Differential Methylation

Introduction: Differential Methylation of Sample Groups

Differential methylation analysis was conducted on site and region level according to the sample groups specified in the analysis.

Comparisons

The following comparisons were made:

P-values

In the following anlyses, p-values on the site level were computed using the limma method. I.e. hierarchical linear models from the limma package were employed and fitted using an empirical Bayes approach on derived M-values.

Site Level

Differential methylation on the site level was computed based on a variety of metrics. Of particular interest for the following plots and analyses are the following quantities for each site: a) the difference in mean methylation levels of the two groups being compared, b) the quotient in mean methylation and c) a statistical test (t-test or limma depending on the settings) assessing whether the methylation values in the two groups originate from distinct distributions. Additionally each site was assigned a rank based on each of these three criteria. A combined rank is computed as the maximum (i.e. worst) rank among the three ranks. The smaller the combined rank for a site, the more evidence for differential methylation it exhibits. This section includes scatterplots of the site group means as well as volcano plots of each pairwise comparison colored according to the combined ranks or p-values of a given site.

The following rank cutfoffs have been automatically selected for the analysis of differentially methylated sites:

Rank Cutoff
ESC vs. derived (based on cmp_ESC_vs_differentiated) 4320614
normal vs. tumor (based on cmp_colon_tumor_vs_normal) 0
AD vs. control (based on cmp_brain_AD_vs_normal) 4105800
ectoderm vs. mesoderm (based on cmp_derGermLayer_meso_vs_ecto) 3783408
endoderm vs. mesoderm (based on cmp_derGermLayer_meso_vs_endo) 3882234
ectoderm vs. endoderm (based on cmp_derGermLayer_endo_vs_ecto) 3636816
comparison
differential methylation measure

Figure 1

Figure 1

Scatterplot for differential methylation (sites). If the selected criterion is not rankGradient: The transparency corresponds to point density. If the number of points exceeds 2e+06 then the number of points for density estimation is reduced to that number by random sampling.The1% of the points in the sparsest populated plot regions are drawn explicitly (up to a maximum of 10000 points).Additionally, the colored points represent differentially methylated sites (according to the selected criterion). If the selected criterion is rankGradient: median combined ranks accross hexagonal bins are shown as a gradient according to the color legend.

comparison
difference metric
significance metric

Figure 2

Figure 2

Volcano plot for differential methylation quantified by various metrics. Color scale according to combined ranking.

Differential Methylation Tables

A tabular overview of measures for differential methylation on the site level for the individual comparisons are provided in this section. Below, a brief explanation of the different columns can be found:

The tables for the individual comparisons can be found here:

Region Level

Differential methylation on the region level was computed based on a variety of metrics. Of particular interest for the following plots and analyses are the following quantities for each region: the mean difference in means across all sites in a region of the two groups being compared and the mean of quotients in mean methylation as well as a combined p-value calculated from all site p-values in the region [1]. Additionally each region was assigned a rank based on each of these three criteria. A combined rank is computed as the maximum (i.e. worst) value among the three ranks. The smaller the combined rank for a region, the more evidence for differential methylation it exhibits. Regions were defined based on the region types specified in the analysis. This section includes scatterplots of the region group means as well as volcano plots of each pairwise comparison colored according to the combined rank of a given region.

The following rank cutfoffs have been automatically selected for the analysis of differentially methylated regions:

tiling genes promoters cpgislands
ESC vs. derived (based on cmp_ESC_vs_differentiated) 4931 2317 349 145
normal vs. tumor (based on cmp_colon_tumor_vs_normal) 0 0 0 0
AD vs. control (based on cmp_brain_AD_vs_normal) 165 1395 265 218
ectoderm vs. mesoderm (based on cmp_derGermLayer_meso_vs_ecto) 669 1062 74 225
endoderm vs. mesoderm (based on cmp_derGermLayer_meso_vs_endo) 1774 1399 162 328
ectoderm vs. endoderm (based on cmp_derGermLayer_endo_vs_ecto) 439 1056 91 143
comparison
regions
differential methylation measure

Figure 3

Figure 3

Scatterplot for differential methylation (regions). If the selected criterion is not rankGradient: The transparency corresponds to point density. The 1% of the points in the sparsest populated plot regions are drawn explicitly. Additionally, the colored points represent differentially methylated regions (according to the selected criterion). If the selected criterion is rankGradient: median combined ranks accross hexagonal bins are shown as a gradient according to the color legend.

comparison
regions
difference metric
significance metric

Figure 4

Figure 4

Volcano plot for differential methylation quantified by various metrics. Color scale according to combined ranking.

Differential Methylation Tables

A tabular overview of measures for differential methylation on the region level for the individual comparisons are provided in this section.

The tables for the individual comparisons can be found here:

tiling genes promoters cpgislands
ESC vs. derived (based on cmp_ESC_vs_differentiated) csv csv csv csv
normal vs. tumor (based on cmp_colon_tumor_vs_normal) csv csv csv csv
AD vs. control (based on cmp_brain_AD_vs_normal) csv csv csv csv
ectoderm vs. mesoderm (based on cmp_derGermLayer_meso_vs_ecto) csv csv csv csv
endoderm vs. mesoderm (based on cmp_derGermLayer_meso_vs_endo) csv csv csv csv
ectoderm vs. endoderm (based on cmp_derGermLayer_endo_vs_ecto) csv csv csv csv

Enrichment Analysis

Enrichment Analysis was conducted. The wordclouds and tables below contains significant GO terms as determined by a hypergeometric test.

comparison
Hypermethylation/hypomethylation
ontology
regions
differential methylation measure

Figure 5

Figure 5

Wordclouds for GO enrichment terms.

comparison
Hypermethylation/hypomethylation
ontology
regions
differential methylation measure

GOMFID Pvalue OddsRatio ExpCount Count Size Term
GO:0050911 0 162.9451 0.1211 4 368 detection of chemical stimulus involved in sensory perception of smell
GO:0009593 0 134.7397 0.1454 4 442 detection of chemical stimulus
GO:0050906 0 132.2511 0.1481 4 450 detection of stimulus involved in sensory perception
GO:0007606 0 130.4425 0.15 4 456 sensory perception of chemical stimulus
GO:0007186 1e-04 50.5984 0.3675 4 1117 G-protein coupled receptor signaling pathway
GO:0050877 2e-04 48.0274 0.3856 4 1172 neurological system process
GO:0044707 0.0093 Inf 1.9639 5 5969 single-multicellular organism process

References

  1. Makambi, K. (2003) Weighted inverse chi-square method for correlated significance tests. Journal of Applied Statistics, 30(2), 225234