Question

Batch correction and DESeq2

0

Entering edit mode

Alessia • 0

@alessia-24912

Last seen 3.9 years ago

Hi all,

I would like to use DESeq2 for my DE analysis. I have a quite big dataset of about 200 samples, divided into 16 treatments. For each treatment, I have 3 biological replicates and 4 time points.

Because of space issues, we decided to split our samples over 3 different blocks. In each block, we had 1 biol rep for all treatment and all time points.

After data exploration, It is evident we need to perform a batch correction, and for this reason, I would not be able to use raw data as input for DESeq2.

Do you have any suggestions on packages that will still allow me to use it or if there is a way to correct my data within DESeq2?

Thank you very much.

DESeq2 • 1.8k views

ADD COMMENT • link 4.0 years ago Alessia • 0

score 1 · Answer 1 · 2021-03-02

1

Entering edit mode

swbarnes2 ★ 1.4k

@swbarnes2-14086

Last seen 1 day ago

San Diego

It's actually easier than you think. You do give DESeq your raw counts, and include batch as an element in your design. Then call your contrasts as normal. You do not change the counts, you let DESeq model batch effect as part of its calculations.

ADD COMMENT • link 4.0 years ago swbarnes2 ★ 1.4k

0

Entering edit mode

Group  Batch
A   B1
A   B2
B   B1
B   B2
C   B1
C   B2

In this little example it would simply be designed as ~Batch + Group as also suggested in the Quick Start section of the vignette. You can use something like limma::removeBatchEffect to explicitely remove the batch effect from the vst-transformed counts and then visualize the batch correction via PCA as a sort of "quality control" on what the effect of the batch correction is going to be. There are plenty of threads available on the exact same topic, please use the search function for it.

ADD REPLY • link 4.0 years ago ATpoint ★ 4.7k

score 0 · Answer 2 · 2021-03-04

0

Entering edit mode

Alessia • 0

@alessia-24912

Last seen 3.9 years ago

Thank you very much for your suggestions. I will have a look at the option you pointed out.

ADD COMMENT • link 4.0 years ago Alessia • 0