Hi, I've used DiffBind to analyze differential binding analysis for cut&tag data.

The result is showed in MAplot:

I noticed some peaks with 0 log fold change are differentially biding sites, while it could be a little bit confusing.

I see this result should be caused by the overall gaining of binding affinity, as the red curve indicates.

However, this data is spike-in normalized, and the overall gaining of binding affinity might be reasonable in our case.

So how can I get differential binding sites without considering the overall gaining of signal?

I can manually set a cutoff (e.g. abs(log FC) > 1) to determine, however the p-value and FDR should be re-caculated, which I don't know how to do in DiffBind.

Here is my code, just regular steps but with a spike-in normalization:

sample = read.csv("samplesheet.csv")
CTCF <- dba(sampleSheet=sample)
CTCF <- dba.count(CTCF)
scale_factors = read.delim("scale_factors.txt", header=T, row.names=1)
CTCF <- dba.normalize(CTCF, normalize=scale_factors$scale_factor)
MUT2_vs_WT2 <- dba.contrast(CTCF, design=FALSE, group1 = CTCF$masks$M2, group2 = CTCF$masks$W2, name1 = "MUT2", name2 = "WT2",  minMembers = 2)
MUT2_vs_WT2 <- dba.analyze(MUT2_vs_WT2)

Thanks for your kindly help in advance!

I'm wondering about how you calculated the scaling factors, they really need to be correct!

It would be good to compare the MA plot using bNormalized=FALSE.

Also, is there a reason you're using design=FALSE? I would think you'd want to set the contrast, ie

MUT2_vs_WT2 <- dba.contrast(CTCF, contrast=c("Condition", "M2","W2"))

Have a look at the data returned by dba.report(), especially the Fold values. Do these values look right gievn the Conc_ values for the two groups?