4.3 years ago by
You can't do both. Algebraically things will just cancel out. So if you normalize by cpm and then normalize by housekeeping genes, the library sizes cancel out (because the housekeeping genes are normalized by cpm as well), and it is the same as if you just normalized by the housekeeping genes.
In my experience (here 'experience' means working with NanoString data, using CodeSets that varied from ~200 to ~800 genes), once you get past about 400-500 genes, normalizing to library size is a better way to go (where 'better' is defined by lower variance of technical replicates).