Is bisulfite-seq data from different studies comparable with each other?

Question

[CTLong]

Hi all,

I am conducting a study which involves the comparison between more than 3 experimental conditions. In particular, I would like to study the methylation status of samples and to find differentially methylated regions between these conditions. Let's say if I only have bisulfite-seq data for one of these conditions but have found bisulfite-seq datasets in the public for the other two conditions, can I analyse them together and potentially find some "real" DMR regions between them?

Potentially, I am also looking to evaluate the genomic differences between these conditions. Is WGS data also comparable between different studies?

Thanks!

Answer 1

[colindaven]

Probably not. Batch effects are a huge issue so read up on the statistical literature.

You can try, but be very skeptical. Start with heatmaps, check the samples globally and across a range of relevant genes. Look at different normalization methods. There are batch effect correction tools, but these are somewhat simplistic and controversial in my experience.

Potential DMR regions should also be checked via another method.

Also - if you want to stop your advisor pushing you towards this - reviewers easily catch this kind of weakness, so it may not be publishable ....

[CTLong]

Hi Colin, thanks for the reply. What about integration of RNA-seq expression from different studies? I guess this is more widely accepted right (looking at the citations of Combat and Combat-seq)?

[Basti]

There is no general rule, what you can apply depends on the nature of each dataset you intend to integrate (tissue, condition, technique, sample size, ...). You need to ask precise biological questions and evaluate whether you have suitable datasets to answer those questions.

Answer 2

[NextGenSeek]

The efficiency of bisulfite treatment itself can differ significantly, so I would definitely not compare bisulfite-seq data from different datasets.