Cardiovascular disease biomarkers derived from circulating cell-free DNA methylation

Figures

• 

  Figure 1.. (A) Total amount of extracted ccfDNA normalized to plasma volume. (B) Misclassification of ACS type by classical biomarkers and ccfDNA levels. (C) Correlation analysis of the classic cardiac biomarkers and ccfDNA concentration. LVEF, Left ventricular ejection fraction. CRP, C reactive protein. Troponin T, troponin T high-sensitive. CK, creatinine kinase. CK max, maximum value of CK measured. CK-MB, creatinine kinase isoform MB. CK-MB max, maximum value of CK-MB measured.

• 

  Figure 2.. Predicted cell type/tissue proportions per ACS group. Using a comprehensive methylation atlas and non-negative least squares method, the putative origin of ccfDNA could be deconvoluted. (Only cell/tissue types of interest are shown).

• 

  Figure 3.. (A) DMRs shared across the three ACS groups (compared to healthy control). (B) CpG regions, genomic and regulatory annotations for DMRs.

• 

  Figure 4.. (A) GREAT gene set enrichment for DMRs. Enriched canonical pathways from MSigDB CP (curated gene sets), with adjusted P-values <0.05 in at least one group are shown. The colors represent the adjusted P-values (–log10 based). (B) DisGeNET enrichment analysis showing cardiovascular/inflammatory diseases with adjusted P-values <0.05 in at least one group. The colors represent the adjusted P-values (–log10 based). The stars (*) indicate P-value >0.05.

• 

  Figure 5.. (A) Percentage of the discovery DMRs showing significant association with biomarkers and disease group on linear model adjusted by ccfDNA concentration on plasma. (B) PCA on methylation levels of 254 DMRs found significantly associated (P < 0.05) with ACS types encoded as disease severity using linear models. (C) Correlation plot with the top five DMRs most significantly associated with disease groups.

• 

  Figure 6.. DMR validation in validation cohort. (A) DMRs from DM analysis compared to healthy controls (n = 1589). Using a 25% methylation difference cutoff 37.3% (115 out of 308 DMRs) of NSTEMI, 30.0% (171 out of 570) of STEMI and 32.5% (82 out of 252 DMRs) of UA DMRs from the discovery cohort could be validated in the validation cohort. Linear Pearson correlation coefficients are stated and dotted lines indicate absolute 25% cut offs. (B) PCA on DMRs found significantly associated (P < 0.05) with disease groups on linear models in discovery cohort (n = 193). Clustering of samples according to condition without regard to the sequencing method.