Hi folks,
I am new to ChIP-seq analysis hence this may be a basic question. I want to use DiffBind to get peak summits, but after running dba.count with my own peaks, I find the number of peaks has decreased. Is this normal? How did this happen?

library(tidyverse)
library(DiffBind)

tf <- dba(sampleSheet=tf_samples)

peaks=read.table('final_merge.bed')
colnames(peaks)=c("Chr","Start","End")
dim(peaks)
[1] 1271    3

tf <- dba.count(tf , peaks = peaks,bUseSummarizeOverlaps=TRUE,
                score=DBA_SCORE_READS)
dim(tf$peaks[[1]])
[1] 1261    8

sessionInfo( )

DiffBind merges overlapping peaks. Even if there were no overlap in the original peakset, the default behavior of dba.count() is to re-center peaks around consensus summits. These re-centered peaks could overlap if there were original peaks closer together than the re-centered width (default 401bp). Any peaks within 400bp of each other are candidates for being merged, depending on where the computed consensus summit is located.

You can avoid this behavior by changing the default value of the summits parameter by setting summits=FALSE. Alternatively, you can see what it is doing in more detail by setting summits=TRUE and score=DBA_SCORE_SUMMIT_POS, then retrieving the resulting peaks by calling dba.peakset() with bRetrieve=TRUE and seeing where the computed summit is for each peak. If the summits of two adjacent peaks are within 400bp of each other, they will be merged in the default case.