Wrong avg_logFC?

[JerryZhang-1222]

Dear developer,

I've been using Libra for several projects and I noticed that the avg_logFC does not equal to log(group1.exp/grou2.exp).

It's also weird that in this screenshot, the first row, foldchange equals to 5.4788/1 while the second row equals to 1.9859155/0.3636364 = 5.461267, which is smaller. However, the logFoldchange of the second row is bigger than that in the first row.

I wonder how's that happened and how did foldchange been calculated? Thank you very much!

Best,
Feiyang.

[mph270]

Hello,

I am also getting odd results with avg_logFC. For some rows, the directionality of avg_logFC matches that of the change in expression (WT.exp vs KO.exp) but for other rows it is the opposite directionality (for example below, rows 1 and 3). N.B. This is a separate issue from the assignment of the reference group.

Thank you for your help!

Best,
Molly

[mph270]

Here is another example after making genotype and animal ID (orig.ident) factor levels both alphabetical and filtering out lowly expressed genes from the results. The avg_logFC value for the third row seems very off.

I used the following code:

DE_edgeR_LRT = run_de(mapped_seurat, cell_type_col = "new_ident", label_col = "genotype", replicate_col = "orig.ident")

Thanks again for your help and for creating this package!

-Molly