Correlate number of cells with gene expression of one gene in scRNA Seq

Question

[Bine]

Good afternoon,

I would really appreciate any feedback on the following:

We have identified a gene and we want to understand whether a higher percentage of endothelial cells is correlated with a higher expression of this gene.

So far I managed to obtain a table of the percentage of endothelial cells per individual from our Seurat object. Now, in order to run the correlation, I need to extract the gene expression of this gene per individual. I am struggling to continue here as

a) I havent found a way yet to filter the Seurat object on a particular gene

b) I wonder how to combine the gene expression of this gene from different cells

I would like to end up with a table like this in order to run the correlation:

Sample  |   Percentage endothelial cells   | Gene expression 
1       |      20%                         |  30                                                                                                                                                                         
2       |      13%                         |  67 

....

Does it make sense? Any input highly appreciated!

Thank you!

[Zhenyu Zhang]

No comment on the method, but I am interested in the biology here. Aren't you just looking for genes specifically expressed or not expressed in endothelial cells? Don't they already exist in the Seurat DE tables?

[Bine]

I am interested in whether a higher expression of my gene of interest is correlated with a higher amount of endothelial cells. Where do I find the Seurat DE tables? I just started working with scRNA Seq! Thank you!

[Zhenyu Zhang]

https://satijalab.org/seurat/articles/de_vignette

Answer 1

[jv]

a) I havent found a way yet to filter the Seurat object on a particular gene

assume scrna is your seurat object:

# define a vector of genes of interest
genes <- c('gene_1', 'gene_2', 'gene_3')

# subset seurat obj for genes of interest
scrna_filtered <- scrna[genes,]

b) I wonder how to combine the gene expression of this gene from different cells

What you seem to be asking about is aggregating expression across a set of cells, for that you can use Seurat::AggregateExpression https://satijalab.org/seurat/reference/aggregateexpression

[antonioggsousa]

Hi,

Are you sure you can filter genes from a Seurat object?

I didn't try the new version 5 of the package, but at least in previous version 4 this was not possible (see old github issue: https://github.com/satijalab/seurat/issues/4958)

You would need to filter the matrix and create a new object from scratch.

I know you can do such thing (as your code suggests) with a SingleCellExperiment object though.

Just curious!

Best,

António

[jv]

Filtering by genes as shown above works for me using SeuratObject v4.1.3 and Seurat v4.3.0

> packageVersion("Seurat")
[1] ‘4.3.0’
> packageVersion("SeuratObject")
[1] ‘4.1.3’
> class(scrna)
[1] "Seurat"
attr(,"package")
[1] "SeuratObject"
> genes <- head(rownames(scrna))
> test <- scrna[genes,]
> dim(scrna)
[1] 16313  6569
> dim(test)
[1]    6 6569

[jv]

Worth noting that I'm not sure how the above filter applies if the Seurat object has multiple assays. The example above only has a single assay in scrna.

[Bine]

Thank you very much for your reply. For b) I think thats exactly what I need For a) this doesnt work for me (Seurat Object 5).

But I think I can actually combine it with step b) via

 kid.filtered_BAZ2B_aggregated<-AggregateExpression(
                                kid.filtered_new,
                                assays = "RNA",
                                slots="counts",
                                features = "gene_of_interest",
                                return.seurat = FALSE,
                                group.by = c("orig.ident")
                                )

I believe that in the @X slot I should have the aggregated values for that gene then, or?

[jv]

since you are using return.seurat = FALSE I believe kid.filtered_BAZ2B_aggregated should have a matrix stored under kid.filtered_BAZ2B_aggregated$RNA but things might be different in newer versions of Seurat. Run str(kid.filtered_BAZ2B_aggregated) and you'll see what's returned.