complex heatmap with CPM

Question

[biology_inform]

Hi all, To draw complex heatmap, I started to use ComplexHeatmap package in R. I have a RNASeq data which has 2 cell lines with 2 conditions (CellLine1-WT, CellLine1-KO, CellLine2-WT, CellLine2-KO). When I try to draw heatmap with log2(CPM+1) values, the h2052 doesnt cluster together but they should cluster together. Is there a specific way to do that?

[Mensur Dlakic]

Are you telling us that H2052_ATG_1 and 2 should be clustering with H2052_REN_1 and 2? Because my eyes are telling me that H2052_REN_1/2 are clustered exactly as they should be.

[Ming Tommy Tang]

true. Also the clustering of the samples can change depending on the genes you choose. I have a tutorial on it at

and deep understand heatmap at

Answer 1

[ATpoint]

You seem to use logcpm rather than scaled counts for the heatmap. This causes clustering by expression level rather than by differences in expression. See my answer here for details:

Scaling RNA-Seq data before clustering?