# Characterizing hub biomarkers for metabolic-induced endothelial dysfunction and unveiling their regulatory roles in EndMT through RNA sequencing and machine learning approaches

## TL;DR

This study identifies key genes and long non-coding RNAs linked to metabolic-induced endothelial dysfunction and atherosclerosis using RNA sequencing and machine learning.

## Contribution

A novel EndMT-associated gene diagnostic signature with high accuracy for atherosclerosis diagnosis is developed and validated.

## Key findings

- 306 mRNAs and 523 lncRNAs were differentially expressed in high glucose serum-exposed cells.
- A machine learning model using CD36, ISG15, HSPB2, and IRS2 achieved an AUC of 0.997 for atherosclerosis diagnosis.
- LINC002381, VIM-AS1, and ELF-AS1 lncRNAs were significantly upregulated in peripheral blood samples.

## Abstract

Metabolic disorder and endothelial dysfunction (ED) are key events in the development and pathophysiology of atherosclerosis and are associated with an elevated risk of Cardiovascular disease (CVD). The pathophysiology remains incompletely understood.

Leftover serum samples were collected and stored at −20 °C until study. Serum specimens were mixed to obtain pooled high glucose serum (GLU group) (11.97 ± 2.09 mmol/L); pooled elevated low-density lipoprotein serum (LDL group) [3.465 (3.3275, 3.6425 mmol/L)]; pooled high triglycerides serum (1.15 ± 0.35 mmol/L) (TG group); Subsequently, Human umbilical vein endothelial cells (HUVECs) were exposed to culture media supplemented with these pooled serum or control serum for 72 h. Whole transcriptome sequencing was performed to characterize gene expression profiles and data were analyzed using GSEA, GO, KEGG. qPCR was used to validate the gene expression.

A total of 306 mRNAs and 523 lncRNAs were identified as differentially expressed in the GLU group, 335 mRNAs and 471 lncRNAs in the LDL group, and 364 mRNAs and 562 lncRNAs in the TG group, compared to the control group. These genes are primarily involved in inflammation, lipid metabolism, and EndMT pathways. By integrating differentially expressed mRNA and curated EndMT-related gene sets from the KEGG, GO, and dbEMT2.0 databases, we identified 52 differentially expressed genes associated with EndMT under metabolic stress conditions. Utilizing machine learning techniques, we established an EndMT-associated gene diagnostic signature comprising CD36, ISG15, HSPB2, and IRS2 for the diagnosis of AS, which achieved an AUC of 0.997. The model was subsequently validated across three independent external cohorts (GSE43292, GSE28829, GSE163154), in which it consistently demonstrated strong diagnostic performance, with AUC values of 0.958, 0.808, and 0.884, respectively. The ceRNA networks associated with EndMT are constructed and related lncRNAs including LINC002381, VIM-AS1, and ELF-AS1 were significantly upregulated in peripheral blood samples.

This study identified novel biomarkers for ED. These findings may provide both a potential biomarker and therapeutic target for the prevention and treatment of atherosclerosis and CAD.