Association between number of exons and read density

Question

[lkmklsmn]

Hi,

I am analyzing human cell line RNAseq data. I have generated gene level read densities by dividing total read counts by gene length.

Now I have an extremely strong correlation (P<1e-100) between this read density and the number of exons contained in a gene. Something seems wrong here. Is this normal? Any ideas where or how I could have messed up the normalization? Has anybody recognized this relationship before?

Answer 1

[mikhail.shugay]

Hello! RNA-Seq should normally contain a relatively small fraction of intronic reads. On the other hand introns are far longer then exons, so it is the major factor that determines gene length. So one should normalize by CDS length, not the total length of gene.