Should we apply DropletUtils on control case separately or on a common sce object.
1
0
Entering edit mode
@rohitsatyam102-24390
Last seen 23 hours ago
India

Hi Everyone!!

I recently started with Single Cell preprocessing step and I have cells in two batches. Day 16 (control vs Knockout) and Day 40 (control vs Knockout). I plan to use seurat after preprocessing.

My queries are as follows:

1. Should I separately apply dropletutils on each control case like

p16C_sce <- read10xCounts(p16C, col.names=TRUE) #dim: 9250 6794880
p16K_sce <- read10xCounts(p16K, col.names=TRUE) #dim: 9250 6794880
p40C_sce <- read10xCounts(p40C, col.names=TRUE) #dim: 9250 6794880
p40K_sce <- read10xCounts(p40K, col.names=TRUE) #dim: 9250 6794880


or should I combine like it has been done here? This is because, when I apply Dropletutils separately, I get sce object with different dimensions and I am not sure if that can be ingested by Seurat.

2. One of my Batch showed ~10% of barcodes with high mitochondrial gene expression, higher than the threshold both in Control and Knockout. Should I retain those cells too because we see similar pattern in both control and knockout?

DropletUtils • 93 views
3
Entering edit mode
Aaron Lun ★ 26k
@alun
Last seen 1 hour ago
The city by the bay

I don't know what you mean by applying DropletUtils, which is the package name rather than the name of any particular function.

If you mean to ask whether they should be loaded in separately and then cbinded, or loaded in together by read10xCounts(), then it doesn't matter; both will yield the same result, give or take some automatic colData filling.

If you mean to ask how to perform cell calling with emptyDrops(), then yes, the cell calling needs to be done separately on each 10X run. Different runs have different ambient profiles so it doesn't make sense to combine and call.

One of my Batch showed ~10% of barcodes with high mitochondrial gene expression, higher than the threshold both in Control and Knockout. Should I retain those cells too because we see similar pattern in both control and knockout?

The fact that it occurs in the control and knockout doesn't mean anything; in fact, given that it's batch-specific, the most obvious explanation is that the processing for that batch was rougher and yielded more damaged cells.

0
Entering edit mode

Yes, I was referring to the usage of the emptyDrops() function on the combined sce object where we read both batches together (control-case) using read10xCounts(). We observed that that rescued nearly ~3000 more cells in total (control+case) than when they were preprocessed separately. But now I know that's not the right way. Thanks for your prompt response sire!!