markers to define a cell population in a single UMAP

Question

[Morris_Chair]

Hello guys,

I'm doing single cells analysis using Seurat pipeline and after the cluster distribution (UMAP) I found markers unregulated and downregulated in each of them. I want to know if it's possible to define in a single UMAP plot a specific cluster by including or excluding specific markers for example:

Cluster 1: A+B+C- Cluster 2 A+B-C+ Cluster 3 A-B+C+

Does exist a code that when I give the information (A+B-C+) automatically defines and shows in UMAP the cluster 2?

Or is it possible represent all of three markers in only one plot?

Thanks for help

[ATpoint]

Can you explain a bit more? Do you want to make a new clustering with all cells (or only some cells?) based on these genes? Or do you want to color cells by gene expression? I do not really understand.

[Morris_Chair]

Hi ATpoint,

I want to make a new clustering with all cells based on three selected genes.

Thank you

[ATpoint]

It is possible to use custom dimensional reductions in Seurat, please check https://satijalab.org/seurat/v3.0/dim_reduction_vignette.html

Storing a custom dimensional reduction calculation is the passage you probably want. This then would include only the genes you want. Not discussing if a clustering based on three genes actually makes sense (in fact I do not know).

[Morris_Chair]

ok sorry I was not clear, I don't have to make a UMAP only using three genes but I want to show where the cells that express this genes are located. For example if I have three population of cells in the UMAP and the population 2 is A+ B- C+, I want to show cells A+ and C+ instead of two separated UMAP with a single marker A and C

[sydnor.lindsey]

This is quite old, but there's always the magick package where you can make overlays of the two plots, in the worst case scenario. You can create a feature map showing the genes you want, and have that overlayed with high transparency on the UMAP clustering you want. To get the dimensions to match well, you'd either have to extract and configure the legends yourself or just plot these both w/ no legend.