In cases the same cluster is estimated as initial and terminal states

[st-tky]

...
Hi Cellrank team,

Thank you for developing this wonderful tool!
I have a general question about initial and terminal state estimation using GPCCA.

In short:
I am wondering if it is acceptable that the same cluster is estimated as initial and terminal states.
If it is acceptable, that would be OK for me.
If not acceptable from the theoretical point of view, I would appreciate any suggestions to workaround.

The following is a supplemental explanation:
I am following the instruction of "Computing Initial and Terminal States" on the website now.
In my datasets, the same cluster is estimated as initial and terminal states.
Because of the overlap, I need to set an option as follows: g.predict_initial_states(allow_overlap=True)
I understand that is why I get the same cluster, but I have to do so; otherwise, the error will come up due to the overlap.
In my case, it may be possible that stem cells (initial state) can differentiate into other cell types (terminal states) and also self-renew (possibly terminal state, too). Also, it may be possible that de-differentiation occurs or that other stem cells can transform into the other type of stem cells.
Because I am trying to find a new stem cell population that is currently not fully appreciated, it is very difficult to judge if this result is not correct or biologically acceptable.

I would appreciate it if you could give me any suggestions or explanations about it.

[Marius1311]

Hi @st-tky, this is a difficult question, and it depends a bit of what you want to learn from the data. If you already know that certain populations are stem cells, but there might be some self-renewal going on within that population, then there's a question of whether that's a biological aspect you're interested in. If you're not interested in that, because you know that these cell self-renew and you want to focus rather on their differentiation potential, then I would treat this state purely as an initial state, and remove it from the list of terminal states. It's okay to use prior biological knowledge where available to guide the inference process in CellRank, you should just be transparent about that in your manuscript.

[st-tky]

Thank you for your quick and kind reply!
OK, I will try several patterns by including or excluding those cells as terminal state.
I guess we can exclude the specific macrostates from the list of terminal state by g2.set_terminal_states(states=["A", "B", "C", "D"]).
Can we remove specific macrostates from the candidates of initial state, on the other hand?
I want to check as many possibilities as possible before diving into further experiments.

[Marius1311]

Sure, this should work the same way for initial and terminal states.

[st-tky]

Thanks a lot!
I will give it a try to see if there are any different outputs by changing those parameters.