[BioC] edgeR/DESeq for ChIP-seq analysis

Gordon K Smyth smyth at wehi.EDU.AU
Sat Nov 10 07:22:57 CET 2012


Dear Davide,

My understanding is that the edgeR glm functionality was used to evaluate 
statistical significance of differential binding in Chandra et al (2012), 
although edgeR is not mentioned in the paper.  In other words, DiffBind 
organized the counts and edgeR did the parameter estimation and 
statistical tests.

Best wishes
Gordon

> Date: Thu, 8 Nov 2012 12:15:35 +0000
> From: Rory Stark <Rory.Stark at cancer.org.uk>
> To: "bioconductor at r-project.org" <bioconductor at r-project.org>
> Subject: Re: [BioC] edgeR/DESeq for ChIP-seq analysis
>
>
> As Mark said, DiffBind provides a straightforward workflow for this type 
> of ChIP-seq analysis, with all the statistical heavy lifting done by 
> edgeR and/or DESeq.
>
> Regarding histone marks, we have had success using DiffBind to analyze 
> wider regions of enrichment. For example, in Chandra et al (Mol Cell 
> 2012 47:2), we found biologically meaningful differences in five histone 
> marks over enrichment regions as wide as 500Kb (cf Figure 3d). The 
> tricky part with this type of enrichment is in peak calling, as the most 
> popular peak callers (esp. ones that rely on strand information) assume 
> that the enriched area (peak) is shorter than the sequenced fragment 
> length. There are a number of peak callers that are designed to find 
> wider areas of enrichment. We have used these peak callers, or avoided 
> peak calling all together using general windowing schemes or genomic 
> annotations (eg windows oriented around transcription start sites to 
> capture binding profiles in promoter regions).
>
> Cheers-
> Rory
>
> ----------------------------------------------------------------------------
> Dr. Rory Stark
>
> Principal Bioinformatics  Analyst
>
> Cancer Research UK
> Cambridge Research Institute
> Robinson Way
> Cambridge CB2 0RE
> United Kingdom
> +44 1223 404 311
>
> rory.stark at cancer.org.uk
> ----------------------------------------------------------------------------
>
>> Hi,
>
>
>> I want to continue this discussion.
>
>> I saw in some papers, people used edgeR and DESeq to analysis
>> differentially bound between different sample groups following ChIP-seq.
>
>> But most of them are studying transcription factors.
>
>> Is it the case for histone modifications ChIP-seq (H3K4me1, H3K4me2 or> H3k9me3)?
>
>> Regards,
>> Sheng
>
>
> On Thu, Nov 8, 2012 at 8:59 AM, Cittaro Davide <cittaro.davide at hsr.it>wrote:
>
>> Dear Mark
>>
>> On Nov 8, 2012, at 8:53 AM, Mark Robinson <mark.robinson at imls.uzh.ch>
>> wrote:
>>
>>> Dear Davide,
>>>
>>> Indeed, edgeR and DESeq can be (and have been) used in this mode.  We
>> published something recently on this:
>>>
>>> http://www.ncbi.nlm.nih.gov/pubmed/22879430
>>> http://imlspenticton.uzh.ch/robinson_lab/ABCD-DNA/ABCD-DNA.pdf
>>>
>>
>> I've missed that :-( Thanks for the paper
>>
>>> You can apply that approach regardless of copy number being a factor ...
>> basically, we counted tiled bins of the genome, but yes, you could focus in
>> on regions of interest.  The function abcdDNA() is really just a wrapper
>> for the edgeR GLM.  As usual, "normalization" can be delicate, depending on
>> the type of data.
>>>
>>> Also note that the DiffBind package already does something similar, but
>> has a lot more machinery to collect and organize the sets of enriched
>> regions.
>>
>> I wonder why I've never used DiffBind before :-)
>>
>>>
>>> Hope that helps.
>>
>> It does, thanks!
>>
>> d
>

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}



More information about the Bioconductor mailing list