DESeq on BS-Seq data

[brentp]

Hi, would it be reasonable to use DESeq on Bisulfite data?

I was thinking of using counts of the number of unconverted Cytosines (for different treatments) covering each gene as the input.
Any reason I should *not* do that?
Any thing I should be aware of in doing so?

Thanks,
Brent

[brentp]

Or perhaps the counts would be better as number of Cytosines with > X% methylation within a gene?

[Wolfgang Huber]

Dear Brent

probably the answer is yes, but "use DESeq on Bisulfite data" is vague, and it will be good to first clarify: what is the biological question that you want to answer with this? How do you translate it into the hypothesis test(s) of DESeq?

Whatever you do, can you first apply it to a comparison between biological replicates ("as if" they were the two conditions that you want to compare) and make sure that the type-I error (false positive rate or false discovery) is as advertised?

Best wishes
Wolfgang