Woooo! Okay - so there have been some surprises in working with actual human cells that everyone is running into as their getting going. A lot of the things you know will work from your years of doing proteomics with basically unlimited material....don't seem to work quite as well.
Check out the image at the top of the post - doubling their gradient gets them what? a 10% increase? You doubling your acquisition time for 10% more peptides? Me either. You can look at the window optimization inside, it's cool stuff. Also -- gas phase fractionation (where you make libraries on an system by doing DDA on little narrow windows of your samples, for example only doing DDA [or, I guess, DIA] from 400-600 m/z, then running that sample again and only doing 600-800 m/z, etc.,) has a lot of value.
The coolest part here is the biology as they track the proteome of this cellular population as their differentiating. We're getting out of the methods development side right now and it's time to start rewriting some text books (or whatever they use now).