How to extract binding affinity matrix in Diffbind

Question

[1614999088]

Hi everyone, I'm performing ChIP-Seq analysis using Diffbind3.0, after doing:

>     dbObj <- dba(sampleSheet="peaksfile_merged.txt")
>     dbObj <- dba.count(dbObj, bUseSummarizeOverlaps=FALSE, summits=FALSE)
>     dbObj <- dba.normalize(dbObj, normalize=DBA_NORM_LIB)

then use plot(dbObj) for clustering among samples，and generate a heatmap of clustering among samples.

However, we want to make some adjustments to the generated heatmap. What are the parameters in the plot() function, or can we extract the original data used to draw the heatmap?

help will be highly appriciated!

Answer 1

[Rory Stark]

plot() calls dba.plotHeatmap(); parameter documentation is available on the help page.

You can also extract the normalized count data using dba.peakset() with bRetrieve=TRUE

[1614999088]

Hi Rory,

From the latest update, I can retrieve the data from dba.peakset() ,after doing:

data <- dba.peakset(dbObj, bRetrieve=TRUE)

Then use the following code to calculate the correlation of each sample and draw the heatmap：

> result <- cor(data) ## use pearson method 
> heatmap.2(result )

However, the heatmap obtained is not consistent with that drawn by using the default function dba.plotHeatmap(dbObj), and the clustering results are also different.Does it need to set other parameters.

Thank you very much!

[Rory Stark]

This is likely due to the default setting in dba.plotHeatmap() of bLog=TRUE, which uses the log2() normalized count values for the correlations.