I would like to do a unsupervised heatmap using deseq2 counts.

If I use :

topVarGenes <- order(rowVars(assay(rld)),decreasing=TRUE)[1:1000]
mat <- assay(rld)[ topVarGenes, ]
mat <- mat - rowMeans(mat)
pheatmap(mat, annotation_col=df)

Which is the right way to do? Using scale="row" or scale="none"?

We recommend scale="none" in the workflow.

Here is a previous post where I give some rationale for this decision:

A: Biclustering Normalizing by Row in Heatmap of DESeq2

Thanks ! I read the that but I have this situation:

topVarGenes <- order(-rowVars(assay(rld)))[0:1000]

mat <- assay(rld)[ topVarGenes, ]
mat<- mat - rowMeans(mat)
colnames(mat) <- paste0(metadata$CONDITION,"-",metadata$ID)

#draw heatmap
nome<- "Image_analisi/Unsupervised_1000highvariation.png"
pheatmap(mat,method="complete",main = "Unsupervise 1000 genes ", show_rownames = F, color=color1,annotation_legend = TRUE, legend=T, cluster_cols=TRUE,file=nome)

If I use scale none or scale row  I have a radical change of the cluster of my groups :

Here you have the images: http://imgur.com/a/uxxx8 The first one is the use scale=none e scale row. In the first case seem I have all my data is down-regulate...