DESeq2 VST question

Question

[MarcHiggins]

Hello everyone,

I am using the DESeq2 command variance stabilizing transformation to normalise raw read counts across 4 RNA-Seq experiments. Once normalised I will use these VST counts in Matrix eQTL, however I am having an issue with VST.

After carrying out VST, some of my results are genes have higher VST values than others which had higher raw read counts, i.e. re-ranking has occurred for some samples. Would anyone be able to shed some light on why this might have happened?

Thanks.

Answer 1

[Kevin Blighe]

This is most likely due to the adjustment for library size, which occurs during normalisation. For example, the samples that originally had higher raw counts may just have been sequenced to a higher overall read depth - through their adjustment for library size in relation to all other samples during normalisation, it was probably then found that these higher raw counts did not actually equate to higher expression.

For recommended reading, I would look at:

• Differential expression analysis for sequence count data
• A comprehensive evaluation of normalization methods for Illumina high-throughput RNA sequencing data analysis.

[MarcHiggins]

Hi Kevin,

Many thanks for your reply. After contemplating this further myself I agree that this is the case.

Marc.