When is a gene considered unexpressed?

[lukas1848]

Hi all,

I just have a quick question concerning Illumina RNAseq. If you have a gene that gets about 1 read in 7 biological replicates, would you consider this gene to be expressed or not expressed? What is the minimal number of reads (or normalized expression) that you would consider it expressed? Are there any general rules I can orientate myself on?

Thanks,

Lukas

[rskr]

Hi all,

I just have a quick question concerning Illumina RNAseq. If you have a gene that gets about 1 read in 7 biological replicates, would you consider this gene to be expressed or not expressed? What is the minimal number of reads (or normalized expression) that you would consider it expressed? Are there any general rules I can orientate myself on?

Thanks,

Lukas

The way I think about it is how did the read map. If you have a 54bp read that maps unique without any mismatches and has a fair amount of complexity, then I would say there is enough evidence to conclude that that RNA was expressed in the sample, because the probability of just randomly getting that sequence is very low. Now how well you can quatitate the concentration of a single read gene may be suspect, however if you have multiple maybe even dozens of high quality unique mappers in one sample and none in another, you might start to think that it wasn't just random chance, that indeed there was a significant difference between the two samples...

[lukas1848]

Your argument is indeed convincing! In my experiment, the treatment has ~1000 genes very weakly expressed which are not expressed at all in my control (i.e. virtually no reads in any sample). Since I have so many genes with such a pattern, I would argue likewise that the probability that these all are just random hits is low.

Thanks for you input, rskr.