I have an Infinium methylationEPIC dataset that I use R package minfi to analyse.

I normalized (noob, swan, func, quantile)(sex info used for func and quantile) the data and convert to beta values. XY probes are now removed. Beta values for autosomal probes are then converted to M values. MDS plots are created using beta and M values.

I find that beta values and M values cluster differently. Beta values do not show clustering by sex but M values show clustering by sex. I find it hard to figure out how going from beta to m-values can change the clustering profile. Is this normal or is something wrong somewhere? There is also this weird stripy clustering going on with noob, swan and func beta value mds.

Fig: MDS on beta values (left) and M values (right). Each plot shows different normalisations (raw, noob, swan, func or quantile).

Other info related to the various normalisations are in this question.

Thanks.