I'm trying to cluster rows and columns in DESeq2 and show it in a heatmap. I've tried this:
heatmap.2(assay(vst)[select,], col = hmcol,
Rowv = T, Colv = T, scale="row", trace="none", margin=c(8, 8))
I think my problem is revealed when the scale = 'row', the genes don't seem clustered, yet when scale='none' I see high medium and lowly expressed genes in separate clusters. I think this means that the rows are not normalized, while clustering columns, each column is normalized, right?
I think, when the rows aren't normalized the counts that are low tend to cluster to each other and high counts would cluster each other. For example:
geneW, Sample A = 10, B=20
geneX, Sample A = 15, B=15
geneY, Sample A = 1000,B=2000
GeneZ, Sample A=1500,B15000
if each gene is normalized, W and X are closer to Y and Z. But in realized W changes more similar with Y; and W remains constant like with Z
Is my interpretation correct? How would I normalize the rows?