Question

Batch effect removal with Deseq2

1

Entering edit mode

lirongrossmann ▴ 80

@lirongrossmann-13938

Last seen 3.5 years ago

Hello everyone,

I was wondering if there is a way to remove batch effect from an expression matrix (raw counts) for downstream analysis (like pca, clustering, machine learning) using the Deseq2 package.

I know I can do it using limma with removeBatchEffect, but wasn’t sure if there’s a similar function in Deseq2.

Thanks!

deseq2 batch effect removebatcheffect() limma • 3.1k views

ADD COMMENT • link updated 6.4 years ago by Michael Love 41k • written 6.4 years ago by lirongrossmann ▴ 80

score 0 · Answer 1 · 2017-12-18

0

Entering edit mode

Michael Love 41k

@mikelove

Last seen 15 hours ago

United States

Yes, for PCA, visualization, you can run limma's removeBatchEffect on the assay of the transformed data. This remove linear shifts associated with batch variables. We recommend to use the limma function, see here:

C: Batch correction in DESeq2

ADD COMMENT • link 6.4 years ago Michael Love 41k

0

Entering edit mode

Thanks! Can I use the output for classification purposes as well or would you recommend a different strategies for accounting for confounders?

ADD REPLY • link 6.4 years ago lirongrossmann ▴ 80

0

Entering edit mode

Depends on application. For statistical testing, we recommend to include confounder variables when possible as part of model as opposed to removing batch. But then for many possible tests, you could use DESeq() on counts. Otherwise, it seems easy to remove batch for e.g. visualization and prediction tasks.

ADD REPLY • link 6.4 years ago Michael Love 41k