Ambient RNA removal method that generates whole (integer) counts
1
0
Entering edit mode
12 weeks ago
EK • 0

I am envisioning a project that requires removal of ambient RNA counts from cell-containing droplets. On the other hand, I will also need to aggregate counts from individual mice into "pseudobulk" profiles, to use as input to between-condition* DE analysis in DESeq2. (*: between-condition DE as opposed to DE for detecting celltypes—for the latter I simply use a method such as FindMarkers in Seurat)

I have used SoupX in the past for the ambient RNA removal, but I cannot use it with DESeq downstream because DESeq will accept a counts matrix with integer values only.

I am wondering if there are any ambient RNA-correction methods that produce integer counts. If anyone has familiarity with one or more of the following ambient-RNA correction methods, and could tell me which, if any, manage to remove whole counts (entire UMIs), I would be very appreciative. These are some of the methods I am checking one by one. Other methods I'm not thinking of would be good too!

DropletUtils, removeAmbience()

Cellbender

DecontX

Alternatively... would anyone with a statistical background be able to chime in on whether rounding the SoupX-generated counts matrix into whole numbers could be an acceptable input for the purpose of DESeq? Cells are being pseudobulked; i.e., all the cells that came from one sample—as identified by antibody-based hashtagging—will be summed together to generate one counts matrix for that sample. Which I think might remove any unique concerns that have to do with zero-counts in individual cells. But I am not confident about that.

Thank you in advance!

pseudobulk soupX DESeq2 single-cell • 476 views
ADD COMMENT
4
Entering edit mode
11 weeks ago

I've definitely just rounded soupX output and had no issues with pseudobulking.

A colleague speaks highly of Cellbender, though it takes much longer to run and really performs best when run on a GPU. It provides additional metrics though, and can be useful for cell calling.

ADD COMMENT
0
Entering edit mode

Thanks, Jared. Sounds like rounding counts is the simplest approach for now. :)

ADD REPLY

Login before adding your answer.

Traffic: 2002 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6