Positive and negative marker genes for integrated data

[dahun73]

Hello all,

I have a question about marker genes. When I get marker genes using "FindAllMarkers", there are two types of log2FC, positive and negative. In this case, when I combined two samples such as control and stim, are these marker genes from comparing with all other cells? Or, from comparing with two different condition? I don't know exactly about these log2FC values whether these are from comparing with entire cells or not. I want to check out differentially expressed genes between two samples each cluster.

I'm sorry for confusing question.
Thank you for reading

[saketkc]

FindAllMarkers takes each ident and compares it with all remaining idents. If you are looking for only positive markers, you can pass only.pos=T.

[saketkc]

Also, FindAllMarkers should be run on the RNA/SCT assay and not the integrated assay.

[dahun73]

Thank you so much :)
I can understand meaning of markers.

You said "FindAllMarkers should be run on the RNA/SCT assay and not the integrated assay.". If so, FindAllMarkers should be run separately such as control and stim, right..? Or, you mean I should on the RNA assay in seurat object?

I confused whether I can use FindAllMarkers after integration of two data.

[joseale2310]

@dahun73 No, you need to revert back to the RNA assay to use FindAllMarkers. Integration is useful for visualization.

[dahun73]

Thank you for reply. I totally agree that integration is useful for visualization.
Then, you suggest that I should run FindAllMarkers individually, right?
But I'm curious about running FindAllMarkers after integration. Because some of clusters have totally different cell numbers between control and case. In this case, I think those markers in the integrated data may be meaningful.
Also, if I run the FindAllMarkers with each sample, how can I compare with those markers between control and case? I think it is hard to compare side-by-side which genes are significantly different pattern.

[joseale2310]

Check the basic integration tutorial. After integration, revert back to the RNA Assay and perform DEA there.

DefaultAssay(immune.combined) <- "RNA"
nk.markers <- FindConservedMarkers(immune.combined, ident.1 = 6, grouping.var = "stim", verbose = FALSE)
head(nk.markers)

It says FindConservedMarkers but you can run FindAllMarkers instead. I really do not think you should use the markers from the integrated data, since the underlaying data has been modified for the integration.