Identification of hub biomarkers in coronary artery disease patients using machine learning and bioinformatic analyses

Abstract

Understanding the molecular underpinnings of CAD is essential for developing effective therapeutic strategies. This study aims to identify and analyze differentially expressed hub biomarkers in the peripheral blood of CAD patients. Based on RNA-seq datasets from the Gene Expression Omnibus database, machine learning algorithms including LASSO, RF, and SVM-RFE were applied. Furthermore, the hub biomarkers were enriched to ascertain their roles in immune cell expression and signaling pathways through GO, KEGG, GSVE, and GSVA. An in vivo experiment was conducted to verify the hub biomarkers. Eleven hub biomarkers (ITM2B, GNA15, PLAU, GNG11, HIST1H2BH, SLC11A1, RPS7, DDIT4, CD83, GNLY, and S100A12) were identified and associated with CD8 + T cells and NK cells. They were mainly involved in immune responses, cardiac muscle contraction, oxidative phosphorylation, and apoptotic signaling pathways. Moreover, ITM2B had the most importance and significance to be the biomarker of CAD patients. In conclusion, these findings point to the possibility of ITM2B as a biomarker on the inflammatory pathogenesis of CAD and suggest new options for therapeutic intervention.

Keywords: Apoptosis; ITM2B; Inflammation; NK cell; T cell.

Fig. 1

Identification of hub biomarkers in CAD patients. (A) PCA cluster plot of GSE9820, GSE10195, GSE18608, GSE20686, GSE42148, GSE98583 and GSE66360 before batch effect removal and correction. (B) PCA cluster plot showed that batch effect has been removed. (C) Volcano plot of hub biomarkers between CAD patients and healthy controls. (D) Heatmap for the top 30 hub biomarkers CAD patients and healthy controls. Green, down-regulation (n = 93); Red, up-regulation (n = 36); Grey, not significant.

Fig. 2

Three machine learning algorithms for identifying hub biomarkers. (A-B) LASSO regression algorithm. (C-D) RF algorithm; Mean Decrease Gini score > 2 was used as the threshold to determine whether a gene was selected. (E-F) SVM-REF algorithm. (G) Venn diagram identified eleven hub biomarkers that were shared by three feature selection algorithms.

Fig. 3

Expression of eleven hub biomarkers. (A-K) Violin plot of hub biomarkers expression between CAD patients and healthy controls. (L) Line plot of all the eleven hub biomarkers expression between CAD patients and healthy controls. ^***P < 0.001.

Fig. 4

ROC curves and validation of the eleven hub biomarkers. (A-G) ROC curves of the eleven hub biomarkers. (H) Expression of ITM2B in CAD patients compared to normal controls in the validation dataset (GSE 61144 dataset). (I) ROC curve of the predictive efficacy of ITM2B in the validation dataset. (J) The ITM2B mRNA relative expression. CON, control group. MOD, AS model group. ^**P < 0.01.

Fig. 5

Grouping based on hub biomarker ITM2B. (A) Nine related biomarkers decreased. Green, down-regulation (n = 9). (B) Correlation between ITM2B and the other nine related biomarkers. (C) Heatmap of relationship between the ITM2B and the nine biomarkers.

Fig. 6

GO and KEGG term analysis. (A-B) Bar and bubble plots of GO analysis. (C-E) GOcircos, GOchord, and GOcluster. (F-H) KEGGcircos, KEGGchord, and KEGGcluster. (I-J) Bar and bubble plots of KEGG analysis.

Fig. 7

Gene set enrichment and immune function. (A) The top five pathways with ITM2B high expression by GSEA. (B) The top five pathways with ITM2B low expression by GSEA. (C) Bar plot of differentially expressed biomarkers enriched in the pathways based on ITM2B by GSEA. (D) Scores of immune-related functions between the ITM2B high and low expression groups. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

Fig. 8

Immune cell infiltration analysis. (A) The relative percentage of 22 immune cell subpopulations between CAD patients and healthy controls. (B) Boxplot illustrates different fractions of 22 immune cells in CAD and control samples. (C) Correlation between ITM2B and immune cells. (D-E) ITM2B negatively correlates with activated NK cells and CD8 + T cells. (F) Correlation heatmap of 22 immune cells. Red and blue represent positive and negative correlation, respectively.

Fig. 9

ceRNA network construction. Red rectangle represents the hub biomarker ITM2B. Green triangle represents the miRNAs related to ITM2B and blue oval represents lncRNAs related to miRNAs.