Coronary Artery Disease with Elevated Levels of HDL Cholesterol Is Associated with Distinct Lipid Signatures

Figure 1. Partial least−squares−discriminant analysis (PLS−DA) scores plot of the High HDL CAD (+) and High HDL CAD (−) groups. (A) Two−dimensional scores plot with 95% confidence regions. (B) Sparse three−dimensional scores plots.

Figure 2. (A) Volcano plot of the statistically significant lipids, derived from the comparison between CAD (+) and CAD (−) groups, generated by p value from Wilcoxon test and fold change (FC). Red plots indicate lipids are up−regulated in High HDL CAD (+) group and blue means down−regulated. Gray plots represent insignificantly altered lipids. (B) Heatmap visualization of significantly altered lipid species between the CAD (+) and CAD (−) groups. (C) Pathway analysis of differential lipids between the CAD (+) and CAD (−) groups. GPI= glycosylphosphatidylinositol; S1P = sphingosine−1−phosphate; PE = phosphatidylethanolamine; PS = phosphatidylserine; GlcCer = glucosylceramide; PG = phosphatidylglycerol; GM3 = monosialo−dihexosyl ganglioside.

Figure 3. Differential lipid correlation network using the Spearman correlation coefficients. PE = phosphatidylethanolamine; PS = phosphatidylserine; PG = phosphatidylglycerol; S1P = sphingosine−1−phosphate; GlcCer = glucosylceramide; GM3 = monosialo−dihexosyl ganglioside.

Figure 4. Selection and evaluation of lipid biomarker candidates of CAD with elevated HDL−C levels. (A) Screening potential lipid biomarkers using the AUC (x−axis) and value of specificity + sensitivity (y−axis) calculated by the ROC analysis. Three colored circles represent that AUC of differential lipids are greater than 0.8, and the sum of specificity and sensitivity are greater than 1.6 meanwhile. The white circles represent lipids with the sum of specificity and sensitivity less than 1.6. (B−D) ROC curves and histograms of GM3 (d18:1/22:0), GM3 (d18:0/22:0), and PS (38:4). (E) The ROC curve of the biomarker model generated by combining the three lipid biomarkers based on linear SVM. (F) The mean predicted class probabilities for individual samples across the one hundred cross−validations of the diagnostic model. GM3 = monosialo−dihexosyl ganglioside; PS = phosphatidylserine; AUC = area under the curve; Sens. = sensitivity; Spec. = specificity.