Thursday, May 10, 2018

BOXCAR -- It's time to stop and redesign most proteomics experiments!!

Okay. I'm going to try not to freak out. This is so simple and brilliant that every one of us sitting in front of an Orbitrap should have come up with it on our own.

We've spent all our time focused on how to get more MS/MS signal that we've ignored the fact that our 120,000 resolution;  MS1 scan (that might take 250-500ms to actually do) only had ONE millisecond of fill time.

What happens if we stop being dumb and use all that wasted time for something? 

10,000 proteins in a single injection?  Hmmm....that wouldn't be too bad. That's, I dunno, 5 to 10 times what I'm normally getting per injection.  WHAT!?!?!?!?!?!?!?

That's what happens when you unbias your MS1 acquisition and distribute it across the mass range. The authors use the quadrupole to isolate a series of little windows across the mass range. The next MS1 scan will have a different set of windows. All the sudden that one single albumin peptide isn't consuming the full 3e6 charges that the C-trap can take. It wasn't isolated in that analysis, and in the one it did it was equally distributed between the other boxcar windows. It CANT waste all of your AGC!

What they get? An average of 20-60x higher signal to noise! Which kind of results in more peptide ID's per unit time than anything I've ever seen.

If you're thinking -- wait -- they just butchered the MS1 scan by cutting it into little sections -- you're right. However, the authors are somewhat involved in the MaxQuant project and have written some clever code and optimized the isolation windows to make the quan reproducible and reliable.  This might mean, however, that if you aren't using MaxQuant you're going to need to tweak your LFQ software to make this work. I'll look at the ones I use and let y'all know if I can make it work.

In a human clinical study they quantify around 6,200 proteins with no missing values across 10 samples. That's -- impressive.... On a less limited sample (mouse brains) they crack 10,000. Single shot. 10,000 proteins.

Man -- this paper is awesome, logical and I still feel really dumb -- but I'd feel a lot dumber if I wasn't using this method for some projects next week when I'm back in the lab....

EDIT: Thanks for the comments guys -- I'm out of the office till Monday. I'll be working on standardizing BOXCAR methods so that they can be distributed through   I hope to have something so people not on this great new study can be running next week.


  1. This needs a special acesss to Xcalibur software. Do you know how we can get access to “text method interface”?

    1. You can get a free permission from thermo if you request them or wait for their MaxQuant Live from Cox Lab

  2. I was super excited to try this method today, but the "text method interface" that you need to use to create the method is not widely available. Thermo customer support didn't know what it was and said they'd have to check with the factory.

  3. Dear all,

    Thank you very much for your excitement about the BoxCar acquisition method. We are fixing some last bugs, but we are about to release an Xcalibur plug-in that will enable BoxCar scans. Please follow for updates and download details once available. Data analysis is integrated in the MaxQuant software environment for quantitative proteomics ( as detailed in the publication.


  4. I was wondering if the XCalibur Plug-in is available yet to be used for enabling BoxCar scans.

  5. Anyone heard any news of this as of late? Thanks!

  6. I just attended the MaxQuant Summer School and they said that the BoxCar acquisition mode can be run through a new software called MaxQuant Live. It can soon be downloaded from

    I think the method is promising indeed. But as I understand, after rushing through the paper, they are reducing the number of MS/MS scans as a consequence of spending more time for boxcar MS1 scans. To not lose too many identifications in this reduction of MS/MS scans, they are transferring identifications from a peptide library.

    I wonder, how many identifications would they get if they did not use a peptide library? Would it still outperform traditional DDA top 15 in identifications?

  7. Update: as of yesterday is... well, just that, live.

  8. It sounds great! I am excited to try it.