How to filter single cell RNA data if only 2 of the many cell types to be analyzed? Filtering over all or part?
0
0
Entering edit mode
12 weeks ago
entropy ▴ 50

How to filter single cell RNA data if only 2 of the many cell types to be analyzed? When it is filtered with a standard filtering of Seurat, it filters all the data. But I will perform like DE analysis only over the cells that belongs to 2 cell types out of the many like 16. Since I will not perform any downstream analysis for the rest of the 14 cell types, should I first subset the cells that belong to the 2 cell types only and apply the standard Seurat filtering or should I still filter using all the data including the cells of those other 14 cell types?

single-cell scRNA scRNA-seq • 497 views
0
Entering edit mode

You need to add something to illustrate the problem, like plots. Textual descriptions are hard to follow, we cannot see your screen.

0
Entering edit mode

I am just asking should I filter the data with all the cells of all the cell types, which is the standard way as we see in Seurat. Or should apply that standard filtering to only the cells of the 2 cell types only since I am going to perform DE analysis over only these two cell types? Namely, should I get the subset of the cells of the 2 cell types first and the apply this standard filtering to only that subset of the data?

pbmc <- subset(pbmc, subset = nFeature_RNA > 200 & nFeature_RNA < 2500 & percent.mt < 5)

0
Entering edit mode

Seurat subset will allow you to filter on any of the Seurat object meta data vectors. Do you have the cell types stored as meta data in your seurat object? Alternatively if you know the UMI of the cells of interest you can also use subset with cells paramter:

subset(
x,
subset,
cells = NULL,
features = NULL,
idents = NULL,
return.null = FALSE,
...
)


also see https://satijalab.org/seurat/articles/essential_commands.html for more examples with subset

0
Entering edit mode

Thanks for the reply but I am not asking How to filter. But I am asking should I filter the data with all the cells of all the cell types, which is the standard way as we see in Seurat. Or should apply that standard filtering to only the cells of the 2 cell types only since I am going to perform DE analysis over only these two cell types? Namely, should I get the subset of the cells of the 2 cell types first and the apply this standard filtering to only that subset of the data?

pbmc <- subset(pbmc, subset = nFeature_RNA > 200 & nFeature_RNA < 2500 & percent.mt < 5)

1
Entering edit mode

thanks for the clarification. I imagine either strategy would be fine, the main thing is to tailor your filtering strategy to appropriately filter out low-quality cells from what ever population of cells you are starting with.

Do you see considerable disparity in the nFeature or nCount metrics between your two cell types of interest and/or between your cells of interest and all the other cells? Some cell types are known to yield lower nFeature values like mouse neutrophils so it's good to be mindful of biologically-relevant differences such as this.