DNA methylation preprocessing

Question

[SYOSY]

Hi,

I studying about DNA methylation analysis.

In R, minfi package provide many normalization methods. (for example, SWAN, Funnorm etc..)

but i don't know what is difference of these..

[jared.andrews07]

Have you read the documentation, vignette, and associated papers?

[soyeon]

Yes..

but i don't understand it well..

Answer 1

[Marc]

I happen to be writing a paper that compares preprocessing in minfi, sesame, and methylsuite so I can offer AN answer. However, know that different researches prefer different ways to process data and there is no one universally accepted "right" way.

Minfi has at least 5 ways to clean up the raw probe intensities. Of these, methylprep (of which I am the maintainer) and sesame (another R package we benchmark against) opt for pOOBAH, NOOB, and a nonlinear dye bias correction based on quantile normalization. Minfi doesn't offer 2 of these but does have a preprocessNOOB step. I would use that over SWAN and its quantile normalization and the other options.

The end result is that you get cleaned data that is more comparable across batches and between studies. Differences in background fluorescence are minimized and the beta values you calculate are shifted closer to 0 and 1, better matching what we'd expect.

[Dunois]

Hi, sorry for replying onto an old post. But this paper you mention, has it been published? Or is a preprint available somewhere?

[sup0903]

Hi, has the paper you mentioned been published? I would like to know the comparison results.

Answer 2

[cao510927]

I happend to see one paper doing the similar thing, share it with you guys, "A systematic evaluation of normalization methods and probe replicability using infinium EPIC methylation data". Hope it will help.