search

Tian-Dechao / diffDomain

DiffDomain is a statistically sound method for detecting differential TADs between conditions
MIT License
11 stars 3 forks source link

How to use replicates? #11

Open imerelli opened 9 months ago

imerelli commented 9 months ago

Hi, I have replicates for the two conditions to test. How I can I use them? I don't see option to use replicates.

Tian-Dechao commented 9 months ago

Hi Ivan, in case each condition has multiple Hi-C replicates, DiffDomain takes the combined Hi-C contact matrix from the replicates as the input which is a common practice to generate a large number of Hi-C interactions. So in your case, please merge the replicates in each condition into a combined Hi-C data and then feed the combined data to DiffDomain.

imerelli commented 9 months ago

Hi Tian, thank you for your answer. The problem is that I was explicitly asked to take into account replicates in order to have statistics about the different tads, like in RNA-seq, therefore merging is not an option. I was wondering if you are planning something in this sense, creating tad by tad a sort of distribution and then testing it case vs control. I know it's difficult for intervals, such as for atac-seq, but for example diffbind implements a similar approach.

Tian-Dechao commented 9 months ago

Hey Ivan. Current version of DiffDomain cannot take the replicates as the input. We are sorry about this. Indeed, it is challenging for comparing TADs using replicates Hi-C data. We will take this as a future work. Please stay tuned.

Meanwhile, I would recommend trying DiffDomain on combined Hi-C data. From our extensive analysis, using only combined Hi-C data achieves reasonable results (see our bioRxiv paper here)