Question: Removing Batch effect on the raw counts or normalised counts
0
11 weeks ago by
chudar.chudar0 wrote:

Dear All,

I am a newbie to RNAseq data analysis. Recently I have RNA seq data which include 4 samples and each of the sample include triplicates. I used DESeq2 pipeline for analysing the data and in the PCA plot i could see that all rep1s from all 4 samples are clustered together and rep2 from all samples are clustered together and rep3 from all samples are clustered together(rather clustering to their respective samples). My friend who prepared the samples told that he processed rep1 samples, rep2 samples and rep3 samples in consecutive days.

I used removeBatcheffect using the following code on the vsd transformed data

   vsd <- vst(dds)
plotPCA(vsd, "batch")
assay(vsd) <- limma::removeBatchEffect(assay(vsd), vsd\$batch)
plotPCA(vsd, "batch")


and the PCA looked far better where the replicates clustered according to their respective samples. But is it possible to remove the batch effect from raw counts or normalised counts in DEseq2 or any other package. Any guidance would be really useful. Thanks in advance.

rnaseq deseq2 batch effect • 98 views
modified 11 weeks ago by Michael Love26k • written 11 weeks ago by chudar.chudar0
Answer: Removing Batch effect on the raw counts or normalised counts
0
11 weeks ago by
Michael Love26k
United States
Michael Love26k wrote:

For the purpose of differential expression, we do not remove the batch effect, but instead include the batch variable in the design, e.g. ~batch + condition.

(This example is given in the first code chunk of the DESeq2 vignette)

Hi Michael,

Thanks for your reply. I did include ~batch into my design to regress the batch effect. But I am wondering whether is it possible to have batch effect completely removed from raw counts or normalised counts. This is because I want to use batch effect removed raw or batch effect normalised counts in another package which cannot account for the batch effect. Kindly guide me.

"whether is it possible to have batch effect completely removed from raw counts or normalised counts"

No. DESeq2 doesn't do this.

Thanks for your reply. But is there any other tool apart from DESeq2 which can remove the batch effect from raw or normalised counts ? Kindly guide me

I've given you all the information I have. I provide support for the DESeq2 package only on this site and I don't have any further guidance for you.

Have you tried sva?