[BioC] DEseq for chip-seq data normalisation

[Giuseppe Gallone]

Hello Sheng

I have had biologically interesting results when selecting overlapping 
peaks between replicates using the IDR. I first call peaks with MACS2 or 
SPP, using a very loose q-value threshold, and then pass these to the IDR.

However, in my experience this works fine between pairs of aligned read 
files. If you have many samples, my experience reflects Lucia's: 
plotting an overlap function and then, based on the overlap behaviour, 
choosing a minimum number of carriers for each peak position (in my case 
2-3) works well to create a consensus peakset.

Regards
Giuseppe

On 11/06/13 09:40, pbczyd . wrote:
> Hi,
> I always think I should use IDR from ENCODE to get some <conservative>
> peaks from replicates, then move to Diffbind/DESeq.
>
> How do you think?
>
> PS: IDR: https://sites.google.com/site/anshulkundaje/projects/idr
>
> Regards,
> Sheng
>


-- 
Giuseppe Gallone, PhD
MRC career development fellow
MRC Functional Genomics Unit - DPAG
University of Oxford, UK