Why are lowly expressed genes less likely to detect as DE?

Question

[statfa]

Hi,

In my study, I have realized that the model I use is not able to detect DE genes if genes are lowly expressed. What can be th reason that lowly expressed genes are less likely to be detected as DE even if they are truely DE?

I am not talking about increasing the power of analysis by removing lowly expressed genes as they may be affected by noises. I am talking about the situation in which my lowly expressed genes are important but my model cannot detect differences in expressions. why is that?

Thank you

Answer 1

[Devon Ryan]

The strength of fold change and variance shrinkage scales inversely with the size of the values (and the number of them). As a rule, power increases with sample number and the number of read counts per gene.

[statfa]

Thank you. Do you have any reference for this, please?

[statfa]

I found the reference here. Thank you