Skip to main content
. 2020 Oct 19;11:5270. doi: 10.1038/s41467-020-18965-w

Fig. 5. Tissue-derived 5hmC features are able to classify PDAC cfDNA.

Fig. 5

a Average 5hmC FPKM values over top 50 hyper-hydroxymethylated (hyper-hme) genes with the highest 5hmC occupancy compared with the top 50 hypo-hydroxymethylated (hypo-hme) genes with the lowest 5hmC occupancy in PDAC tumor tissues (n = 17). Boxplot center lines represent median, bounds of box represent 25th and 75th percentiles and whiskers are Tukey whiskers. Each dot represents an individual PDAC tumor tissue. b PCA plots showing non-cancer (blue) and PDAC (red) cfDNA cohorts using top 50 hyper-hydroxymethylated (upper panel) and top 50 hypo-hydroxymethylated (bottom panel) genes as features. c ROC curves for predictive model using features discovered in cfDNA (teal), top 50 genes with highest 5hmC representation in tissue (solid coral), or 50 genes with lowest 5hmC representation in tissue (dotted coral). d Heatmap showing FPKM values of all the gene features (n = 37) used in cfDNA prediction model (represented in each row) in non-cancer cfDNA, PDAC cfDNA, and PDAC tissue 5hmC profiles (represented in each column).