[BioC] DiffBind - Info

Rory Stark Rory.Stark at cruk.cam.ac.uk
Thu Nov 14 16:38:57 CET 2013

Hello Goutham-

You should not do any manipulation of the reads (including normalization).
When you call dba.analyze the data will be normalized by the analysis
package(s) you choose (edgeR, DESeq, DESeq2) using the read counts
computed by dba.count.

There is an explanation of how the normalizaiton works for each analysis
method near the end of the User Guide/Vignette. You have some control over
the normalization in the call to dba.analyze, using the bSubControl and
bFullLibrarySize options.

You can retrieve normalized data using dba.report after running an
analysis, with bCounts=T (and bNormalized kept at its default value of
TRUE). You'll need to set the FDR threshold to 1 (th=1) if you want to
retrieve values for all sites.

You can also get normalized values for all the data using dba.peakset with
bRetrieve=TRUE after counting, but only for some normalization methods,
with the default being the TMM method from edgeR; see the "score"
parameter in dba.count.


On 08/11/2013 15:52, "Goutham atla" <goutham.atla at gmail.com> wrote:

>Dear Dr. Stark,
>I am using DiffBind for Chip-Seq data analysis. I have biological
>replicates between two conditions.
>I would like to know, before giving the reads and peak files of
>different samples ( Replicates and Conditions ), should I perform any
>normalization of the data ( MANorm etc ) ? Or The DiffBind will take
>care of Normalization also ? If so, could we get the normalized values
>that are generated during the analysis ?
>Thanks and Regards,
>Goutham Atla

More information about the Bioconductor mailing list