high variance genes

[jinxixiang]

Dear authors,

I attempted to filter genes based on variance using the following code:

adata = sc.read_h5ad(expr_path)
adata.var_names_make_unique()
sc.pp.filter_cells(adata, min_genes=1)
sc.experimental.pp.highly_variable_genes(adata, n_top_genes=TOP_GENE)

However, I observed that only half of the samples (594 out of total slides, TOP_GENE=128) exhibit high variance genes, while many of the remaining samples show little gene expression levels.

Could you please provide your insights on these findings? Thank you!

[pauldoucet]

Dear @jinxixiang ,
Do you mean that for half of the samples, the top 128 most variable genes have low expression levels?
Can you clarify how small these gene expression levels are compared to the other genes? Normalization might be necessary

[jinxixiang]

Thank you for your reminders. I will review the norm first.