Comparable across batches

[baigal628]

Hi,

Thank you for developing this nice tool. I have mouse data from different time points and strains. I ran through them via cellphoneDB individually. I wonder if the output means and p-values are comparable across different runs.

Thanks, Gali

[luzgaral]

Dear @baigal628

Thank you for using CellphoneDB.

What do you mean by different runs?

P-values should be comparable as long as your background populations are comparable. However, expression means may be susceptible to batch effects (unless you correct the expression counts directly).

Best

Luz

[baigal628]

Hi,

Thanks for the reply. It makes sense to me. I have another question. When I ran on a large sample, I realized that not all the cell types in the input meta file are included in the pvalue.txt/mean.txt. I assume all the cell-cell interactions will be reported in these two files even if they are not significant, right? For example, if I have 36 cell types in my meta file, I expect I have 36x36 cell-cell interactions as the output. Is it possibly due to the downsampling size being too small so that they are not captured? Thanks, Gali

[luzgaral]

Hi,

All cell types should be in the means file unless they are not in all the input files (e.g. counts, meta). The name should be exactly the same. Have you checked that?

Best,

Luz