What is the recommended method for specifying enriched or depleted features?

[jolespin]

I'm following the tutorial here: https://birdman.readthedocs.io/en/stable/default_model_example.html

With other software like ALDEx2 and edgeR you can specify a cutoff using FDR. I understand that working in a bayesian framework is different so I have a few questions:

• What is the recommended way for automating the feature sets using differentials?
• Is there an analog to an FDR cutoff that could be used that is generalizable to every run?
• Does this require manual curation on a case-by-case basis to determine which differentials are statistically enriched or depleted?