Wrong coordinates of peaks in DiffBind

Question

[grant.hovhannisyan]

Hi All,

I'm analyzing ATAC-seq data and use MACS2 for peak calling and DiffBind for occupancy and affinity analysis.

In occupancy analysis when I try to find peaks that are unique to one group, it seems that DiffBind confuses the contigs and coordinates of the peaks. For example I get,

SU68  390602  392045  1444      * 2.191028e-03
SU39  136859  137210   352      * 2.175033e-03
SU79  564195  565710  1516      * 2.161226e-03 

but contigs SU68, SU 39 and SU79 in reality are very short (ca 1kb each), so it is not possible that peaks are located at the coordinates that DiffBind displays.

MACS2 output files seem fine, so I guess this is either a bug in DiffBind or it somehow changes the coordinates of chromosomes. Any ideas?

Here is my code after loading the peaks:

dba.overlap(SU_vs_hybrid, SU_vs_hybrid$masks$Parent, mode = DBA_OLAP_RATE)
dba.overlap(SU_vs_hybrid, SU_vs_hybrid$masks$Hybrid, mode = DBA_OLAP_RATE)


SU_vs_hybrid_unique<-dba.peakset(SU_vs_hybrid, consensus = DBA_CONDITION,minOverlap = 0.33)


SU_vs_hybrid.OL<-dba.overlap(SU_vs_hybrid_unique,SU_vs_hybrid_unique$masks$Consensus)
only_parent_peaks<-as.data.frame(SU_vs_hybrid.OL$onlyA)

Thanks

Disclosure: Cross-posted to Bioconductor support

[grant.hovhannisyan]

Has anybody experienced the same?

[aboyd003]

Hi. I am having this issue with my ChIP-seq data which used MACS2 and then DiffBind. Did you figure out the issue? I tried the suggestion from Rory Stark of explicitly setting the data type to data frame but the issue persists.

[grant.hovhannisyan]

Hi, no I did not find a solution unfortunately. The suggestion of Rory Stark also did not fix it.

[aboyd003]

Thank you for your reply. I'm going to try the HMM based THOR program then.

Answer 1

[grant.hovhannisyan]

Rory Stark, the developer of DiffBind answered the question in Bioconductor Support.