The hypothesis: cancer related changes in DNA methylation affect key drivers (or hub genes) of coexpression modules. How to test: find highly correlated pairs of CpG loci and the first principal components of the coexpression networks.

Affymetrix vs mb:

Based on the preliminary analyses of the coexpression network (PC1 variability) I removed some of the modules from further analyses: brown3, chocolate, cyan2, gold, gold3, gold4, green2, hotpink, lightgreen, mediumblue, plum, seashell, snw, yellow4, blue, brown2, gold (variation of these modules was higher than 0.2 on both sides. This was my arbitrary cutoff. Note: may want to repeat the analyses with all modules.) Furthermore, I removed the modules with the low percent variance explained (less than 30%): blue, brown, firebrick, plum, turquoise and yellow. Further analyses were performed with the remaining 42 modules out of 65. Reduced the size of the DNA methylation dataset to 378 and calculated all possible correlations:

rho	0.8	0.7	0.6	0.5	0.4	0.3
Targets	0	10	229	996	2545	5765

A few scatter plots to show the correlations that passed |rho| >=0.6 threshold:

Note: need to do permutations to calculate FDR.

Affymetrix vs mbc:

rho	0.8	0.7	0.6	0.5	0.4	0.3
Targets	0	7	200	874	2221	4573

A few scatter plots to show the correlations that passed |rho| >=0.6 threshold:

With the threshold of rho = 0.6 the number of modules correlated with CpGs in mb and mbc sets is 170.

Agilent vs mb:

Based on theprevious analyses I removed the following modules from the coexpression network: turquoise (high variability), yellow, sienna, red, pink, orchid, magenta, green, brown, blue, bisque, brown3 (these all have less than 30% variance explained).

rho	0.8	0.7	0.6	0.5	0.4	0.3
Targets	0	33	311	1129	2824	7782

A few scatter plots to show the correlations that passed |rho| >=0.6 threshold:

Agilent vs mbc:

rho	0.8	0.7	0.6	0.5	0.4	0.3
Targets	0	30	406	1256	2758	6255

A few scatter plots to show the correlations that passed |rho| >=0.6 threshold:

Next step in the analysis is to see if CpGs that are correlated with certain modules map to nearby genes and whether these genes are also within the correlated module. Is they are within a module, are they key drivers or hub genes (hub genes were defined based on the top 10% highest k.in values within that module)? Dow DNA methylation status also correlates with expression of that gene?

properties	mb, 229	mbc, 200
Mapped to nearby Affymetrix gene (+/- 2kb of TSS)	122	110
The gene is within the same module	25	25
The gene is in the same module, key driver (Bayesian network)	0	0
The gene is in the same module, hub gene (top 10% based on k.in)	6	6
The gene is in the same module, not a hub, CpG correlates with its expression	8	?
The gene is in the same module, hub, CpG correlates with its expression	5	?

? - analysis hasn't been performed

Genes that are in the same module that is correlated with the nearby CpG, it is a hub, and its expression correlates with CpG status (mb):

KEAP1	9.4085	lightcyan	cg03754063	kelch-like ECH-associated protein 1
KIAA0090	7.52433	tan	cg18224761	KIAA0090
PPP2R2A	5.58591	darkcyan	cg10207787	protein phosphatase 2 (formerly 2A), regulatory subunit B, alpha isoform
RDBP	9.65791	midnightblue	cg04710641	RD RNA binding protein
ZNF45	5.29257	blue2	cg00410831	zinc finger protein 45

Browser not supported