Exploring Genetic Associations of Three Types of Risk Factors with Ischemic Stroke: An Integrated Bioinformatics Study

1.1 Data sources

In accordance with the goals of this study, we utilized the IEU database, conducting searches with the keywords of “VLDL (very low-density lipoprotein)”, “LDL (low-density lipoprotein)”, “IDL (intermediate-density lipoprotein)”, “HDL (high-density lipoprotein)”, “apoliprotein”, “triglyceride”, “fatty acid class”, “systolic pressure”, “diastolic pressure”, “blood glucose”, “glycosylated hemoglobin”, and “IS”. This can identify genetic instruments for blood pressure, blood glucose, and blood lipids and their metabolic products as exposure variables. Concurrently, IS-related genetic instruments were selected as outcome variables for an MR analysis, aiming to explore potential causal relationships between exposures and outcomes. By using “ischemic stroke” as the keyword, setting the data type as array-based expression profiles, and specifying the species as homo sapiens, we retrieved samples from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) to obtain gene expression and clinical data of IS patients and healthy individuals. Perl code was employed for gene symbol annotation and data adjustment, thus deriving expression levels of genes related to blood pressure, blood glucose, blood lipids and their metabolites, and IS, in both the normal and IS groups.

1.2 Mendelian Randomization analysis

We utilized the “TwoSampleMR” package to conduct a two-sample MR analysis on the relationship between blood pressure, blood glucose, and blood lipids and their metabolic products with IS.²⁰ In the analysis, we set the parameters as follows: 1) The selected genetic instruments need to have a strong correlation with the exposure factors, and the corresponding P-value should be less than 5 × 10⁻⁸; 2) In conducting Linkage Disequilibrium (LD) clumping, we set the r² threshold at 0.01; 3) When performing clumping analysis, we set the window size as 10,000 kilobase pairs.²¹ These stringent parameter settings help us to accurately reveal the causal relationship between exposures and outcomes.

1.3 Colocalization analysis

After completing the two-sample MR analysis, we selected positive results that demonstrated significant associations between the exposure factors and outcome in the MR analysis. We used the “ieugwasr” and “coloc” packages to conduct colocalization analysis on these positive results to determine if the exposure factors and outcome might be associated within the same gene region. We set a posterior probability (PP4) greater than 50% as the standard, indicating that the genes in that region are associated with both exposure and outcomes¹⁶. Finally, we employed the “gassocplot2” package to visualize the colocalization results, clearly demonstrating the associated shared genes of blood pressure, blood glucose, and blood lipids and their metabolic products with IS.

1.4 Identification of differently expressed associated genes (DEAGs) and analyses of DEAGs

Expression levels of genes associated with blood pressure, blood glucose, blood lipids and their metabolic products, and IS were extracted from both the normal and IS groups. Differential expression analysis was performed using the “limma”, “pheatmap”, “ggpubr” and other R packages, and results were presented as box plots and heatmaps. Genes with a p-value less than 0.05 were defined as differentially expressed associated genes (DEAGs). Using Perl coding, DEAGs were located on chromosomes and their positions were displayed on a circos plot created with the “Rcircos” package. In addition, the “cor” command was used to calculate the correlation coefficient between each two DEAGs and the results were visualized to illustrate the correlation among DEAGs.

1.5 Analysis of immune cells in IS samples

We performed 1000 simulations using the CIBERSORT command in R to obtain a total relative amount of immune cells equal to 1, then visualized the content of immune cells in each sample using a bar plot. Single-sample gene set enrichment analysis (ssGSEA) was performed using the “GSVA” and “GSABase” packages to compare the differences in immune cell contents between the normal group and the IS group, and the results of ssGSEA are presented as box plot. We matched the differentially expressed associated genes (DEAGs) with the ssGSEA scores, performed a correlation test to obtain the correlation coefficient, and then visualized the results as a heat map.

1.6 Identification of feature genes and construction and validation of nomogram based on ML

Expression data from DEAGs was utilized to construct four predictive models: Random Forest model (RF), Support Vector Machine model (SVM), Generalized Linear Model (GL), and Extreme Gradient Boosting model (XGB). The results from these four models were calculated based on the prediction function. Feature genes within DEAGs were then selected through comprehensive analysis of reverse cumulative distribution plots, residual box plots, and receiver operating characteristic (ROC) curves. After selecting the optimal model, line plots were constructed using the expression levels of the feature genes in both the normal and IS groups. Lastly, decision and calibration curves were drawn to assess the accuracy of the line plots. Another dataset that includes both a normal and IS group was obtained from the GEO database, and a machine learning model was constructed using the same R language approach as before. The ROC was plotted to validate the machine learning model constructed in the test dataset.

