Flanking of modules

A preliminary conclusion so far is that DNA methylation may not be driving formation of coexpression modules because of the very small overlap between the genes of the coexpression and comethylation modules. Even modules that are highly correlated don't have many genes in common. Xia has proposed another idea: that DNA comethylation modules might be flanking coexpression modules and somehow defining the boundaries of the coexpression modules. In order to do a quick analysis, I focused on seashell and thistle modules from the comethylation network and the orchid module from the coexpression network. Majority of the genes of the orchid module (coexpression network) come from chromosome 19, which is rather an unusual characteristic (but might be common for cancer datasets).

Similarly, modules seashell and thistle are made from CpGs of chromosome 19:

In addition, all three modules are associated with survival, orchid is highly correlated with seashell and thistle has some overlap with orchid.
Orchid: 55 genes, Cox pvalue (calculated by Bin): 0.0418022, overlap with Thistle (comethylation, 67 CpGs; Wald's P value: 0.0108) is 13 genes, P value associated with overlap is 0. Rho is 0.7.

Orchid's overlap with Seashell (comethylation, 105 CpGs; Wald's P value is 0.0238) is 1, P value associated with overlap is 1. Rho is 0.4.

Next I binned gene and CpG counts along chromosome 19 to see relative clustering of the module members.

It seems that orchid module is followed by thistle and then seashell modules. I googled "ovarian cancer right arm chromosome 19" ad found the following references:

Amplification of 19q13.1-q13.2 sequences in ovarian cancer. G-band, FISH, and molecular studies, 1996. By FISH showed structural abnormalities, amplification and overexpression of AKT2 oncogene.

Analyses of other correlated modules from the coexpression and comethylation (mb) networks. Is flanking a rule?

Here are few other graphs for the correlating coexpression and comethylation modules along with their stats (module size, overlap, overlap P value, correlation coefficient, association with survival). All Rho correlation coefficients are |Rho|=>0.4

Violet (Affy): size - 63, Cox P value - 0.83, major chromosome - X

Gold2 (mb): size - 27, Wald's P value - 0.24, major chromosome - X.

Overlap: 1, overlap P value: 0.079

Violetred (Affy): size - 53, Cox P value - 0.04, major chromosome - 13

Yellow4 (Affy): size - 26, Cox P value - 0.54, major chromosome - 13

Peru (mb): size - 134, Wald's P value - 0.059, major chromosome - 13.

Overlap (violetred vs peru): 16, overlap P value: 0

Overlap (yellow4 vs peru): 3, overlap P value: 0.54

Black (Affy): size - 133, Cox P value - 0.19, major chromosome - 9

Chocolate (mb): size - 109, Wald's P value - 0.026, major chromosome - 9.

Overlap: 24, overlap P value: 0

Pink (Affy): size - 121, Cox P value - 0.18, major chromosome - 17

Firebrick (mb): size - 53, Wald's P value - 0.5, major chromosome - 17.

Overlap: 4, overlap P value: 0.002

Greenyellow (Affy): size - 109, Cox P value - 0.74, major chromosome - 16

Forestgreen (mb): size - 91 , Wald's P value - 0.47, major chromosome - 16.

Overlap: 12, overlap P value: 0

It seems to me that flanking of coexpression modules with comethylation modules correlated with them is not a rule. What I see is that CpG sited from cometh. module are sprinkled over regions of general clustering of the coexpressed genes.

Browser not supported