Hello, I’m new to this so sorry if any of this is obvious but I would really appreciate any help. I have 6 cancer cell lines, two of each of the following subtypes: mesenchymal, proneural, classical.

I want to find differentially methylated positions/regions, and have tried using limma to compare each subtypes in a contrast matrix but this generates no significant results. This is not totally unexpected as, from inspecting MDS plots/PCA, there probably isn’t enough power to detect changes due to the low sample size. However, I would like to compare the 2 cell lines that are most likely to have significant results from looking at the MDS plot. Can anyone suggest a single replicate analysis pipeline that would allow me to do this? I cannot do it with limma as it cannot do the statistical analysis so just comes up with an error.

However, prior to setting up my contrast matrix, my design matrix (without intercept) generates significant probes for each of the subtypes. I used this design matrix to calculate DMPs and have found a large number of significant probes. But, as this is not the way it’s done in the pipeline, I have no idea if these results are valid? Does anyone know please? From this I can’t calculate DMRs as DMRcate requires the use on an intercept and that alters the whole matrix, hence why I don’t know if using the matrix without intercept is okay for DMPs, but I am guessing not. I also tried using ChAMP but that generated no significant DMPs.

Hopefully this makes sense and someone can help me out, thanks!

fit <- lmFit(mVals, design)
summary(decideTests(fit))

        Mesenchymal Proneural Classical
Down        212395    208081    208239
NotSig      278049    298912    284535
Up          321598    305049    319268

contMatrix <- makeContrasts(Mesenchymal-Proneural,
                        Mesenchymal-Classical,
                        Proneural-Classical,
                        levels=design)
contMatrix
Levels        Mesenchymal - Proneural Mesenchymal - Classical Proneural - Classical
Mesenchymal                       1                       1                     0
Proneural                        -1                       0                     1
Classical                         0                      -1                    -1

fit2 <- contrasts.fit(fit, contMatrix)
fit2 <- eBayes(fit2)
summary(decideTests(fit2))

                 Mesenchymal - Proneural Mesenchymal - Classical Proneural - Classical
Down                         0                       0                     0
NotSig                  812042                  812042                812042
Up                           0                       0                     0



Source link