1.7 Clustering of DEAGs and analysis between DEAG clusters

The “ConsensusClusterPlus” package in R was used for clustering, employing Euclidean distance type and allowing up to nine clusters. Expression levels between clusters were compared using heatmaps and box plots, and principal component analysis (PCA) was applied to assess differences between clusters. Following this, a ssGSEA was performed on the DEAG clusters, generating bar plots of the amount of individual immune cells in each sample within the different clusters and comparing the differences in immune cell content among different clusters. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using gmt files downloaded from the GSEA platform (http://www.gsea-msigdb.org/), and gene set variation analysis (GSVA) was conducted in R language to analyze the expression of enrichment items between clusters. Finally, under the filtering conditions of |logFC| > 1 and an adjusted P-value < 0.05, differential expression analysis was performed on the gene expression of DEAG clusters. Intersection of DEAG clusters through a Venn diagram yielded differentially expressed genes (DEGs).

1.8 Clustering of DEGs and analysis between DEG clusters

The same clustering method as in section 1.7 was applied to cluster DEGs, and the DEG cluster with the highest precision was selected. Based on the DEGs clustering results, we compared the expression levels of DEGs, the differences in DEAGs expression, and the immune cell content in different clusters. These results were visualized using heatmaps and box plots.

1.9 Construction of DEAG scores

The PCA method was employed to calculate the expression levels of DEAGs for each sample, yielding DEAG scores (The formula is shown below).²² Using R packages such as “limma” and “ggpubr”, we performed differential analysis on the scores of DEAGs in both significantly differentially expressed core gene clusters and DEG clusters. We created box plots to illustrate the scores of significantly differentially expressed core genes in samples within the significantly differentially expressed core gene and DEG clusters. In addition, an alluvial diagram was drawn using the package “ggalluvial” to visualize the relationships and overall processes among DEAG clusters, DEG clusters, samples with higher scores, and samples with lower scores of DEAGs.

1.10 Statistical analysis

This study carried out MR analysis and colocalization analysis using R V4.2.1. In the MR analysis, if the number of SNPs for the exposure factor is less than 5 after filtering, we will use the Wald ratio method for analysis; otherwise, we will opt for the inverse variance weighted (IVW) method for analysis. In colocalization analysis, to maximize the acquisition of shared genes between “exposure” and “outcome”, we set the chromosomal locus parameter range from 75 kilobase pairs to 500 kilobase pairs, aiming to include a wider area around the SNP to enhance the likelihood of discovering shared genes related to both exposure and outcome. In the bioinformatics part, we used Strawberry Perl 5.32.1.1 for GEO dataset extraction and data annotation, and all other statistical analyses were performed with R V4.2.1. For two independent samples, we applied a t-test, while for two paired samples we used the Wilcoxon paired-rank sum test, and for three or more groups of data, we employed one-way analysis of variance (ANOVA) and the Kruskal-Wallis rank sum test. The Spearman rank correlation test was used for correlation analysis. We set a P value < 0.05 or a false discovery rate (FDR) < 0.05 after correction by the Benjamini-Hochberg method as being statistically significant.

2.1 Results of collection of GWAS and GEO datasets

We retrieved 516 relevant exposure datasets from the ieu database based on the set keywords, and the IS dataset came from 440,328 participants of European ancestry, including 34,217 cases and 406,111 controls. These datasets were used in two-sample MR analysis for IS to explore their potential causal relationships. By searching the GEO database for IS datasets, we filtered out two datasets that met the requirements, namely GSE16561 and GSE22255, using the former as the test set and the latter as the validation set. The former is total RNA data from human peripheral blood samples, including 39 IS samples (diagnosed by MRI) and 24 normal samples; the latter is RNA expression data from human peripheral blood monocytes, including 20 IS samples and 20 normal samples.

