renaming genes in Seurat v3

[zolotarovgl]

Is it possible to rename genes (features) in the seurat v3?
When I'm trying to run:

rownames(object) = new_names Error in dimnames(x) <- dn : 'dimnames' applied to non-array.

The problem here is that there is no @raw.data slot anymore. When trying to set gene names using GetAssayData(): rownames(GetAssayData(object, slot = "counts")) = new_names

Error in rownames(GetAssayData(object, slot = "counts")) = new_names : could not find function "GetAssayData<-"

I would be very grateful for any help with this issue.

[mojaveazure]

Hi Grygoriy,

It is not currently possible to rename features in Seurat v3. If you want to rename features, we recommend renaming the features in the expression matrix before creating a Seurat object. We have looked into adding this functionality, but trying to keep track of feature names across all Assay, DimReduc, and Graph objects is a tall order. We may build this functionality at a later date, but have nothing to report at this time.

[zolotarovgl]

Got it, thank you very much for a quick response. I've found a way around by just re-creating the datasets from the data that could be extracted from the objects. I also have a question left: if it's possible to perform batch correction in Seurat v3?

Thank you very much.

[andrewwbutler]

Please see this vignette for our new integration procedure in Seurat v3.

[mlukosev]

Hi all. I am having the same problems. I am trying to plot velocities on my integrated Seurat object based on anchors (following this vignette https://satijalab.org/seurat/v3.1/immune_alignment.html). I have a merged loom file for them, too. However, in order to project velocities on the integrated Seurat object, colnames on loom and Seurat objects need to match. Since I have integrated Seurat object from control and mutant my colnames are for example AAACCCAAGGTATTGA_1 or TTTGTTGTCCGTTTCG2 to mark from which original dataset it came from. Therefore ideally I would want to remove *. However, I am getting the same error as @zolotarovgl when I try this command: colnames(seurat.object) <- paste(substring(colnames(seurat.object),1,16),sep=""). Error in dimnames(x) <- dn : 'dimnames' applied to non-array Is there a way around it? @zolotarovgl could you explain more how re-created the datasets from the data that could be extracted from the objects? Thanks a lot for your help! :)

[andrewwbutler]

Hi,

Please see the RenameCells function to rename the cells. This should allow you to remove the trailing _* as long as that doesn't result in duplicate cell names.

[nh-codem]

Hi, I used to abstruct counts to reaname featrues, and then reconstruct seurat object: counts <- GetAssayData(sc,assay = "RNA",slot = "counts") colnames(counts) <- newid sc <- CreateSeuratObject(counts = counts,meta.data = sc@meta.data)

[genecell]

Hi, I used to abstruct counts to reaname featrues, and then reconstruct seurat object: counts <- GetAssayData(sc,assay = "RNA",slot = "counts") colnames(counts) <- newid sc <- CreateSeuratObject(counts = counts,meta.data = sc@meta.data)

You could also use:

colnames(sc@assay$RNA@counts)<-newid
colnames(sc@assay$RNA@data)<-newid

without creating the object again, it works for me.

[mihem]

@genecell actually it's rownames and not colnames; and assays and not assay; but otherwise you are right, works for me too.

This small helper function works for my purposes and may be useful for some other Seurat users

convertRownames <- function(seu_object) {
  lookup <- homologene::mouse2human(rownames(seu_object), db = homologene::homologeneData2)
  new_rownames <- lookup$humanGene[match(rownames(seu_object), lookup$mouseGene)]
  rownames(seu_object@assays$RNA@counts) <- new_rownames
  rownames(seu_object@assays$RNA@data) <- new_rownames
  rownames(seu_object@assays$RNA@scale.data) <- new_rownames
  return(seu_object)
}

[mihem]

While this worked well for me for Seurat v4, I had some troubles with this script with Seurat v5. So first remove genes that were not converted (so rownames with NA), and subset the object, and then rename @meta.features

convertRownames <- function(seu_object) {
  lookup <- homologene::mouse2human(rownames(seu_object), db = homologene::homologeneData2)
  new_rownames <- lookup$humanGene[match(rownames(seu_object), lookup$mouseGene)]
  rownames(seu_object@assays$RNA@counts) <- new_rownames
  rownames(seu_object@assays$RNA@data) <- new_rownames
  rownames(seu_object@assays$RNA@scale.data) <- new_rownames
  #remove columns with NA
  features_keep <- rownames(seu_object)[!is.na(rownames(seu_object))]
  obj_new <- subset(seu_object, features = features_keep)
  rownames(obj_new@assays$RNA@meta.features) <- rownames(obj_new)
  return(obj_new)
}

Note: You need to remove the line with scale.data, if you don't have scaled data. And you would need to adjust it, if you have further assay, like a sketch assay.