vst of DESeq2 for bulk UMI counts
1
0
Entering edit mode
sabine • 0
@98111de7
Last seen 4 days ago
Germany

Hello, Is it correct to use the variance stabilizing transformation from DESeq2 to normalize bulk UMI counts? I have a matrix samplesXgenes with UMI counts and i was wondering if DESeq2 can handle UMI counts too or only read counts.

This is what i'm simply doing:


dds <- DESeqDataSetFromMatrix(countData=as.matrix(umi_matrix), colData=targets, design=~subtype)
dds <- dds[rowSums(counts(dds))>=10,]
vsd <- varianceStabilizingTransformation(dds, blind=F)


Thanks in advance for any help!

DESeq2 normaliz Normalization • 72 views
1
Entering edit mode
@mikelove
Last seen 18 minutes ago
United States

Should be fine with UMI counts. Check meanSdPlot (see DESeq2 workflow) if you want to assess how the transformation did.

0
Entering edit mode

I'm working on two datasets, and while the meanSdPlot of the first one looks good, the second one has a different trend. Do you think it can be linked to the fact that they're UMI counts?

1
Entering edit mode

Both of these look pretty good to me. Compare to log(x+1)

mat <- matrix(rpois(1e6, 2^seq(0,8,length=1e4)), ncol=1e2)
meanSdPlot(log(mat+1))