Is it okay to just average the log2FC values across different cell types in pseudobulk scRNA-seq data?

0

Entering edit mode

14 months ago

JLee • 0

Hi!

I downloaded a differential gene expression summary data table like this from brainSCOPE.

enter image description here

All I need are the gene and log2FoldChange column. However, each gene's data is split into multiple different cell types, as indicated in the cell_type column. These are too many cell types to work with- so I just want to merge some cell subtypes into larger cell types- e.g., I want to merge Vip, Sst, Pvalb cells into just "interneuron".

In this case, can I simply get the mean of log2FoldChange values of Vip, Sst, Pvalb for each gene? I don't care about p-values because I will analyze log2FC values with threshold-agnostic methods.

scrna-seq rna-seq log2fc • 577 views

ADD COMMENT • link 14 months ago by JLee • 0

0

Entering edit mode

logFCs without stats are meaningless since logFCs can have large standard errors. Combining individual logFCs can introduce bias as high logFCs with large SEs can overwrite reliable but moderate logFCs, I don't see the point. If the data format does not allow your analysis then get the underlying data and analyse as needed.