search

sqjin / CellChat

R toolkit for inference, visualization and analysis of cell-cell communication from single-cell data
GNU General Public License v3.0
637 stars 145 forks source link

How are receptor-receptor interaction counted as singal out/in? #502

Open flde opened 2 years ago

flde commented 2 years ago

Hi @sqjin,

Many thanks for the great tool! I checked Vcam1 in the mouse DB and it is registered as receptor while its agonists Itga4+Itgb1 are registered as ligands. The later however assemble the cell surface receptors α4β1 integrin.

How is the signal in/out computed in this case when you have two cell surface receptors interacting with each other? In my understanding, right now, you count the cells expressing Vcam1 as receiver (receptor) and the Tga4+Itgb1 (ligand) positive cells as sender?

That would be a great information since we have some obscurely low signaling cells in the signal in/out plot due to receptor-receptor interactions.

Best, Florian

sqjin commented 1 year ago

@flde Thanks for pointing it out. Yes, CellChat will compute the average expression of Tga4+Itgb1 positive cells as the sender cells. In CellChat, this signaling is in the category of cell-cell contact. What is the best way do you think we should do for these receptor-receptor interactions?

flde commented 1 year ago

Dear @sqjin,

Many thanks for your help! If I understand correctly the annotated type of interaction (e.g. l-r, r-r, ecm-r) should not matter for computing the communication strength and associated p-values. It should also not matter for the count of interactions. However, it might bias the weighted-directed network with incoming and outgoing signals, right? Because by chance a receptor-receptor (r-r) interaction could either be annotated as incoming or outgoing signal for a given cluster. Maybe r-r interaction could be annotated as bidirectional or just duplicated with reversed labels for the ligand and receptors?

On another note, do you think it’s a good idea to try computing an cell-cell spatial proximity score based on the number and strength of r-r interactions? I would just use such score to approximate which cells co-localize but it could also be incorporated into the computation of the interaction probability? Again, many thanks for your great work and the tutorials!

Best, Florian