The false discovery rate corresponds to the expected proportion of items that will be falsely
detected as responsive. With a very large data set it is important to define a selection step based
on an FDR not only to reduce the number of items to be further processed, but also to remove
too noisy dose-response signals that may impair the quality of the results.
We recommend to set a value between 0.001 and 0.1 depending of the initial number of items. 
When this number is very high (more than several tens of thousands),
we recommend a FDR less than 0.05 (0.001 to 0.01) to increase the robustness of the results 
(Larras et al. 2018).

-- REFERENCES --

Larras F, Billoir E, Baillard V, Siberchicot A, Scholz S, Wubet T, Tarkka M,
Schmitt-Jansen M and Delignette-Muller ML (2018). DRomics: a turnkey tool to 
support the use of the dose-response framework for omics data in ecological 
risk assessment. Environmental science & technology.