retaining only the clusters of interest

Question

[shamza]

Hey I am trying to subset only the clusters of my interest by the following command

diff2 <-  subset(merged_seurat_filtered, idents = c(0:16),)

There are 21 clusters in total and I want to omit the last 4 clusters from my Seurat object, but when I run the above command it gives me the following.

table(diff2$seurat_clusters)

   0    1    2    3    4    5    6    7    8    9   10   11   12   13   14   15   16   17 
9288 8539 8342 7552 6088 4697 4645 4251 3857 3621 3144 3045 2241 1460 1446  591  505    0 
  18   19   20 
   0    0    0

the diff2 object is retaining all the clusters and just removes all the cells but I am trying to retain only 0:16 clusters which should look something like this

 0    1    2    3    4    5    6    7    8    9   10   11   12   13   14   15   16   
9288 8539 8342 7552 6088 4697 4645 4251 3857 3621 3144 3045 2241 1460 1446  591  505

Does anyone know what mistake I am making?

[Ram]

idents is probably a factor so even though you've removed cells with those level values, the factor level still remains. Don't worry about it.

[seidel]

Could it be that idents needs a character string instead of integers? Also, you have an extra comma in your call after c(0:16).

[rfran010]

I usually use select to choose which columns to keep, but not sure if this subset function is specific to seurat ( I am completely unfamiliar with it)? I am thinking of what I use from base R.

subset(merged_seurat_filtered, select = c(0:16))

Edit: after looking at the seurat function, I think what Ram said makes sense. I think with your command you are selecting cells (not clusters) with Ident = 1 to 16, which is why the cluster columns are still there. For example, you could subset this object with cells that have gene expression > 3, and wouldn't expect any columns to drop, just the cells. At least as far as I understand.

[Ram]

You're close. The subset function serves a similar functionality as the base, but since the Seurat object is an S3 object, it subsets based on a slot instead of on a column like it would a data frame. However, the slot contains a factor and the factor is not re-created with new levels, hence the 0 values for levels that existed before subsetting.