Hi James,

I am not sure if I undestood your question completely. There's an explosion of the availability of tools for scRNAseq analyses (https://www.scrna-tools.org/), (https://www.nature.com/articles/s41581-020-0262-0).

Earlier this month, a comprehensive study about 1000 scrnaseq tools was reported in bio-archive (https://www.biorxiv.org/content/10.1101/2021.08.13.456196v2).

I assume that you are already familiar with Seurat which is one of the most widely used packages for scRNAseq data analyses. Monocle and Magic are also popular.

'Cell type annotation' should be based on cell type specific markers which can be obtained from the literature. Unsupervised clustering often leads to many potential cell types which should be confirmed by other methods and if possible also by functional assays (to make sure that they represent biologically relevant cell types and not just cell states). Findallmarkers function in Seurat gives genes which are differentially expressed between clusters. One has to be aware that sometimes immediate early genes (i.e.dissociation induced) can create un-real cell types. It helps to perform functional enrichment analysis of cluster markers.

All the best,

Tushar

Thanks, Tushar Deshpande for the recommendation of tools! Definitely I will check them out. I guess my question is more about automatic cell type annotation, rather than manually looking at the marker list ourselves and refer to the literature, cluster by cluster. Have you tried cell type prediction in Seurat, or do you suggest any tool to do this? Appreciate your recommendations!

James Kaczerski

Perhaps you know this:

https://satijalab.org/seurat/articles/integration_mapping.html

This approach involves reference based annotation.

You can try SingleR

Go through their documentation, it's well self explained;

https://bioconductor.org/packages/release/bioc/html/SingleR.html