Identification and validation of inflammatory response genes linking chronic kidney disease with coronary artery disease based on bioinformatics and machine learning

Figures

• 

  Fig. 1. Flow chart of this study. CAD, coronary artery disease; CKD, chronic kidney disease; DEGs, differentially expressed genes; WGCNA, Weighted gene co-expression network analysis; LASSO, least absolute shrinkage and selection operator; RF, random forest; ROC, receiver operating curve.

• 

  Fig. 2. The procession of CAD dataset and differentially expression analysis of CAD. (A) Box plot of normalization of data from 141 samples. (B) UMAP for GSE113079. (C) The volcano plot illustrating CAD DEGs in GSE113079. Upregulated gene were represented by red dots, while downloaded genes were represented by bule dots. (D) Heatmap of the top 20 upregulated and 20 downregulated DEGs in GSE113079. CAD, coronary artery disease; DEGs, differentially expressed genes.

• 

  Fig. 3. Identification of key module genes in CAD dataset by WGCNA and determination of key genes in CAD via overlapping DEGs and key module genes. (A, B) Identification of the best β value through scale independence and mean connectivity. (C) The dendrogram and module eigengenes. (D) Correlation between WGCNA modules and status of CAD. (E) The intersection of DEGs and key module gens through venn diagram. CAD, coronary artery disease; WGCNA, Weighted gene co-expression network analysis; DEGs, differentially expressed genes.

• 

  Fig. 4. Functional enrichment analysis of key genes in CAD. (A) Bar plots of GO biological process enrichment analysis. (B) Bar plots of GO cell component enrichment analysis. (C) Bar plots of GO molecular function enrichment analysis. (D) The chord plot showing KEGG enrichment pathway results. CAD, coronary artery disease; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

• 

  Fig. 5. The integration of CKD datasets and identification of IRGGs in CKD related CAD. (A, B) PCA plots of two CKD datasets before and after batch-effect correction. (C) The volcano plot showing DEGs of integrated CKD dataset. Upregulated gene were represented by red dots, while downloaded genes were represented by bule dots. (D) The heatmap presenting the top 20 upregulated and 20 downregulated DEGs of integrated CKD dataset. (E) The intersection of key genes in CKD related CAD and IRGGs through venn diagram. CKD, chronic kidney disease; IRGGs, inflammatory response-related genes; CAD, coronary artery disease; DEGs, differentially expressed genes; PCA, principal component analysis.

• 

  Fig. 6. Identification of potential diagnostic genes for CKD related CAD via two machine learning methods. (A, B) The LASSO logistic regression algorithm was applied to determine the minimum and lambda values of diagnostic genes in CAD. (C) MeanDecreaseGini analysis was performed on 14 genes using the RF algorithm, and nine biomarkers with scores greater than 2 were selected in CAD. (D, E) The LASSO logistic regression algorithm was applied to determine the minimum and lambda values of diagnostic genes in CKD. (F) MeanDecreaseGini analysis was performed on 14 genes using the RF algorithm, and ten biomarkers with scores greater than 2 were selected in CKD. (G) Venn diagram showing the overlap in results between two algorithms. CKD, chronic kidney disease; CAD, coronary artery disease; LASSO, least absolute shrinkage and selection operator; RF, random forest.

• 

  Fig. 7. Diagnostic effectiveness of three hub genes in CAD dataset. (A, B) ROC curve of diagnostic biomarkers including NUPR1, GCLM and PTGER1 in training group. (C, D) ROC curve of diagnostic biomarkers including NUPR1, GCLM and PTGER1 in validation group. CAD, coronary artery disease; ROC receiver operating characteris tic.

• 

  Fig. 8. Immune cell infiltration analysis in CAD. (A) Stacked histogram showing the immune cell kinds and ratios in CAD. (B) Violin plot displaying the expression level of 22 types of immune cells between CAD and control group. (C) Heat map illustrating the correlation among 22 kinds of immune cells in CAD. The size of the colored squares represents the strength of the correlation, with red indicating positive and blue indicating negative correlations. Darker hues signify stronger associations. (D) Correlation map demonstrating the correlation between immune cells with GCLM, NUPR1 and PTGER1 in CAD. CAD, coronary artery disease. *p < 0.05; **p < 0.01; ***p < 0.001; ns not significant.

• 

  Fig. 9. Validation of the expression patterns of three hub genes in CKD patients and assessment of their diagnostic value for CAD in this population. (A–C) Expression levels of three hub genes from hemodialysis patients with or without CAD. (D, E) ROC curve of diagnostic performance including GCLM and NUPR1 in hemodialysis patients. (F, H) Expression levels of three hub genes from patients with CKD stage 3–5 with or without CAD. (I, J) The ROC curve of diagnostic performance including GCLM and NUPR1 in patients with CKD stage 3–5. CKD, chronic kidney disease; CAD, coronary artery disease. *p < 0.05; **p < 0.01; ***p < 0.001; ns not significant.