What is the ideal FDR and fold change cutoff to consider in a RNA-seq experiment and which one should be the priority for selecting the DEGs?

Question

[aishu.jp]

I have a control vs treated time course RNA-seq plant data for which I am trying to construct gene coexpression network.I have used DESeq2 package to identify differential expression of genes.

My aim is to identify genes that are exclusively differentially expressed in the time course experiment.

What is the safe FDR and fold change values to consider in a RNA-seq experiment ?

Is filtering the DEGs using FDR cutoff >0.05 is alone sufficient to get significant DEGs?

In FDR and fold change, which one should be given the priority for selecting the DEGs ? or should one consider both in order to filter significant DEGs.

[Joe]

What is the safe FDR and fold change values to consider in a RNA-seq experiment ?

There is no such thing. You need to determine this empirically for your data, such that you can tell a coherent biological story.

A fold change of 0.1 still (loosely speaking), means there is 10% more of that RNA (and therefore potentially protein). This could be a profound increase for one phenotype, and next to nothing for another.

Start with some arbitrary thresholds, such as FDR <= 0.05 and Fold Change >= 1, see what the data looks like, and iterate on the process until a story starts to emerge that makes sense.

[corend]

You should definitely use a threshold on both metrics.

Depends on how many DEGs you want to retrieve, take a look at the litterature, which thresholds are used in other papers. Then look how many DEGs you have with those thresholds and modify if necessary.

Also you should consider for the Fold change what is biologically significant for you. A gene 1.1 times more expressed is DE ? Or 2 times ? or 4 times ?

See also: Recommended cutoff for FDR - 0.05 or 0.1

What Is A Good Threshold For Log2 Fold Change?