DESeq, elevated SD in lower count range even after VST

[pinki999]

Hi,

I am using DESeq for differential expression analysis. When I transformed the data for variance stabilization using varianceStabilizingTransformation from DESeq, I still see an elevated standard deviation in the lower count range. Does anyone have come across similar issue? I am using appox. 50 vs 50 replicates. Does large number of replicates creates such elevation?..



Thanks

[Richard Barker]

Can you share the code/script you used as it might make the issue a bit clearer?

[pinki999]

Hi,

I used the code according to the vigenette:

countTable = read.table("count_data.txt", header=TRUE, row.names=1)
condition = factor(c(rep("A", 44), rep("B", 21), rep("C", 55)))
cds = newCountDataSet( countTable, condition )
cds = estimateSizeFactors( cds )
cdsBlind = estimateDispersions( cds, method="blind" )
vsd = varianceStabilizingTransformation( cdsBlind )

I am also attaching the meanSdPlot. First plot is from log transformed data and the second is from VST.

thanks,

[Richard Barker]

Could you explain what the numbers and letters refer to in the following line?

condition = factor(c(rep("A", 44), rep("B", 21), rep("C", 55)))

I

[pinki999]

Hi,

A, B and C are the conditions and the numbers represent the replicates.

[Richard Barker]

Wow, thats a lot of replicates, so your initial table has 120 columns? And in the first row 44 of them just have A, 23 have B and 55 are C?
I've heard of an experiment with so many rep's before so i'm trying to imagine the table structure.