Compare absolute cell_abundance in two different visium sample is allowed?

[erganzi90]

Thanks for this briliant job, cell2location team.
10X visium sample1: colorectal cancer, celltype_A abundance A1
10X visium sample2: colorectal cancer, celltype_A abundance A2
If two sample has the same hyperparameters [N_cells_per_location and detection_alpha], can I compare A1 with A2 directly? Or just compare their proportion?

[vitkl]

Hi @erganzi90

to compare you need to run cell2location on both samples together - otherwise, the analysis regresses out differences between them. This applies to both comparing absolute values and proportions.

We did condition comparisons in the Nanostring hackathon https://github.com/vitkl/SpaceJam and unpublished COVID work. It helps to have some way to assess replication such as several ROIs within each sample rather than comparing averages across all locations in the sample.

Differential analysis is also likely better with segmentation priors #344 (comment).