Fig. 3

DMRs overlapping different genomic features had distinct methylation profiles. a: Upper heatmap and dendrograms show unsupervised hierarchical clustering of 709 DMRs (rows) based on the combinations of genomic features they overlapped (columns). Green colored cells indicate DMR-feature overlap. Bottom heatmap shows Pearson correlations between the first 5 principal components (PCs, rows) of the DMR correlation matrix and genomic features (columns). The name of each PC is followed in parentheses by the amount of variance it explained within the DMR correlations. Red and blue denote strong positive and negative correlations, respectively. Correlation coefficients are shown in cells where the p-value was significant after Bonferroni-correction for 55 tests (5 PCs × 11 feature types). b-e: Boxplots show comparisons of DMR baseline methylation levels (b-c, y-axis) and the absolute value of ND - D methylation differences (d-e, y-axis) within different groupings of the DMRs. Baseline methylation is defined as the mean of all smoothed methylation values within the DMR across all fish and denoted as all.mean in Additional file 1: Figure S2 and Additional file 2: Table S1, while the absolute value of ND - D methylation differences is denoted as meanDiff.abs in Additional file 1: Figure S2 and Additional file 2: Table S1. First, DMRs are grouped by whether they were contained within one part of a gene body or regulatory region, including introns, exons, within 5 kb upstream of transcription start site (5′ type-1), or within 5 kb downstream of 3′ end (3′ type-1) (b, d), and second, DMRs were grouped by whether they contained transposable elements (TEs) or conserved non-coding elements (CNEs) (c, e). The number of DMRs in each grouping is reported in parentheses below group names in each boxplot. Horizontal grey lines denote the median value across all DMRs tested. Kruskal-Wallis omnibus p-values are reported above each plot and asterisks indicate significance levels of pairwise comparisons after Dunn’s test with Bonferroni-correction; p < 0.05*, 0.01**, 0.001***