Classifiers for Predicting Coronary Artery Disease Based on Gene Expression Profiles in Peripheral Blood Mononuclear Cells

Figures & data

Figure 1 Flow chart of the present study. PCA, principal component analysis.

Abbreviations: LASSO, Least absolute shrinkage and selection operator; GSVA, Gene set variation analysis; ROC, Receiver operating characteristic curve.

Figure 2 Differential gene expression analysis and functional enrichment analysis. (A) Volcano map. Red indicates genes that were up-regulated in coronary artery disease (CAD) patients compared to healthy samples, green indicates genes that were down-regulated, and gray indicates genes with similar expression between both groups. (B) Heat map of differentially expressed genes (DEGs). The depth of color reflects the level of differential expression. (C) The biological processes where DEGs may be involved according to enrichment analysis. (D) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways in which DEGs may be involved. (E) Biological processes where DEGs may be involved according to Clue GO analysis. (F) KEGG pathways enriched in CAD samples.

Table 1 The 20 Feature Genes Selected by Least Absolute Shrinkage and Selection Operator (LASSO) Method

Feature Gene	Coefficient
MTRNR2L9	0.001973493
IL1A	0.045115135
HOPX	0.41611678
SH2D2A	0.094835346
XLOC_001392	−0.001822156
PAX8	−0.59780644
XLOC_010247	−0.516403532
LGR6	0.097081831
DLG3	0.099024616
PIK3R1	0.006243573
XLOC_009416	0.482317217
DDX23	0.044157509
DUX4	−0.94458075
PPT2	0.203846629
THOC4	0.019290091
NEURL1B	−1.053835989
KCNK9	−1.060114652
KRTAP6.1	−0.56392575
SIK1	0.225232958
GNG2	1.565124099

Figure 3 Least absolute shrinkage and selection operator (LASSO) model and principal component analysis (PCA). (A) Ten-times cross-validation of parameter selection in the LASSO model. (B) PCA after LASSO variable reduction. (C) Receiver operating characteristic (ROC) curve of the training set. (D) ROC of the test set.

Abbreviations: AUC, area under the ROC curve; var., variance.

Figure 4 Support vector machine (SVM) model. (A) Receiver operating characteristic (ROC) curve of the training set. (B) ROC of the test set.

Abbreviation: AUC, area under the ROC curve.

Figure 5 Evolution of the Z-score during Boruta operation. (A) The Boruta function was used to further select features in the 20 differentially expressed genes. (B) Receiver operating characteristic (ROC) curve of the training set. (C) ROC of the test set.

Abbreviation: AUC, area under the ROC curve.

Figure 6 Gene set variation analysis (GSVA) score model. (A) Receiver operating characteristic (ROC) curve of the training set. (B) ROC of the test set.

Abbreviation: AUC, area under the ROC curve.

Table 2 Performance of the Four Classifiers for Diagnosing CAD

Classifier	Dataset	Se	S p	PPV	NPV	Accuracy	AUC
LASSO	Training set	1.00	1.00	1.00	1.00	1.00	1.00
	Test set	1.00	1.00	0.960	1.00	0.980	0.997
SVM	Training set	1.00	1.00	1.00	1.00	1.00	0.997
	Test set	0.960	0.930	1.00	0.930	0.980	0.997
Random forest	Training set	1.00	1.00	1.00	1.00	1.00	1.00
	Test set	0.980	0.970	1.00	0.97	0.990	1.00
GSVA	Training set	–	–	–	–	–	0.971
	Test set	–	–	–	–	–	0.989

Abbreviations: AUC, area under the receiver operating curve; NPV, negative predictive value; PPV, positive predictive value; Se, sensitivity; Sp, specificity.

Figure 7 Gene set expression analysis (GSEA) of optimal genes and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enriched in the 10 differentially expressed genes.

Figure 8 Gene set expression analysis (GSEA) of optimal genes and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enriched in the 10 differentially expressed genes.