How to understand the two results when converting to the control and treatment

[Nuturetree]

when A sample as a control and B sample as a treatment I can get a lot of split and a litter merge condition1 (A sample). But when B sample as a control and A sample as a treatment I also can get a lot of split and a litter merge condition1 (B sample).

such as:
A control B treatment
chr1 0 100000 split condition1
chr1 0 50000 nan condition2
chr 50000 100000 nan condition2
when B as control， the two TAD （chr1 0 50000 ，chr 50000 100000 ） not identified merge status？

[Tian-Dechao]

Your observation is right. DiffDomain cannot guarantee symmetric results when switching conditions A and B. The reason is that, using your example, the interactions between the two TADs ((i, j), i in chr1:0-50000, j in chr1:50000-100000) have the most significant differences between conditions A and B, compared to interactions within the chr1:0-50000 region and within the chr1:50000-100000 region. When B as control, DiffDomain only tests the two B TADs (chr1 0 50000 ，chr 50000 100000 ), excluding the interactions between the two TADs ((i, j), i in chr1:0-50000, j in chr1:50000-100000) since they are not part of B TADs.

Depending on your research goal, you can merge the two TADs lists in conditions A and B into a consensus TAD list and using the consensus TAD list as the input for DiffDomain. This can guarantee symmetric results. Alternatively, do both tests (A vs B, B vs A) and merge DiffDomain results.

[Nuturetree]

Thanks your reply！！