2.2 Results of Mendelian randomization analysis

Among the 516 exposure datasets, 114 showed a positive association with IS. These positive results further suggested that increased levels of Apolipoprotein A-I and HDL may reduce the risk of IS, while elevated levels of VLDL, LDL, IDL, Apolipoprotein B, triglycerides, fatty acids, systolic blood pressure, diastolic blood pressure, fasting blood sugar, and glycated hemoglobin might increase the risk of IS. (Please see Attachment 1 for detailed analysis results)

2.3 Results of colocalization analysis

Upon completion of the two-sample MR analysis, we further subjected the 114 positive results to colocalization analysis to study the potential shared genes between the exposure factors and IS. The results showed that the PP4 was greater than 50% for 70 of the positive results, including exposure factors such as LDL, IDL, VLDL, Apolipoprotein B, fatty acids, systolic and diastolic blood pressure. This suggests that these exposure factors may have genes shared with IS. The colocalization results were visualized using the “gassocplot2” package, and we finally identified 64 shared genes between these exposure factors and IS. These genes may play a key role in the association between the selected exposure factors and IS. (Please see Attachment 1 for detailed analysis results and Attachment 2 for specific shared genes situations)

2.4 Results of DEAGs identification and analysis of DEAGs

Through colocalization analysis, we identified 64 associatied shared genes between blood pressure, blood sugar, and blood lipids and their metabolites with IS. Differential analysis between the IS group and the normal group in the test dataset revealed that 13 associatied shared genes exhibited statistically significant differences. Among them, ERCC2, FEN1, HDDC3, TOMM40, TRAPPC6A were highly expressed in the normal group, while ALDH2, FADS2, FES, FURIN, MAN2A2, PVRL2, SYT7, and FAM109A were highly expressed in the IS group, as shown in Figures A and B. The specific positions of these significantly differentially expressed association genes on the chromosome are shown in Figure C. The correlation analysis between each pair of DEAGs in the IS samples showed that there is a certain correlation between DEAGs, which are mainly positive correlations, as shown in Figures D and E.

2.5 Results of Immune cell infiltration analysis, immune cell difference analysis and correlation analysis of IS samples

Through immune cell infiltration analysis, we obtained the types and content of immune cells expressed in each sample, as shown in Figure A. ssGSEA analysis revealed (Figure B) that the expression differences of 11 types of immune cells in the control group and IS group were statistically significant. Among them, B cells naive, T cells CD8, T cells CD4 memory activated, T cells follicular helper, Natural killer (NK) cells activated, and Dendritic cells activated were highly expressed in the control group, while T cells gamma delta, Macrophages M0, Macrophages M2, Mast cells activated, and Neutrophils were highly expressed in the IS group. The correlation analysis between DEAGs and immune cells showed that there is a certain correlation between DEAGs and immune cells, and the levels of positive and negative correlations are comparable, as shown in Figure C.

2.6 Results of selection of machine learning models, construction of nomogram, and verification

The results of building SVM, RF, XGB, and GL machine learning prediction models using DEAGs data were obtained. From the residual boxplots, inverse cumulative distribution plots, and ROC curves (Figure A, C, and D), it can be observed that the GL method had the highest area under the ROC curve, the lowest residual values, and the lowest inverse cumulative values. Therefore, the GL method was considered the most accurate and chosen as the best model for further analysis. The GL model provided importance scores for the selected feature genes, as shown in Figure B, revealing nine feature genes. Among them, the top five genes with the highest importance scores were used to construct the column line graph (including FURIN, MAN2A2, HDDC3, ALDH2, and TOMM40). Then, separate scoring scales were obtained for these five feature genes. The risk rate of the co-located feature genes associated with blood pressure, blood glucose, and blood lipids and their metabolic products in the occurrence of IS was assessed by calculating the sum of the feature gene expression scores (Figure E). The predictive accuracy of the model was evaluated by the distance between the solid line and the dotted line in the calibration curve (Figure F) and the distance between the red and gray lines in the decision curve (Figure G), indicating high accuracy. The GL model was validated using the validation dataset (GSE22255) and the top five feature genes with the highest importance scores were included in the model validation and ROC curve analysis. The results showed an Area Under the Curve (AUC) value of 0.528, with a 95% confidence interval of 0.167-0.889. Therefore, based on the validation of the model, it can be concluded that the model based on the GEO dataset has good accuracy.

