Methylation Analysis EPIC array

Question

[Will Yam]

Hello,

I have run a dm analysis using the workflow described here:

https://www.bioconductor.org/packages/devel/workflows/vignettes/methylationArrayAnalysis/inst/doc/methylationArrayAnalysis.html

I have a few questions and would be grateful if i could get any feed back or input.

I'm working with AML project with samples of patient at different stages (healthy, first diagnosis, first or second relapse, etc..) i have run a pca and plotted using plotMDS but i see not cluster except the healthy one. I feel uncertain about the lack of clustering. any advice?

then I've run the DM using limma:

modelMatrix <- model.matrix(~0+sampleGroupFactor+sampleSourceFactor, data=targets)
mlmFit <- lmFit(mvalues.filtered, modelMatrix)
DMPs <- topTable(mlmFit.Contrasts.Ebayes, num=Inf, coef=1, genelist=annotation.Filtered)

now once again, when i plot the top 4 most dm'ed cpg sites:

plotCpg(beta.Filtered, cpg=cpg, pheno=targets$Sample_Group, ylab = "Beta values")

except the top1, i see very little difference, the beta values are kind of evenly distributed between 0 and 1 Once again, what give? what do I do?

So then i go on and identify the DMR using dmrcate:

cpgAnnotation <- cpg.annotate(object = mvalues.filtered, 
                              datatype = "array", 
                              what = "M",
                              analysis.type = "differential", 
                              design = modelMatrix,
                              contrasts = TRUE, 
                              cont.matrix = contrastMatrix,
                              coef = coef, 
                              arraytype = "EPIC")
DMRs <- dmrcate(cpgAnnotation, lambda=1000, C=2)
results.ranges <- extractRanges(DMRs)

and plot them using DMR.plot once again, very little difference in methylation appears.

Now I have a few questions regarding methylation analysis:

Q1 what can I do reg lack of clusters in the PCA, and the lack of difference in methylation? unclear how to move forward if I don't see big differences in methylation between different groups. How can I know if I made a mistake or not? how can I check the coherence of my results? I was told that it was not surprising not to see any clustering except for the healthy samples, but at the moment i have no way to see if there is a mistake or not.

Q2 regarding grouping, i have made several DMA comparing different relapses groups, broken down between adults and children, paired or unpaired. Would that make sense to group along other characteristics than the disease stage? sex? (actually I filter out sex chromosomes so i guess no?) age bins? could you recommend another kind of grouping?

Q3 At this point my idea was to check if the average beta of the cg sites of the genes overlapping my DMR was higher on non expressed genes (using the RNA counts). Does that make sense to you?

Q4 Any idea what else I could do?

Any advice very much appreciated!