Detection of expression of a gene in rna-seq

Question

[Alex Voronin]

How can I say that I gene is expressed or not using RNA-seq data?

We know here may be background expression. So I think even if a few reads are mapped to the region a gene corresponds to it is not easy to claim the gene is really expressed. Is there a statistical method or an empirical way to do that?

[Ryan Dale]

There's a similar question over at To What Degree Could We Determine Whether A Gene Is Expressed? with some discussion about conventions and empirical thresholds.

Answer 1

[Sean Davis]

A few reads mapped to a gene in an experiment that otherwise "worked" (not DNA contamination) is often going to represent expression. The opposite (no reads) does not imply lack of expression, though. To answer your question directly, I don't know of an ideal systematic way to call a gene expressed or not expressed in RNA-seq. As usual, the best way to think of expression data is in the setting of differential relative expression, not absolute expression.