RPKM on TSS using DiffBind

Question

[g.persic1991]

Hi everyone, I want to extrapolate RPKM value from Diffbind. Investigated regions are TSS (± 2.5kb) but around 3500 TSS are "lost" because merged. I have read that in the new version of DiffBind is possible to use dbObj_region$config$mergeOverlap with negative value to avoid merge but I'm not able to use.

Below my code

library(DiffBind) 
region <- read.csv("DiffBind_template_peaks.csv")
DBA <- dba(sampleSheet=region)  
DBA$config$mergeOverlap = -1
DBA_Count <- dba.count(DBA, score=DBA_SCORE_RPKM)  
rpkm <- dba.peakset(DBA_Count, bRetrieve=TRUE)

Thanks a lot in advance

ciao Giuseppe

Answer 1

[Rory Stark]

The value of DBA$config$mergeOverlap is the number of overlapping bases required for regions to be merged. So if you are starting with regions that overlap and you don't want them to be merged, you should use a larger positive value. For example if you set DBA$config$mergeOverlap <- 100, then regions which overlap by 100 or fewer bases will not be merged.

Negative values for DBA$config$mergeOverlap are used to merge non-overlapping regions that are close to each other. So if you set DBA$config$mergeOverlap <- -100, then regions which are within 100 bases of each other will be merged even if they don't overlap.

[g.persic1991]

Hi Rory,

Thanks for your suggestion, however I cannot resolve the problem.

After retrieving TSS list (n = 23069) and put this as bed file for each samples, I proceed with DiffBind

library(DiffBind)

file<-read.csv("DiffBind_template_H3K9me3_23362_TSS.csv") 
DBA <- dba(sampleSheet=file)

I set this parameters (as suggested)

DBA$config$mergeOverlap <- 10000

and proceed with dba.count

DBA_Count <- dba.count(DBA, score=DBA_SCORE_RPKM, summit=FALSE, filter=0)

but if I check the regions I found 19996 (at beginning I have 23069)

Where I'm wrong?

Thanks in advance

ciao

Giuseppe