported by LMM results) and solid lines (indicating correlations supported by both the LMM and Mendelian randomization (MR) analysis). Green lines indicate positive correlations between proteins and metabolites in the LMM, while red lines indicate negative correlations. b Scatter plot showing the correlation between aceturic acid concentration and LEP, color-coded by sex. c Scatter plot showing the correlation between 1-Arachidonoylglycerol (1-AG) concentration and MDGA1, color-coded by different genotypes of rs9349050. Pearson correlation coefficients were reported for both b and c. d Manhattan plot for MDGA1 protein, showing the associated genetic variants with plasma levels of MDGA1. One of the most significant SNP (rs9349050) was used as an instrumental variable in the MR analysis. e UMAP clustering of the protein-metabolite profiles of the analyzed samples, color-coded by individual with lines connecting the visits for each individual. f, g Bar plots showing the variance explanation fraction of different genetic, clinical and lifestyle factors, calculated from linear mixed modeling, for UMAP1 (f) and UMAP2 (g). h, i Distribution of betweenness score of proteins (h) and metabolites (i) in three tiers. j, k Dot plots highlighting the Tier 1 proteins (j) and metabolites (k)
Most of these protein-metabolite associations were connected to lifestyle and physiological conditions. For example, we found a significant association between leptin (LEP) and aceturic acid (Fig. 6b). LEP, a hormone secreted by adipose tissue, plays an important role in regulating hunger and energy balance [81], with higher concentrations observed in females [82]. Aceturic acid, also known as N-acetylglycine, is a derivative of the amino acid glycine and has been reported to modulate weight and associated adipose tissue immunity [83]. Our findings suggested a significant association between LEP and aceturic acid with sex stratification.
Subsequent Mendelian Randomization (MR) analyses were conducted to investigate potential genetic drivers of causality between plasma proteins and metabolites. A total of 87 putative causal associations were identified between 38 proteins and 61 metabolites (FDR-adjusted P < 0.05, Additional file 1: Figure S5a, Additional file 2: Table S15). As an example, we detected a significant MR association between MDGA1, a cell surface glycoprotein involved in cell adhesion, migration, axon guidance, and neurodevelopment [84,85,86], and 1-Arachidonoylglycerol (1-AG) (Fig. 6c), a stable regioisomer of the endocannabinoid 2-AG engaged in physiological functions such as emotion, cognition, and neuroinflammation [87, 88]. Our analyses showed that the cis instrumental variable (rs9349050, Fig. 6d) for the MDGA1 protein consistently differentiated both MDGA1 and 1-AG levels in individuals carrying different genotypes (Additional file 1: Figure S5b-c). This stable and positive association between MDGA1 and 1-AG indicated a potential genetic basis for the co-regulation of proteins and metabolites in the nervous system.
Using Uniform Manifold Approximation and Projection (UMAP), we clustered individual molecular profiles based on the integrated metabolite-protein network. We noted that each individual possessed a unique and stable protein-metabolic profile (Fig. 6e). Regression analysis of the two UMAP components (Fig. 6f-g) revealed that UMAP1 was most significantly associated with genetic factors, followed by kidney function, lipid profile, body composition and erythrocytes. UMAP2 showed significant associations with body composition, blood pressure, genetics, and erythrocytes. Additionally, immune response, vitamin D, urate levels, liver functions, and glucose homeostasis were moderately associated with the metabolite-protein profiles. Lifestyle factors such as stress, drug intake, smoking, and physical activity also exhibited minor influences on these profiles.
To identify key proteins and metabolites with the highest connectivity of the protein-metabolite network, we quantified their importance using the centrality betweenness score and categorized them into three tiers based on their deviation from the median level, measured in median absolute deviations (MAD). These categories include Tier 1, high importance (beyond two MADs); Tier 2, moderate importance (beyond one MAD); and Other: low importance (within one MAD). (Fig. 6 h-i). Notably, ANGPTL4 and LEP were the top two proteins with the highest scores (Fig. 6j). ANGPTL4 has been reported as a key regulator in lipid metabolism, primarily by inhibiting the activity of lipoprotein lipase (LPL) [89], and also involved in angiogenesis, vascular permeability, and inflammation processes. In our analysis, we revealed that ANGPTL4 was associated with a broad spectrum of metabolites, involving those related to lipid and glucose metabolism, tissue functions, and immune responses (Additional file 1: Figure S5d). LEP, which is a well-known hormone, has been shown in recent studies to reflect systemic alterations of the human metabolome [90, 91]. In our analysis, we observed that LEP was associated with a spectrum of metabolites related to blood pressure and glucose homeostasis, as well as stress (Additional file 1: Figure S5d). On the other hand, N-Acetylaspartate (NAA), one of the most abundant metabolites in the mammalian brain, and 3-Hydroxyhippurate, a microbial aromatic acid metabolite derived from dietary polyphenols and flavonoids found in normal human urine, were identified as high-importance metabolites through network analysis and found to be associated with kidney function in the study (Fig. 6k, Additional file 2: Table S13).
Variability of protein-metabolite profiles and the associations with metabolic healthTo investigate the associations between protein-metabolite profile variability and metabolic health, we stratified the analyzed samples into two risk groups (high-risk group and low-risk group) based on the clustering patterns of five key indicators of metabolic syndrome: high density lipoprotein (HDL), body mass index (BMI), systolic blood pressure (SBP), triglycerides (TG) and fasting glucose (Gluc) [92]. The high-risk group was characterized by increased levels of BMI, SBP, TG, and Gluc, alongside decreased levels of HDL, in comparison with the low-risk group (Fig. 7a). We then explored the differences in clinical chemistry measurements between these two groups. We found significantly higher levels of two liver function biomarkers (alanine aminotransferase, ALAT; gamma glutamyltransferase, GGT), one kidney function biomarker (urate), one heart function biomarker (troponin T, TNT), and one immune biomarker (white blood cells count, WBC) in the
Comments (0)