How to normalize 450K data for cell type composition

Question

[Nicolas Rosewick]

Hi,

I've a bunch of 450K (Infinium HumanMethylation450 BeadChip) methylation data from leukemic cells. I read that cell type composition (CD8, CD4, NK, ...) could biais the analysis. I was able to compute the cell composition using the estimateCellCounts function from minfi package. But now how can I normalize my beta (and M) values using these informations ?

Thanks you

Answer 1

[andrew.j.skelton73]

See this post - Kasper Hansen replied to a similar question I wrote about this (admittedly it was about a year ago), and stated that although the estimateCellCounts function can identify biases, it isn't clear what the best way to adjust for this is.

Answer 2

[igor]

This is very delayed, but it seems there is another package celltypes450 that can adjust beta values for cell-type composition: https://github.com/brentp/celltypes450

It was originally created before estimateCellCounts was available. Perhaps there is a good reason why estimateCellCounts doesn't offer such functionality (see andrew.j.skelton73's answer).