Question

Ploting of normalized counts after batch removal

0

Entering edit mode

altuda ▴ 10

@91f63ac0

Last seen 5 months ago

Czechia

Hello, I tried to search for a correct answer, but I don't think I've found one. I have an experiment with paired samples, I've performed DESeq2, and now I would like to plot DESeq2 normcounts. All the discussions and tutorials I found are regarding plotting heatmaps and PCA plots, where you remove the batch effect of the paired samples from vst. Now, how do you think I should proceed regarding the normcounts? Since I am introducing negative values, when I apply limma::removeBatchEffect it doesn't seem correct.

Thank you for your time.

DESeq2 • 477 views

ADD COMMENT • link updated 5 months ago by Gordon Smyth 52k • written 5 months ago by altuda ▴ 10

0

Entering edit mode

You can use ComBat-seq from the sva package to regress covariates such as the pairing information. Advantage is that it returns "raw corrected" counts that are still integers, so you can put them into vst again and plot that.

ADD REPLY • link 5 months ago ATpoint ★ 4.6k

score 0 · Answer 1 · 2024-06-02

0

Entering edit mode

Michael Love 43k

@mikelove

Last seen 3 days ago

United States

You can't really plot counts with variance from nuisance covariates removed. I would go the route of using VST data instead (see FAQ).

ADD COMMENT • link 5 months ago Michael Love 43k

score 0 · Answer 2 · 2024-06-03

0

Entering edit mode

Gordon Smyth 52k

@gordon-smyth

Last seen 1 minute ago

WEHI, Melbourne, Australia

I am introducing negative values, when I apply limma::removeBatchEffect it doesn't seem correct.

The values input to removeBatchEffect() are always on the log-scale. Getting negative log-scale values is no problem. The unlogged expression values that they correspond to are always positive.

ADD COMMENT • link 5 months ago Gordon Smyth 52k