Difference between cell search and cell type annotation

[Mobina-Kazemi]

Hi,
I have been using your tool for annotating my single cell data. I have been doing the analysis using your celltype annotation tutorial. I noticed as this process is cell by cell, it might not be in line with my clusters inferred from seurat analysis. I saw you also have a tutorial for cell search which as far as I understood is different from celltype annotation. My question here is if there is a way that I can do the celltype annotation based on clusters and get one label and score of assignments for each cluster? In my mind this way should be more accurate as I have multiple cells in each cluster that can be used to support the results of celltype annotation. thank you.

[javh]

Greetings @Mobina-Kazemi , sorry for the slow reply.

No, sorry, there's no way to directly way annotate clusters of cells instead of individual cells. We also find that cell-level annotations are often more accurate, as clustering comes with some "contamination". If you plot out a few canonical marker genes for differing annotations within a cluster, you'll sometimes find cells clustered together that you didn't expect (or doublets, dying cells, etc).

That said, you could use the most abundant annotation in your cluster for the full cluster annotation or, alternatively, collapse your clusters into a pseudobulk profiles (summed UMI counts) and annotate the pseudobulk profiles. I haven't tried the latter, so YMMV.