search

sqjin / CellChat

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

spatial data with cellchat #662

Open chrkuo opened 1 year ago

chrkuo commented 1 year ago

thank you so much for creating this tool

quick question regarding utilization of cell chat with spatial data:

https://htmlpreview.github.io/?https://github.com/sqjin/CellChat/blob/master/tutorial/CellChat_analysis_of_spatial_imaging_data.html

the cell-cell communication inferred in this example it seems like the starting object is a deconvoluted object with integration of scRNA seq data set is that correct? so the resolution is more than spot resolution right?

If I deconvoluted with giotto or CARD - is it possible to use the same object to then use cell chat?

wondering how to interpret a deconvoluted spatial data-set cellchat result

Thank you

sqjin commented 1 year ago

@chrkuo What is the output of giotto or CARD? Are they the proportion/probability of different cell types within each spot?

chrkuo commented 1 year ago

@chrkuo What is the output of giotto or CARD? Are they the proportion/probability of different cell types within each spot?

Yes it's probability. But also in the vignette provided is it spot level or it's cell cell

sqjin commented 1 year ago

@chrkuo In the tutorial, a cell group label is assigned to each spot based on the maximum probability of predicted cell types. Then CellChat computes the average expression of ligands/receptors within each cell group. Does this make sense to you?

chrkuo commented 1 year ago

@chrkuo In the tutorial, a cell group label is assigned to each spot based on the maximum probability of predicted cell types. Then CellChat computes the average expression of ligands/receptors within each cell group. Does this make sense to you?

But within a spot that could be 1-10 cells even with the maximum probability there's a chance that the ligand receptor is coming from a different cell correct? Unless the assumption is that the maximum probability is the ground truth

chrkuo commented 1 year ago

@chrkuo In the tutorial, a cell group label is assigned to each spot based on the maximum probability of predicted cell types. Then CellChat computes the average expression of ligands/receptors within each cell group. Does this make sense to you?

But within a spot that could be 1-10 cells even with the maximum probability there's a chance that the ligand receptor is coming from a different cell correct? Unless the assumption is that the maximum probability is the ground truth

Is it then possible to use other objects that's been deconvoluted?

sqjin commented 1 year ago

@chrkuo Your understanding is correct. I am preparing another version to consider the possibility of multiple cell types within each spot. At this moment, you should use the strategy of maximum probability and you can do it using other objects.

chrkuo commented 1 year ago

@sqjin

Thank you - that would be extremely helpful looking forward to that.

abulislam commented 1 year ago

Hi,

@sqjin @chrkuo

Cellchat for visiam ST data requires a dataframe consisting of the cell labels. We used CytoSPACE for cell annotation using external reference scRNAseq data. The CytoSPACE provides more than one cell type (different fraction) per spot ID.

SpotID | CellType TCGCGTAGCAGTGTCC-1 | mast_cell TCGCGTAGCAGTGTCC-1 | NK_cell

Now how to deal with this?

thanks