How to identify the cell type computationally when you have a list of conserved cell type markers for mouse genome (as discovered by Seurat)?

Question

[mrj]

I have a list of conserved cell type markers discovered from Seurat. I would like to annotate clusters with cell types based on that.

For example, here is a profile of cell type markers (from mouse genome) I am dealing with

Gm42418
Nme2
Rpl19
Malat1
AY036118
mt-Co3
mt-Nd2
Ahnak
Hnrnpu
mt-Atp6
mt-Atp8
mt-Nd5
Ly6g
Hist1h4d
Nedd4
mt-Nd1
Lars2
Xist
Prrc2c
Myof
mt-Nd4l
mt-Nd6
mt-Nd3
mt-Co1
Prpf4b
Hist1h1e
mt-Cytb
mt-Co2
mt-Nd4
Hist1h1b
Gm26917

how can I computationally predict the cell type if the genome is mouse mm10?

Answer 1

[Refine]

Hi ! I'm sorry I'm not good at English. I am not able to answer your question because I do not know how to automatically judge cell types, but I noticed some things in your result.

In my experience, Gm42418 and AY036118 genes are detected in noisy cluster for me, and I sometimes can not identify cell types. Similarly, "mt-"genes are also detected in noisy cluster and these genes are markers of dead cells. You might be confused by these factors ( and I do not know why they occur).

Parhaps they might be improved by some changes as follows.

1. Check whether nFeature_RNA & percent.mt are appriximately same as other clusters.
2. Use SCT method and regress percent.mt.(DE is by RNA assay.)

If possible, please show me your code ? I might be able to advise you a little more.

[mrj]

Thank you so much for your insight. It made me go back to my code and realize that I have not limited the presence of mitochondrial contaminated cells.

[Refine]

I am happy to hear that !