DiffBind dba.contrast with design and/or categories parameters

Question

[Orange]

Hello

I am analyzing bulk ATAC-seq data using ENCODE atac-seq analysis pipeline, then DiffBind.

I have 5 treatments (will be stored in "Condition" column in DiffBind sample sheet) and 3 biological replicates ("Factor" column).

Q1) When I call dba.contrast(), am I correct in using the design as follows? From this previous post, it looks like the categories option should NOT be used together with the design parameter. Is this correct?

dba.contrast(dbObj, design="~Factor+Condition")

Q2) When I was using DESeq2 with bulk RNA-seq, I used similar design formula as above and extracted pairwise results of interest using the result() function. Is the process similar for DiffBind where I run dba.count(), dba.normalize(), dba.contrast(), dba.analyze(), and I can extract any pairwise comparisons between the 5 treatments later?

Q3) If anyone has experience feeding ENCODE atac-seq pipeline results into DiffBind, is this the correct narrowpeak file to use? My understanding is that I should use peak files that have NOT gone through IDR or inter-sample overlapping.

trim.merged.srt.nodup.no_chrM_MT.tn5.pval0.01.300K.bfilt.narrowPeak.gz

Thank you in advance for your inputs!

Answer 1

[Orange]

For anyone looking at this in the future

Q1) No need to use the categories argument when using the design argument.

Q2) The design formula such as above will automatically compute all pairwise comparisons. To extract comparisons of interest, first display the "contrast number" representing each pairwise comparison.

dba.show(dba_obj, bContrasts = TRUE)

Then, the data for the comparison of interest can be extracted with

dba_res <- dba.report(dba_obj, contrast = num)

Q3) I did use the blacklist filtered non-IDR narrowpeak files.

Hopefully it's helpful!

[GenoMax]

Thank you for taking the time to come back and post a solution to provide closure.