Question: deseq2 correction batch effect without design
0
gravatar for mdidish
12 weeks ago by
mdidish0
mdidish0 wrote:

Hi

I would like to correct batch effect using deseq2, to analyze one hundred of RNA-Seq of tumors, without experimental design. 80 tumors were sequenced in 2018, and 20 in 2019; I can see a strong batch effect whent I plot PCA on reads count or on data after DESeq between 2018 and 2019 tumors.

Usualy I used:

coldata <- as.data.frame(rep(TRUE, each=100))
rownames(coldata)<- colnames(COUNT)
colnames(coldata)<- c("group")    
dds<-DESeqDataSetFromMatrix(COUNT, coldata,design=~1)

I read that DESeq can correct batch effect with this kind of command:

dds <- DESeqDataSet(COUNT, design = ~ batch + condition)

But in my case I have no condition, and I tried "design=~batch", but without effect. I can remove efficiently batch effect with ComBat, with very good result on PCA plot. But then I have negatives values in my matrix, which is a problem for further analysis.

Which solution can I try?

Thank you for any suggestion.

deseq2 batch effect • 236 views
ADD COMMENTlink modified 12 weeks ago by swbarnes2330 • written 12 weeks ago by mdidish0
Answer: deseq2 correction batch effect without design
0
gravatar for swbarnes2
12 weeks ago by
swbarnes2330
swbarnes2330 wrote:

DESeq does not correct the batch effect, it incorporates it into its linear model making. And you do that just like above, by adding batch to the design along with the other elements of the experiment.

ADD COMMENTlink written 12 weeks ago by swbarnes2330

Thank you for your response. But by adding batch to the design, I still have a strong batch effect when I plot PCA with DESeq matrix. So I can't use these matrix to clustering tumors.

When I remove batch effect with ComBat, I have very good results very good result on PCA plot, but I can't use ComBat matrix because negative values.

So what solution do you recommend?

ADD REPLYlink written 12 weeks ago by mdidish0

There’s a FAQ exactly covering your question here in the DESeq2 vignette.

ADD REPLYlink written 12 weeks ago by Michael Love25k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 338 users visited in the last hour