2.7 Results of Clustering of DEAGs, Analysis of DEAGs Expression between Clusters, Immune Cell Analysis, and GO and KEGG Enrichment Analysis

Clustering analysis based on the expression of DEAGs revealed two distinct clusters with the highest accuracy, as shown in Figure A. Consequently, the IS samples were divided into two groups: C1 and C2, as depicted in Figures B and C. Subsequently, DEAGs expression analysis was performed between the two DEAG clusters (Figure D and E), indicating that ALDH2, FEN1, FES, and FURIN were significantly upregulated in C1. PCA analysis (Figure F) demonstrated that DEAGs can differentiate between C1 and C2. Moreover, ssGSEA analysis (Figure G) identified three statistically significant immune cell types between C1 and C2. Specifically, NK cells resting and Macrophages M2 were upregulated in C1, while Mast cells resting were upregulated in C2. The respective content of different immune cells in each C1 and C2 sample is illustrated in Figure H.

In GSVA analysis (Figures I and J), it can be observed that compared to C1, C2 showed upregulation of several GO biological processes, including Interleukin 6 production, positive regulation of interleukin 6 production, cellular response to osmotic stress, neutrophil degeneration, response to lipoteichoic acid, positive regulation of reactive oxygen species metabolic process, and macropinocytosis. On the other hand, C2 exhibited downregulation of branching involved in blood vessel morphogenesis, cellular response to potassium ion, myotube cell development, regulation of heart morphogenesis, canonical Wnt signaling pathway involved in osteoblast differentiation, gonadotropin secretion, vascular process in the circulatory system, and blood vessel maturation in GO biological processes. In GO molecular functions, C2 showed upregulation of leukotriene C4 synthase activity and downregulation of sphingosine 1-phosphate phosphatase activity. In GO cellular components, C2 exhibited upregulation of early phagosome, specific granule lumen, and proton-transporting two-sector ATPase complex catalytic domain, while downregulation of photoreceptor disc membrane. In KEGG pathways, C2 demonstrated upregulation of other glycan degradation, toll-like receptor signaling pathway, sphingolipid metabolism, pentose phosphate pathway, pathogenic Escherichia coli infection, glutathione metabolism, Fc gamma R-mediated phagocytosis, neurotrophin signaling pathway, cell signaling in Helicobacter pylori infection, progesterone-mediated oocyte maturation, and endocytosis. Conversely, C2 showed downregulation of alanine, aspartate, and glutamate metabolism, cytokine-cytokine receptor interaction, taste transduction, ECM receptor interaction, hedgehog signaling pathway, arachidonic acid metabolism, neuroactive ligand-receptor interaction, ascorbate and aldarate metabolism, pentose and glucuronate interconversions, and nitrogen metabolism in KEGG pathways.

2.8 Results of DEGs screening, DEGs clustering, and analyzes of DEG clusters

After filtering for significantly differentially expressed genes (DEGs) between C1 and C2 samples, a total of 4 DEGs were identified (Figure A). These 4 DEGs were further subjected to clustering analysis within the DEAGs cluster, resulting in two distinct clusters, namely CI and CII (Figure B). Analysis of DEAGs expression in these two clusters in IS samples revealed that RNASE2, RNASE3, MMP9, and CAMP exhibited statistically significant differential expression, with higher expression in CI and lower expression in CII (Figure C).

The differential expression analysis of DEAGs based on DEGs clustering was performed on the samples, as shown in Figure D. The results revealed that these genes were mainly upregulated in the CI group and downregulated in the CII group. Among them, the significantly differentially expressed genes included ALDH2, FES, MAN2A2 (upregulated in CI), and TRAPPCA (upregulated in CII). Additionally, using ssGSEA based on DEGs clustering (using the same gene set file as above), statistically significant differences in immune cells were observed, including activated B cells (downregulated in CI), activated dendritic cells, immature dendritic cells, myeloid-derived suppressor cells (MDSCs), macrophages, mast cells, monocytes, neutrophils, and plasmacytoid dendritic cells (upregulated in CI), as shown in Figure E.

2.9 Results of DEAGs scoring, differential analysis of DEAGs score between clusters, and construction of the alluvial plot

The differential analysis of DEGs clustering based on the principal component analysis (PCA) scores showed statistically significant differences between the clusters (Figure A). Specifically, C1 had lower scores, while C2 had higher scores. However, there was no statistically significant difference observed in the DEGs clustering (Figure B). The scatter plot in Figure C displayed that the C1 and C2 clusters of DEAGs corresponded mainly to the CI and CII clusters of DEGs, respectively. However, there was no clear correspondence observed between the high and low scores of DEAGs and the DEGs clustering.

