Dear community,

I am annotating one query set to a particular reference. In order to annotate this reference I tested two methods: SingleR (https://www.bioconductor.org/packages/release/bioc/html/SingleR.html) and CellTypist (https://celltypist.readthedocs.io/en/latest/notebook/celltypist_tutorial.html). The both methods annotate with the same classification the query sets, however for celltypist I have high confidence scores (median around 0.90) and in SingleR when I extract the deltascores the median of this scores is around 0.10.

I wonder how is do these methods calculate this confidence scores and how trustable are these?

Best Regards,

Manuel