FDR, as currently implemented, is probably broken. Short gradients, DIA, chimeric spectra, PTMs, emphasis on quant. Everyone talks about "how" to do FDR "right" but maybe we need to ask WHY we're doing it and if we should rethink the entire process based on the purpose.

What I am currently puzzled about: What if the same data should serve multiple purposes where different FDR filters might appropriate? Lets say residual protein level determination from KO screen plus differential expression analysis?

There's a bit of talk about why 1% FDR and maybe we shouldn't subscribe so religiously to 1%. A collectively cosigned letter making it clear that >1% FDR can be perfectly fine, depending on what's happening post-processing with that list.

🔥🔥🔥