3.1. The relationships of blood pressure, blood glucose, and blood lipids and their metabolic products with IS

Clinical studies have demonstrated that effective management of various physiological indicators can significantly impact the risk of IS. For example, blood pressure control has been shown to significantly reduce the risk of IS recurrence.²³ Elevated fasting blood glucose levels are significantly associated with an increased risk of IS.²⁴ Furthermore, increased levels of glycated hemoglobin, regardless of diabetes status, are associated with an increased risk of IS.²⁵ Lipid management also plays a significant role in IS risk. There are significant differences in VLDL levels between the control and IS groups.²⁶ However, the role and mechanisms of VLDL level changes in IS pathogenesis require further investigation. Medications that lower LDL levels can significantly reduce the risk of IS.²⁷ Increased levels of IDL and apolipoprotein B are associated with an increased risk of IS.^28,29 Controlling triglyceride levels can lower the risk of IS.³⁰ Both polyunsaturated fatty acids and saturated fatty acids are also associated with IS.^31,32 However, elevated levels of HDL and apolipoprotein A-I are associated with a lower risk of IS.^33,34 Taking all the above clinical studies into consideration, it can be concluded that blood pressure, blood glucose, and blood lipids and their metabolic products play critical roles in influencing the risk of IS.

The results of MR analysis indicate a potential causal relationship between blood pressure (diastolic and systolic), blood glucose (fasting blood glucose and glycated hemoglobin), blood lipids and their various metabolic products (VLDL, LDL, IDL, HDL, apolipoprotein A-I, apolipoprotein B, triglycerides, Eicosapentaenoate, Docosapentaenoate, Stearidonate, Docosahexaenoic acid) with IS. Increasing levels of high-density lipoprotein (HDL) and apolipoprotein A-I are associated with a decreased risk of IS. However, elevated blood pressure (including systolic and diastolic), blood glucose (fasting blood glucose and glycated hemoglobin), various low-density lipoproteins such as VLDL, LDL, IDL, as well as apolipoprotein B, triglycerides, Eicosapentaenoate, Docosapentaenoate, Stearidonate, Docosahexaenoic acid are associated with an increased risk of IS.

3.2 Discussion of mechanisms associated shared genes

Through MR analysis, we found that blood pressure, blood glucose, blood lipids and their metabolic products have a significant impact on the risk of IS. To further explore the specific mechanisms underlying these effects, we conducted a colocalization analysis. The results showed that there are 64 associated shared genes between LDL, IDL, VLDL, apolipoprotein B, fatty acids, systolic blood pressure, and diastolic blood pressure with IS. Although a positive association between blood glucose and IS risk has been established, we did not identify any associated shared genes due to the PP4 being less than 50%.

The differential analysis of the 64 associated shared genes between normal and IS samples, using the GEO dataset, revealed 13 genes with significant differential expression. Among them, ERCC2, a transcription factor (TF) involved in nucleotide excision repair of damaged DNA,³⁵ has been clinically associated with increased stroke risk,^36,37 and animal studies have shown its neuroprotective role in preventing ischemia-reperfusion injury.³⁸ FEN1, another TF encoding a protein involved in rDNA and mitochondrial DNA repair,³⁹ has been implicated in the pathogenesis of IS and is considered a therapeutic target for IS treatment.^40,41 HDDC3, a protein encoded by HDDC3 gene, participates in the starvation response, and starvation-induced autophagy is known to protect neurons and regulate IS.^42,43 TOMM40, encoding the translocase of the outer mitochondrial membrane 40 homolog, has been identified as a potential susceptibility locus for IS based on a study conducted in Japan.⁴⁴ ALDH2, encoding aldehyde dehydrogenase, has shown neuroprotective effects by clearing 4-hydroxy-2-nonenal and reducing mitochondrial-associated cell apoptosis through JNK-mediated cystathionine-β-synthase-3 activation, making it a potential target for IS intervention. Case-control studies have also suggested an association between ALDH2 polymorphisms and IS risk in the Han Chinese population.^45–47 FADS2, encoding fatty acid desaturase 2, has been associated with IS risk and lipid levels in case-control studies, although the specific mechanisms are still under investigation.⁴⁸ The product of the FES gene exhibits tyrosine-specific protein kinase activity. Inhibitors of this class of enzymes have shown potential therapeutic value for various diseases, including IS, but the involvement of FES in IS is yet to be confirmed.⁴⁹ FURIN, encoding proprotein convertase subtilisin/kexin type 1, has been found to be upregulated within 24 hours after cerebral ischemia in animal experiments,⁵⁰ and case-control studies have revealed lower DNA methylation levels in the IS region, suggesting its association with IS.⁵¹ SYT7, encoding synaptotagmin 7, has been identified to have significant DNA methylation correlation with IS in a large-scale sequencing study.⁵² The roles of FAM109A, MAN2A2, PVRL2, and others in IS require further investigation.

