[BioC] DESeq on CCAT identified chipseq peaks

Rory Stark Rory.Stark at cruk.cam.ac.uk
Fri May 16 16:16:52 CEST 2014


Hello Aditi-

What you want is to use a "matched" design. There is a good explanation of
this design (in a differential expression context) in the edgeR vignette.
Basically, the matched tumour-normal pairs are going to have certain
similarities to each other as they each come from the same patient. A
matched design will model this to detect consistent differences in
enrichment between tumor and normal that is independent of individual
patients.

You can analyze a matched design by setting up the contrast with
block=DBA_REPLICATE:

> h3k4me3_counts = dba.contrast(h3k4me3_counts, categories=DBA_CONDITION,
>block=DBA_REPLICATE)
> h3k4me3_counts = dba.analyze(h3k4me3_counts, method=DBA_DESEQ2)

You'll see that two analyses are run (unmatched and matched):

> h3k4me3_counts

Is is useful to look at the MA plot:

> dba.plotMA(h3k4me3_counts, method=DBA_DESEQ2_BLOCK)

You can get the list of all the sites with statistics relating to how
confidently they can be identified as being differentially enriched:
 
> matchedReport = dba.report(h3k4me3_counts, method=DBA_DESEQ2_BLOCK, th=1)

Cheers-
Rory


On 16/05/2014 06:54, "QAMRA Aditi (GIS)" <qamraa99 at gis.a-star.edu.sg>
wrote:

>Hi Dr. Rory,
>
>I understand now. Thank you !
>
>A last question (hopefully) - Can you explain a little more on how the
>use of a blocking factor works in the case of matched normal tumor pairs
>? Does it mean that using the DBA_REPLICATE condition as a blocking
>factor in such a case adjusts (?) and removes any sort of batch effects
>between replicates ?
>
>Thanks !
>Aditi
>________________________________________



More information about the Bioconductor mailing list