[BioC] DiffBind Question

Rory Stark Rory.Stark at cruk.cam.ac.uk
Thu Oct 17 18:49:15 CEST 2013

Hi Stephen-

1. dba.analyze itself doesn't use any thresholds, it just computes the
p-values and FDRs for every binding site. The thresholds are used in the
reporting and plotting functions: dba.report, dba.plotMA, dba.plotHeatmap,
dba.plotPCA, and dba.plotBox. The generic plot function corresponds to
dba.plotHeatmap and you can specify thresholds. eg:

> DBSites = dba.report(DBA,contrast=1, th=.05, bUsePval=TRUE)
> dba.plotMA(DBA, th=.05, bUsePvals=TRUE)

NOTE: I've just added the ability to set the default threshold for the
object instead of the global FDR < 0.1 currently implemented. This feature
will appear in the development version (as 1.9.2) soon. These values will
initialized as:

> DBA$config$th = 0.1
> DBA$config$pUsePval = FALSE

and can subsequently be changed at will.

2. You can export the entire affinity binding matrix (based on the
consensus set after running dba.count) using dba.peakset. If you set the
"bRetrieve" parameter to TRUE, this will return the entire matrix (the
"DataType" parameter controls the form of how this is returned; default
type is GRanges). If you set the "writeFile" parameter to a filename, the
data will be exported to a tab-delimited file of that name. eg:

> affinities = dba.peakset(DBA,writeFile="BindingAffintyMatrix.bed")


On 17/10/2013 16:55, "Stephen Mack" <stephen.mack at mail.utoronto.ca> wrote:

>Hi Rory,
>I was hoping you could help with with two DiffBind questions:
>1) Is it possible to adjust the p-value / FDR threshold for the
>dba.analyze function ?
>2) Is it possible to export the consensus / affinity binding matrices ?
>Stephen Mack
>Grad Student
>The Hospital for Sick Children
>Brain Tumour Research Centre
>Toronto, ON

More information about the Bioconductor mailing list