vsd values vs normalized counts in DESeq2

Question

[adR]

Hi all!,

I know my question is a bit silly but wondering what major difference I might encounter if I use the VSD values of DESeq2 instead of the normalized counts(counts(dds, normalized = TRUE) for my heatmap? Best, Amare

Answer 1

[Kevin Blighe]

Normalised 'counts' will be positive only, and will follow a negative binomial distribution.

Variance stabilised expression levels will follow a distribution more approaching normality - think logged data.

For heatmaps, you then need to take consideration of:

• distance metric used
• colour scale
• breaks for colour scale

Kevin