macs2 bdgdiff for calling differential binding events without control

[meihl]

I had one treated sample and one non-treated sample, but without Chip-Seq mock (or background control). But I want to identify regions that are differentially histone modification between control and treated. MACS2 bdgdiff is a prefect tool to complete the differential analysis of Chip-seq between two conditions. But how could MACS2 bdgdiff identify differential peaks without ChIP-Seq control. I really appreciate your help. Thanks a lot!

[taoliu]

I haven't tested if bdgdiff works fine on two conditions without control. Theoretically it should be able to run through but I haven't studied if the results look fine. In this case, I would suggest you to generate treatment bdg files for two condition from callpeak or pileup function, then compare them directly using bdgcmp.

[s-rashmi]

Hi Sorry to hijack the post, but we are having the same issue 4 ChIP experiments without mock IP data We are very confused as to which parameters to call. Please help! This is what we did 1) callpeak with --dup all --nomodel -extsize <average frag length from 4 datasets> --nolambda --call-summits -B dup all - because the redundancy rate in the datasets are high ~0.79 nolambda - because no control data nomodel & extsize as mentioned in bdgdiff tutorial

2) Which method should I choose for bdgcmp? I tried ppois and got a bed file. What does the last column indicate?

Also How do I use the -S (scaling factor) option to normalize among the 4 datasets?

[sperbat]

I haven't tested if bdgdiff works fine on two conditions without control. Theoretically it should be able to run through but I haven't studied if the results look fine. In this case, I would suggest you to generate treatment bdg files for two condition from callpeak or pileup function, then compare them directly using bdgcmp.

Hi Tao Liu, Could you please elaborate on how to compare the treatment bdg files for two conditions from callpeak or pileup function directly using bdgcmp, as you have suggested here? In your tutorial on how to build a signal track you explain how to generate the bdg files, how to compare them using bdgcmp, how to convert the _FE.bdg and _logLR.bdg files to bigWig etc. - but at the end you use wigCorrelate on the two _FE.bdg files from the two different samples, which is what we don't have in this case - only one treatment file and one control file, i.e., we only have one _FE.bdg file. Thank you so much for your help!