I have RPK values (RNA seq) and I'm wondering what is a good rule of thumb for what is considered to be noise versus what is considered to be signal? I.e what should I choose as a threshold value for what is considered noise versus not noise? (just as a rough guideline)
Signal and noise are best determined at the count level, rather than after normalisation. I don't think there's an easy way to get a signal level from normalised data.
When I wanted to establish the noise level of counts from Illumina reads from lots of different samples, I looked at the total count distribution for genes that had counts in only one sample, and used the maximum value of that (or possibly an elbow of that distribution) for my signal threshold.