Differential expression SCT

[DzenisKoca]

Hello,

I am performing NicheNet with SCTransformed-integrated data. I run PrepSCTFindMarkers() function before runing NicheNet for Seuratm but I keep getting error:

Error in if (any(unlist(model_median_umis) != min_median_umi)) { : missing value where TRUE/FALSE needed

Looks like that error comes from the fact that NicheNet wrapper subsets data before internally runing FindMarkers(). I silved issue by modifying nichenet_seuratobj_aggregate() function to include recorect_umi=FALSE internaly.

DE_table_receiver = FindMarkers(object = seurat_obj_receiver, ident.1 = condition_oi, ident.2 = condition_reference, min.pct = expression_pct, assay='SCT', recorrect_umi = FALSE) %>% rownames_to_column("gene")

Is this approach ok?

[browaeysrobin]

Hi @DzenisKoca

Sorry for the late reply.

I would guess this is correct, but I think the Seurat developers would be more qualified to answer this question. Could you ask them how they would analyze DE between two conditions on SCT-integrated data? (after subsetting the data to get only cells of one cell type)

[anaccsilva]

Hi, I was able to get a response from them:

"DE analysis should be performed on SCT corrected counts (data slot). You can use PrepSCTFindMarkers() to account for different library depths across datasets as shown in the v2 vignette."

I am not expert in coding, and I appreciate the help. With their response, how should I adapt your code to have the best analysis. Just adding SCT is not working.

Thank you, Ana

[anaccsilva]

Hi. Just checking to see if someone could help me with this issue, please? thanks

[browaeysrobin]

Hi @anaccsilva

We will solve this around the 18th of November

[anaccsilva]

Thank you. I am looking forward to it. :-)

[browaeysrobin]

Hi @anaccsilva

To handle this issue, I suggest you use this NicheNet vignette: https://github.com/saeyslab/nichenetr/blob/master/vignettes/seurat_steps.md

It offers a step-by-step approach on how to perform the analysis on a Seurat object. In step 2, you need to perform the differential expression analysis. In case of SCT-integrated data, you need to run the following code before the chunk of code written in step 2/

seurat_obj = PrepSCTFindMarkers(seurat_obj)

Then, you will have to adapt the FindMarkers command and add: recorrect_umi = FALSE as argument

DE_table_receiver = FindMarkers(object = seurat_obj_receiver, ident.1 = condition_oi, ident.2 = condition_reference, min.pct = expression_pct, assay='SCT', recorrect_umi = FALSE) %>% rownames_to_column("gene")

If you want to calculate DE later on for the sender cell types, you will also have to add the recorrect_umi.

Hopefully, you can do something with this information.

[anaccsilva]

Wow! That is great. You were very precise about the response time. :-) Thank you so much. I will try this code and let you know. Thank you so much again!

[anaccsilva]

Hi, sorry bother again with this issue. According to Seurat developers, we can't rerun PrepSCTFindmarkers more than once. So, if I've already run this code before, how should I proceed without running it again? Considering that I am following this vignette (For, what should I do in this part of the code? DE_sender = calculate_niche_de(seurat_obj = seurat_obj %>% subset(features = lr_network$ligand %>% intersect(rownames(seurat_obj))), niches = niches, type = "sender", assay_oi = assay_oi) # only ligands important for sender cell types

Thank you, Ana

[skyla5]

Hello,@browaeysrobin

I am performing PrepSCTFindMarkers() function before runing FindMarkers(）with SCTransformed-integrated data. I keep getting error: Error in if (any(unlist(model_median_umis) != min_median_umi)) { : missing value where TRUE/FALSE needed It bothered me for days. Please just checking to see if someone could help me with this issue? Thank you very much

[me-orlov]

Hello, I am wondering if anyone sorted out whether the correction suggested by @DzenisKoca for the wrapper above is legitimate? I'm new(ish) to coding in R and would prefer to use the wrapper if possible. Wanted to verify that this works. I'd really appreciate it if we could shed some light on the issue.

[Boehmin]

Also following up on this when running with the new nichentr function. I receive this error:

> DE_table_all = Idents(snRNAseq_GAS) %>% 
+   levels() %>% 
+   intersect(sender_celltypes) %>% 
+   lapply(get_lfc_celltype, seurat_obj = snRNAseq_GAS, condition_colname = "group", condition_oi = condition_oi, condition_reference = condition_reference, expression_pct = 0.10, celltype_col = NULL) %>% reduce(full_join) # use this if cell type labels are the identities of your Seurat object -- if not: indicate the celltype_col properly
Error in FindMarkers.SCTAssay(object = data.use, slot = slot, cells.1 = cells$cells.1,  : 
  Object contains multiple models with unequal library sizes. Run `PrepSCTFindMarkers()` before running `FindMarkers()`.

Any suggestions?

[csangara]

Hi @Boehmin,

Please refer to #258 for the solution.