TPM or TMM for gene expression correlation and distribution

Question

[confused but trying]

I am trying to see how the expression of two genes correlates across patient samples. To do this, I first used plotted the expression of the two genes and saw that they were positively correlated across samples, and that some samples had much higher expression of both genes, and other samples had much lower. However, since I know TPM values are affected by the global RNA composition, I was wondering if these differences in expression could just due to changes in RNA composition, and not changes in expression of those genes. So I was wondering if I should redo my plots using TMM values instead. I was wondering if this would be a good idea, or if there could be any problems with doing this.

Answer 1

[LChart]

I think this is a very good idea. The TPM calculation can vary; and a number of methods do things like estimating effective gene length with GC or other covariate corrections. If the genes are strongly correlated under both TMM and TPM, you can have peace of mind in a technically robust result. If they are discordant, then you need to start to think about which one should be considered a more faithful estimator of RNA concentration -- absolute or relative.