Open rocanja opened 3 months ago
When selecting multiple 'tissues', the gene_sets_prepare() function somehow drops some cell types and doesn't include them in the analysis ... and depending on the order the tissues are selected, different cell types get dropped (don't know why, but it happens) - hence, DO NOT SELECT MULTIPLE TISSUES!, it corrupts your analysis! Workaround is to make a custom data base, add all the cell types you need but label them with one tissue identifier. If you select this 'mock' tissue, all cell types get imported correctly and included in the analysis. It would still be nice to have the capability of selecting multiple tissues in a future version of the package, as there are applications where this is needed and makes sense e.g. metastatic samples.
For now, we do not support the explicit selection of multiple tissues simultaneously. While this functionality may occur coincidentally, it has not been formally implemented. We plan to include this feature in the upcoming ScType 2.0. Thank you for your suggestion; we are keeping this issue open for now.
It appears to be possible to select more than one tissue for the analysis (at least the code runs fine in R), but is there any problem or big drawback of this approach and does the order of the selected tissues matter in some way (EDIT: YES and YES!, see below) ? I am looking at metastatic samples and hence, would ideally like to include all cell types from the tissue of the metastatic site, as well as the organ the tumor initiated from. I am currently exploring the provided option to use custom tissues and cell type marker genes, but I would love to see an expansion on the current tissue data base ... and could you please bring back 'tissues comprising less than 5 cell types', which were excluded during the curation of the current version? Thanks for this great tool!