I am using Diffbind to compare H3K27ac peaks between two time points. Recently, we added 4 more time points to the existing 2 to do time-series comparison. From my understanding, Diffbind could conventionally compare and provide differential bound sites between 2 samples .

Q1: Is there a way to compare and contrast 6 time-points, considering each has 2 biological replicates. Q2 If the answer to Q1 is no, when I compare A to B and separately A to C , why do I get different counts for A in the binding affinity matrix arising from A vs B to A vs C for the same peak interval.

DiffBind supports the use of any GLM design formula accepted by the underlying analysis packages DESeq2 and edgeR. I try to avoid getting involved on telling people how to set up their GLM design.

Regarding the different numbers of counts, that depends on how you are setting things up, Are you talking about the values you see when you call dba.peakset() with bRetrieve=TRUE? In that case there is only one matrix to see. Are you talking about normalized read counts returned by dba.report() when bCounts=TRUE? In that case it depends on how the contrasts are set up. For example, if design=FALSE, the normalization will be different for each contrast.

If you can show some more of the script that you are using where you are seeing different counts for the same sample, I may be able to see what is happening in your case.