Clustering 10X scRNA data based on selected markers

Question

[A]

Hi

I am trying to cluster my single cell data using known markers with the aim of finding cell populations based on expression of specific markers. I am analyzing my data using the Seurat package and ran the following lines of code to get my desired output.

ctrl_BL <- RunPCA(ctrl_BL, approx=FALSE, verbose = FALSE)

ctrl_BL <- RunUMAP(ctrl_BL, dims=NULL, features=c("CCL3","CCL5"),verbose = FALSE)

ctrl_BL  <- FindNeighbors(ctrl_BL, dims=NULL, features=c("CCL3","CCL5"), verbose = FALSE)

ctrl_BL <- FindClusters(ctrl_BL, verbose = FALSE)

DimPlot(ctrl_BL, label = TRUE) + NoLegend()

However, the resulting plot looks a bit weird to me (see attached). Could you advise me if this is expected or if I am doing anything wrong. I am new to this type of analysis so suggestions are welcome.

[ATpoint]

You are running the analysis based on two genes, that is in no way meaningful, and this is not how the analysis usually goes. You select genes that are informative (highly-variable, can be thousands), then do clustering and manifold embedding (UMAP), and then use markers (either a priori or based on differential expression) to assign identities to your clusters. Please go through the Seurat vignettes and follow them, there is most likely no reason to change the usual workflow unless you have expert knowledge and strong reason to do so.

As you used two genes here the analysis runs essentially without information (as two genes are basically no information), therefore I guess this is why it looks oddy.

[A]

Thanks ATpoint. Yes I have gone through the vignettes and have done analysis in the standard recommended way. This was an additional analysis that we wanted to explore. I first tried with two markers but eventually will have about 10-15 markers that I would like to use. Will try and see if that looks different.