Annotation of spatial data

Question

[abiuma]

Hi,

I have data for mouse tissue sections using 10x visium. But I don't have the corresponding single cell RNA seq data for the same tissue. There are no publicly available scRNA-seq dataset for the same type of tissue from mice but there are scRNA-seq data available in humans. Can one annotate mouse tissue using scRNA seq data from humans? If yes, what are the shortcomings one can expect?

Thanks!

Answer 1

[Ming Tommy Tang]

you can map the human genes to the mouse genes and do a reference label transfer.

You may work with a wet biologist to help you annotate the clusters using marker genes. I have an example using Merfish data https://divingintogeneticsandgenomics.com/post/how-to-construct-a-spatial-object-in-seurat/

[abiuma]

Thanks! Your example using Merfish is helpful. I have one more query. If I am to annotate using label transfer, the human ref data will have gene names in different format. Is it okay to change it to mouse genome format and then perform label transfer?

[Ming Tommy Tang]

yes, you have to map the human gene name to the mouse name.

[abiuma]

Hi, I have been testing the mapping that you had suggested. I found tutorials using biomart to do the mapping.

convertHumanGeneList <- function(x) { require("biomaRt")

human <- useMart(biomart="ensembl", dataset = "hsapiens_gene_ensembl", verbose = TRUE, host = "dec2021.archive.ensembl.org")

mouse <- useMart(biomart="ensembl", dataset = "mmusculus_gene_ensembl", verbose = TRUE, host = "dec2021.archive.ensembl.org")

genesV2 <- getLDS(attributes = c("hgnc_symbol"), filters = "hgnc_symbol", values = x, mart = human, attributesL = c("mgi_symbol"), martL = mouse, uniqueRows = TRUE)

mouseGenes <- unique(genesV2[, 2])

print(head(mouseGenes)) return(mouseGenes) }

But when I do that I get a huge loss of genes. For example if the human data set has 20,000 genes, after mapping for mouse genes using biomart, I get only 8000 genes mapped. How to work around this issue?