Wednesday, December 4, 2019

Revisiting BoxCar!

Now that a large number of little Orbitraps appear to be showing up all over that essentially have a "BoxCar button" (it's a premade workflow) I think it's worth talking about it again.

What's BoxCar? 

It's a method to improve Match Between Runs (MBR) using MS1 library matching. It improves the S/N of your MS1, but it does nothing to improve your MS/MS scans (they're mostly just there for chromatographic alignment stuff for MaxQuant)

Will BoxCar help my results? 

Yes, but only if you are using MBR and are comfortable reporting IDs using MS1 library matches

How do I process BoxCar data? 

Your dd-MS2 events will trigger normally and those peptides will be ID'ed. However -- if you aren't using MaxQuant and MBR you will get substantially fewer identifications. If you don't have an awesome library and MaxQuant and MBG -- expect maybe 1/2 to 1/4 the number of identifications per file. With MaxQuant and MBR? On 2ug injections you can get numbers just as good as what is reported in the original paper. No joke. I've totally done it. 7-10k proteins in a single shot.

What happened to BoxFahrt? 

Oh...that...I'm dumb and forgot about making mutated versions of BoxCar.


Disclaimer -- you have to pick your application for this to be useful at all -- but there are applications where I swear this totally works (more results and applications coming soon!)  Conor and I still write like Appalachian 4th graders, but since this workflow is SUPER EASY to set up on the Exploris we thought it would be helpful to the world to push this one up the queue and out the door.

The concept is super easy, right? BoxCar does multiple sweet BoxCar scans but that doesn't help your MS/MS, cause you pick your MS/MS from the included MS1 scans. Your cycle time suuuuuucks.  In the original BoxCar paper there are 3 BoxCar MS1 and 1 regular full scan MS1. Even on an HF-X at 60k resolution you're at 128ms transient per scan. That's over 0.5 seconds per MS1. But the S/N on the MS1 is incredible. You can get clear isotope distributions on ions that you can't even see in the MS1s.

Can you get the 10,000 protein ID's in 100 minutes with MaxQuant and MBR on that data? Totally. (In my analysis of their data I get more like 8,800, but I think that's my FASTA database and that I've never attended a MaxQuant summer school). And that's amazing and I'd personally run this workflow all day. But what about the super low abundance stuff? Like when people come to you and say  "I cut the salivary glands out of this weird mosquito, please give me a proteome of the parasite that is living in it?"

The first thing I saw during my BoxCar obsession was that -- holy cow -- yes on this super low material, BoxCar's boost in S/N is amazing. I can see SO MANY PEAKS, but I don't have an MS1 library for this -- or lots of low abundance weird things.


....triggers the MS/MS off of the BoxCars....BOOM....

1) If you run it on something where you've got more than 20ng of material you're going to think I'm more dumb than you already do (this number may be a lot lower on the Exploris -- I think the vendor has been understating the sensitivity on this thing a little. The numbers I'm getting are ridiculous. So...maybe below 10ng of material? I don't know yet, but hope to find out soon!)

2) Your cycle time still suuuuucks. Unless you do this on a Fusion and acquire your MS/MS scans in the ion trap. Then it sucks less.

How do you set it up? Starting methods for Fusion/Lumos are included in the supplemental. I'll try to get one made up for the Eclipse -- I haven't been able to get my hands on one (if you have one and you'd like to see if I'm as strange in person as I appear on these pages [I sure hope not!] I would be happy to discuss a field trip! Unless your in Boston right now. Half meter of snow? Best of luck with that, yo.)

On the Exploris -- you open the BoxCar method and then you add a ddMS2 to each msx-TSIM. That's it. Since the Exploris already has the BoxCars worked out all you have to do is add your MS/MS events. When I get time I'll work on titrating sample downwards, but in limited runs at higher injections it looks like the MS1 signal intensity and cycle time are better than the HF-X and Lumos.

Should go without saying? If you had to do a magic MS1 to see your ion you aren't going to fragment it well with 14ms of fill time. You'll need to crank that up. One mosquito's salivary glands? 300ms MaxIT on the Fusion 1, so around 100ms on the Lumos. Also -- there is a lot of junk in the noise. I'd recommend lowering your quad isolation to the lowest safe limit. 2.2Da on Fusion 1? 1.4Da/Th on Lumos? Something like that.

(Methods and RAW files, of course, are all available. I'm getting great at PRIDE uploads finally and I'll get these up here for the real paper submission. I've got 2 computational proteomics workshops to teach and my 2019 commitments wrap up and I can budget more than 30 minutes per day for writing! This took 28???)

BoxCar ramble number 713 complete.

1 comment:

  1. Hey Ben. We got the API licence from Thermo to try and do Boxcar on our QE+. Any advice or just follow the BioRivX? Or are we just kidding ourselves?