CPM or pesudo.counts, which is the normalized counts

Question

[lilingjoyo]

Hi everyone,

I'm using edgeR performing normalization and differential expression analysis. I want to use the normalized counts data to draw heatmap. But I'm counfused which one shoud I used as normalized data, cpm or pesudo.counts.

Some people say cpm is approriate. But in my data, I found two groups' cpm value are alomse same.

so, which one is the right one?

Answer 1

[h.mon]

The edgeR user guide, on section "2.7.7 Pseudo-counts" (page 13) states pseudo counts main use are for internal calculations. I would use cpm. See here, a comment on answers from this post.

To draw heatmaps, see comments on section "2.15 Clustering, heatmaps etc" (page 25), it gives some tips (and some words of caution) on how to proceed.

Answer 2

[shengxin1015]

pesudo.counts, if you check with cpm function in edgeR, using pseudo.counts you will get the similar cpm value by formula.