Hi all

I am wondering if for ATACseq data, its better to merge all files [controls & disease] into one large consensus peak set for DiffTF/TOBIAS etc analysis or is having consensus peak sets for them separate a better idea - I ask this as DiffTF merges into one large consensus peak set however other tutorials say to merge yourself into two. If someone could explain what way is the logical / sound way I would greatly appreciate it.

For differential analysis tools like DiffTF and TOBIAS, the standard and statistically sound approach is to create one large consensus peak set that represents the union of all conditions (Controls Disease). This is necessary because differential analysis relies on constructing a rectangular count matrix where every sample is quantified against the exact same set of genomic coordinates; without a unified reference, you cannot validly compare signal intensities across groups or calculate fold changes. By creating a union set, you ensure that you capture regions accessible in at least one condition, allowing the tools to statistically measure crucial "on/off" events (e.g., a peak that is open in Disease but has near-zero counts in Control). The tutorials you saw recommending "merging into two" likely refer to an intermediate peak-calling step (pooling reads by group to increase sensitivity), which should still be followed by merging those group-specific lists into a final, single master consensus set for the actual quantification and analysis.