search

jinworks / CellChat

R toolkit for inference, visualization and analysis of cell-cell communication from single-cell and spatially resolved transcriptomics
GNU General Public License v3.0
295 stars 52 forks source link

Uneven cell numbers and subsampling #68

Open LinearParadox opened 10 months ago

LinearParadox commented 10 months ago

Hi all,

I had another question! Is it worth subsampling single cell objects when you have uneven cell numbers in conditions?

Basically our experiment has 5 conditions:

primary tumor(pt) -- 18000 cells pt_treated -- 22000 cells drug1_resistant (drug1R) -- 9000 cells drug1_treated -- 11000 cells drug2_resistant(drug2R) -- 9000 cells

All the treated samples were treated with drug2. What's driving this question is when we look at the interaction strength and number of interactions, this seems to be somewhat driven by cell number. For example:

num_interactions (1)

I'm more concerned about the first two conditions, as they seem to be around 2 times more cells than the other two, and I'm not sure if that is driving the effect we see, or whether this is a nontechnical effect.

sqjin commented 10 months ago

@LinearParadox When you run computeCommunProb by using the default setting of population.size = FALSE, the population size will not affect the results.

LinearParadox commented 10 months ago

@LinearParadox When you run computeCommunProb by using the default setting of population.size = FALSE, the population size will not affect the results.

Got it. What setting do you recommend setting it to true vs false? Or is it almost always better to just use False?

sqjin commented 10 months ago

@LinearParadox It is better to just set it as FALSE because single cell sequencing often produces bias estimation of the truely cell proportion in the tissue, in particularly when doing cell sorting.