search

satijalab / sctransform

R package for modeling single cell UMI expression data using regularized negative binomial regression
GNU General Public License v3.0
212 stars 33 forks source link

FindAllMarkers on SCT V1 assay or RNA on a single sample (non-integrated)? #168

Closed fahmadimoughari closed 1 year ago

fahmadimoughari commented 1 year ago

Hi,

I am using SCTransform V1 on individual samples (no integration is supposed to be done). When I wanted to find the markers of clusters within a sample that has been normalized using SCTransform V1, should I go back to RNA assay or use SCT assay for differential expression analysis (FindAllMarkers)?

Regards, Fatemeh

saketkc commented 1 year ago

You should still be able to use 'PrepSCTFindMarkers' to perform DE using the SCT assay. I would recommend installing the latest develop branch of Seurat where we have introduced substantial speedup to that function.

fahmadimoughari commented 1 year ago

Hi Saket,

Thanks for your quick reply. As I understood 'PrepSCTFindMarkers' should be run on the merged object, where we have different samples in one object. But What I was asking for was a single object containing a single sample. Do I still need to run 'PrepSCTFindMarkers' before DE analysis?

saketkc commented 1 year ago

Right, you don't need to run PrepSCTFindmarkers if you have one sample (your initial post seems to suggest otherwise). It won't do any changes even if you were to run it.

fahmadimoughari commented 1 year ago

Thank you, Saket. I appreciate your help. Can you help me with my main question: Considering that this is an object contacting only one sample and I ran SCTransform V1, which slot shall I use for DE analysis? SCT or RNA?

saketkc commented 1 year ago

SCT is fine in this case as long as the object contains only one sample (i.e. it is not a merged object with multiple samples which were individually processed through SCtransform) - it throws an error in any case.

fahmadimoughari commented 1 year ago

I am grateful for your help, Saket . Just have a slight confusion regarding your answer. I can infer three meanings from it:

Sorry for asking many questions, I really appreciate your time.

saketkc commented 1 year ago

In the v2 paper we show clear advantages of using the SCT assay. That said, it is not incorrect to use the RNA assay as long as you have run NormalizeData. Hope this helps.

fahmadimoughari commented 1 year ago

I know that SCT V2 is much better to be used both for normalization and DE analysis, but for some reason, we decided to try SCT V1, My question was in the case of using SCTransform ""V1"", which slot is the right slot for DE analysis?

saketkc commented 1 year ago

You should use the data slot when using SCT assay.

fahmadimoughari commented 1 year ago

Sorry for the mistake in my last message, I meant which assay (not slot) to be used for DE analysis (RNA or SCT) when using SCT V1 (not V2)?

saketkc commented 1 year ago

I would recommend using SCT assay but in such scenarios it is not a bad idea to run both.

fahmadimoughari commented 1 year ago

Thank you so much. I appreciate your help and quick reply.