Applicable to spatial transcriptomics?
Opened this issue · 1 comments
Congratulations on the preprint.
I am quite interested in judging segmentation quality for my data.
I work with spatial transcriptmomics, where were retrieve discrete counts for measured genes in the cells.
The technology is very highly multiplexed, I could probably pick out 50 high quality markers.
Do you think your model would be applicable to that usecase? Should I transform my input counts somehow?
Thank you!
Hello,
Surely. Your data needs to be in a continuous distribution, so I suggest normalizing first.
The best use case for this model would be to define a subset of your high quality "lineage/important" markers to subset your expression matrix (for instance a cell by 10 marker matrix if say you have 10 lineage markers) and have the expected expression being defined in your prior matrix (this is to define them 1, 2, 3 as defined in the readme, so a expected cluster count by 10 markers). The important part would be to first define your cell types/clusters which is essential for your prior matrix.