Open apposada opened 2 years ago
Hi @apposada, this is correct. In that sense we have the very naive assumption in ANANSE that factors that decrease in expression are not directly responsible for transcriptional changes in the target cell type. Similarly, we currently do not consider repressors. This is something that we plan to look at, but haven't found the time or opportunity to do so.
Do you have any specific examples of TFs where this is an issue. Usually it helps to have examples with known biology to troubleshoot and/or explore these kind of changes. You could try to see what happens when you remove this filter, and see how that influences the results. Adding the option to remove this filter is relatively important, but it's exploring how this affects the output in a variety of biological systems that takes time.
One other option would be to also run ananse influence
"the other way around", ie switching target and source cell types.
Hi,
When looking at the results of calculating ananse influence over two networks, we (in the lab) have noticed that none of the factors in the output have a negative logFoldChange (meaning, only factors more highly expressed in the second network are reported). Upon checking in the code of ananse influence, we have seen this: https://github.com/vanheeringen-lab/ANANSE/blob/18995f01657db5e92d4558eff4c1e81d30ff088e/ananse/influence.py#L302
We are not sure we are understanding the rationale behind choosing only the more highly expressed factors. Couldn't it be possible to check transcription factors whose expression has decreased compared to the first network? We would be interested in checking this as well.
Thanks a lot for your time and effort!