Questions about the cell abundance

[crazybiology]

Dear Developers,

Thank you for creating this fantastic tool for spatial transcriptomics! I greatly appreciate the effort and innovation behind its development.

I have two questions regarding cell abundance in my dataset:

1. I am working with seven cell types and extract the cell abundance from q05_cell_abundance_w_sf. When I sum the individual cell abundances across these types, the result does not match the total cell abundance (sum_f w_sf) output by the cell2location function.

To calculate the relative abundance of each cell type in a spot, should I divide the individual cell abundance by the sum of the individual cell abundances, or by the total cell abundance (sum_f w_sf) provided by cell2location?

2. My dataset consists of tumor samples, where I expect the majority of cells to be tumor cells. However, the deconvoluted tumor cell abundance appears lower than anticipated. Could this suggest a potential issue in the analysis, or might there be other biological or technical explanations for this observation?

Thank you for your time and support. I look forward to your guidance!

[vitkl]

Hi @crazybiology

I am glad to hear that you find cell2location useful.

1. The model reports several values describing the posterior distribution - 5% quantile, mean, 95% quantiles, std. Its possible that the two ways of computing the total used a different summary - eg you used 5% quantile (default in the tutorial) vs total cell abundance (sum_f w_sf) provided by cell2location used mean.

should I divide the individual cell abundance by the sum of the individual cell abundances

This is correct ^^

2. It could be both. In general, 7 cell types is quite a small number for most tissues - using insufficient cell annotation granularity can lead to weird results - like trying to put on a 3-fingered glove on a 5-fingered hand - reference signatures for one cell type might have to explain the data for several cell types with distinct spatial patterns.