MeDIP-seq and Seqmonk

[Grenouille]

Hi everyone,

I am looking for some advice regarding the analysis of a MeDIP-seq dataset. I generated paired-end sequences of two biological replicates of the same genotype, while just one Input sequence is available.
I was planning to use Seqmonk to have a first glance but despite it's very user-friendly I am having a hard time understanding which tools I should use; in particular, I do not see at which step the Input dataset is taken into account to correct the IP datasets. The only option that makes the distinction between IP and Input seems to be the use of MACS to define probes but I was wondering if the MACS algorithm is good at finding what I expect to be rather wide peaks.
I hope my queston is not too confused.
Thanks in advance.

[turnersd]

Haven't used SeqMonk. Take a look at MEDIPS. What are others using for MeDIP-Seq analysis with biological replicates?

[Grenouille]

Thank you for the reply. I am already using MEDIPS but just to compare different samples, apparently it is not appropriate for the identification of sample-wise local enrichment. I was told that to look for methylation-rich regions it is better to use a regular peak finder tool.

[turnersd]

I've used MACS and ZINBA on MEDIPS data, but this data doesn't really look "peaky" the way other xIP experiments might, and not sure how methods like these that were mainly designed for ChIP-seq are performing on MeDIP.

[Grenouille]

That's exactly my point. I would say that MeDIP peaks shouldn't look too different from ChIP peaks for histone modifications (such as H3K27me3) but as I am a beginner I wanted to ask for the opinion of more experienced people!

[paolo.kunder]

Hi,
I am also using MEDIPS for sample-wise enrichment, from the manual:

In principle, it is possible to calculate genome wide coverage and methylation
proles for only one MEDIPS SET or for only one group of MEDIPS SETs
using the function
MEDIPS.meth

But in ther meantime I used also MACS to calls for peaks refining the bandwitch and results are consistent. Then I called peaks with some other programs and extracted a consensus peakste with R package DiffBind

Anyway, I am really interesting in sharing knowledge with you guys..

[Grenouille]

Hi Paolo,

thanks for replying.
Did you include Input samples in your analysis? I know that the usefulness of such data is controversial but in my case I think I am really obliged to take Inputs into account; I am working with several different genotypes, while only one reference genome is available. That means that, due to structural variants, for some samples I could have some regions that show no coverage because the sequence is absent in that particular genotype, and here is where the Input becomes essential.
Are you working with similar data?

[tanaybhatt]

Can you direct me to a published paper where DMR analysis has been done without taking the input in consideration (or without input subtraction/normalization)?