Getting the normalized counts after running DESeq()
1
0
Entering edit mode
@33ce88fe
Last seen 2 days ago
United Kingdom

Hi! I am trying to get the normalized counts after having run DESeq() with my DESeqDataSet, in order to then use these normalized counts to run GSEA analysis using their app. The code below is what I have used, but I was wondering if this normalization is only done considering the transcript lengths or whether it also considers the experimental design (~line + treatment in this case).

#First run Tximeta and summarize the results to gene level
se <- tximeta(coldata)
gse <- summarizeToGene(se)
#create the DESeqDataSet
ddsTxi <- DESeqDataSet(gse, design = ~ line + treatment)
#Filter out the genes with low counts
dds <- ddsTxi[rowSums(counts(ddsTxi)) >= 20]
#Run DESeq()
dds <- DESeq(dds)
#Get the normalized counts
norm_counts <- counts(dds, normalized=TRUE)


Is this correct? Thank you for your help!

DESeq2 Normalization • 69 views
0
Entering edit mode

The size factors are independent of the experimental design. The only "normalization" method that takes a design is vst and rlog, see the vignette.

0
Entering edit mode
Last seen 13 days ago
Singapore

counts(dds, normalized=TRUE) returns raw count matrix normalized by size factors. You can check the normalized library size with colSums(norm_counts).

AFAIK, GSEA requires a sorted feature list instead of normalized counts as input.

What do you mean by 'GSEA analysis' here? Do you want to say sample-wise 'GSVA analysis'?

If it is the case, please refer to the kcdf parameter in GSVA manual for optimal data format.