These genes primarily function through DNA repair, DNA methylation, mitochondrial repair, cell apoptosis, and autophagy to regulate lipid levels, blood pressure, and neuronal protection, thereby affecting the occurrence, development, and prognosis of IS. The associated shared genes are predominantly upregulated in the IS group, and they exhibit mostly positive regulatory relationships, indicating potential synergistic effects among these genes in IS. Differential expression analysis of immune cells between the IS and normal groups reveals significant expression differences in 50% of the immune cells, suggesting the involvement of immune cells in the regulation of IS by these associated shared genes. Moreover, the comparable levels of positive and negative regulation of immune cells by DEAGs suggest that DEAGs may have bidirectional regulatory effects on the immune system’s role in IS.

The feature genes obtained from the GL model constructed through machine learning (FURIN, MAN2A2, HDDC3, ALDH2, TOMM40) have significant importance in regulating the associated shared genes and the risk of developing IS. The column plot provides a quantitative prediction of the aforementioned importance and risk of disease. Furthermore, the construction of the GL model using the validation dataset indicates that the column plot has high accuracy.

The IS samples were clustered into C1 and C2 groups based on the expression of DEAGs, with C1 showing predominantly high expression of DEAGs and C2 showing low expression. The two clusters were significantly distinguishable according to the PCA results. Furthermore, immune cell infiltration analysis of DEAGs clustering revealed significant differences in immune cell expression between C1 and C2 groups of IS patients. The GSVA results indicated significant expression differences between C1 and C2 groups in various biological processes and pathways, including immune inflammation, vascular development and formation, reactive oxygen species metabolism, carbohydrate metabolism, lipid metabolism, fatty acid metabolism, amino acid metabolism, and neural regulation.

Differential analysis between the C1 and C2 groups identified 4 DEGs, which have been extensively studied and shown to be associated with IS incidence, IS prognosis, and IS complications.^53–55 Subsequently, IS samples were further clustered into CI and CII groups based on the expression of DEGs. Differential analysis revealed that DEGs were highly expressed in the CI group and lowly expressed in the CII group, while DEAGs were mainly upregulated in the CI group and downregulated in the CII group. Immune cell infiltration analysis of DEGs clustering demonstrated significant differences in immune cell expression between the CI and CII groups of IS patients, with upregulation observed in the CI group and downregulation observed in the CII group. PCA-based DEAG scoring and subsequent differential analysis revealed significant differences in DEAG scoring between the C1 and C2 groups, while no significant differences were observed between the CI and CII groups.

3.3 Limitations of this study

Although MR analysis and colocalization analysis provide valuable insights into the risk factors and underlying mechanisms of IS from the perspective of exposure factors and genes, these methods also have their limitations. The pathogenesis of IS is extremely complex and involves intricate interactions among multiple environmental factors, physiological indicators, and genetic factors. Therefore, in order to comprehensively understand and elucidate the risk factors and mechanisms of IS, a holistic and multifactorial research approach is needed. Additionally, the GWAS data included in our study primarily originate from sample populations of European ancestry. Hence, extrapolating these results directly to non-European populations may have limitations. Similarly, bioinformatics methods also have certain limitations. Although they have provided multi-faceted and multi-dimensional analyses of human blood samples, elucidating specific mechanisms of interaction between blood pressure, blood glucose, blood lipids and their metabolic products in relation to IS, further research and clinical applications require careful consideration. Moreover, the blood samples used in this study were only collected from the United States and Portugal, and whether the results can be extrapolated to other countries, ethnicities, and populations remains to be investigated and confirmed. Given the relatively small sample size, there is a greater risk of bias in the conclusions. Therefore, large-scale studies or animal experiments can be conducted to further validate and deepen the conclusions.