Celltypist showing invalid expression matrix

[aditya-sarkar441]

I tried running celltypist on the anndata object and it wqs giving me this error. The data is from a research paper : https://github.com/scCOVID-19/COVIDPBMC/

Error:

Exception: Invalid expression matrix, expect log1p normalized expression to 10000 counts per cell

Can you please help me out with this?

[ChuanXu1]

@aditya-sarkar441, as I downloaded and checked that dataset - please remove the proteome expression in that data, by command such as adata = adata[:, ~adata.var_names.str.startswith("AB_")]

[aditya-sarkar441]

ok. why do we remove this proteome data ?

I got these cell types : ILC, Tcells, Monocytes, B cells. Why am I not getting NK and subtypes of T cells (CD4T, CD8T) ? Also how can i keep only those cells which are predicted as T cells ?

[ChuanXu1]

@aditya-sarkar441, the prediction is purely based on gene expression, so proteome data should be removed. There should be subtypes showing up, did you select the right model (for example, Immune_All_Low)?

[aditya-sarkar441]

Yeah, I selected Immune_All_High

Get Outlook for Androidhttps://aka.ms/AAb9ysg

---

From: Chuan Xu @.> Sent: Monday, March 14, 2022 8:23:50 PM To: Teichlab/celltypist @.> Cc: Aditya Sarkar @.>; Mention @.> Subject: Re: [Teichlab/celltypist] Celltypist showing invalid expression matrix (Issue #14)

@aditya-sarkar441https://github.com/aditya-sarkar441, the prediction is purely based on gene expression, so proteome data should be removed. There should be subtypes showing up, did you select the right model (for example, Immune_All_Low)?

— Reply to this email directly, view it on GitHubhttps://github.com/Teichlab/celltypist/issues/14#issuecomment-1066899389, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ANQQQYUNLVW7L66HQARFKWLU75HH5ANCNFSM5QUPGZPA. You are receiving this because you were mentioned.Message ID: @.***>

[aditya-sarkar441]

@ChuanXu1 My Postdoc mentor is telling me that celtypist can predict dozens of cell types. I am not sure how to do this. Can you please help me with this ?

This is the command I'm using : pred_cont = celltypist.annotate(nature_cont, model = 'Immune_All_High.pkl', majority_voting = True)

[ChuanXu1]

@aditya-sarkar441, as I said, you can choose the "Immune_All_Low.pkl" model instead of "Immune_All_High.pkl"

[aditya-sarkar441]

Ok thanks, I'll try this out.