Wednesday, December 12, 2018

OpenTrons OT-2 in the house!

YEAAAAAAAH!  Finally here! More details will certainly follow as we (well... @jenkins_conor) writes the software to make this thing:

1) Do BCA assays (whatever those are)
2) Digest things
3) Phosphopeptide enrich things
4) Extract metabolites
5) Dilute extracted metabolites

So far I'm impressed with the thought that went into the setup, the quality of the materials, and Conor said the developer software for it exceeds expectations. The fact it was 1/10 the price of any other one I wanted didn't hurt either...

You don't get a PC with it, but it connects via WiFi and you may need to have some Python guru hanging around to make it do new things it doesn't know how to do. Of course, method development and validation will take some time, but the manufacturer hosts methods developed and posted by other labs. We'll obviously upload ours as we develop them, if the robot is any good.

We're currently using an older Chromebook that is now running Linux, but the thought of getting a Linux tablet for it once we're confident of the performance and reproducibility and sticking it to it with a magnet is seeming like a solid idea.

Tuesday, December 11, 2018

Nonhuman proteins in human milk?!?!??

(Image is from @Confused_Cow, a very strange Twitter account I am now definitely following)

What I actually want to put here is this confusing new JPR paper.... the biological implications that sound like this would be completely and totally biologically impossible (all the proteins we eat get chopped up by that stomach acid stuff, right?), let's look at the amount of work that went into this ---

6 humans donated milk.
This milk was separated into 6 fractions by SDS-PAGE
These SDS-PAGE fractions were analyzed using a passive split flow Agilent nanoLC system in what looks like around 45 minute gradients on a Q Exactive Plus system
MS1 resolution of 35,000
Top10 experiment
MS/MS resolution of 17,500 with a 120ms fill time.

The data processing was done this way --

--I've highlighted the things that I find interesting here, and that I would have done differently.

Wooooooo.....okay.... In one sentence -- This is a really big search matrix searched against masses obtained at medium level resolution, using search tolerances that are strikingly high and searched against a software package version that has been lampooned for it's inability to accurately quantify FDR when matrices get really complex. Yo. If you're Q Exactive Plus is off by 48ppm, get a fire extinguisher. No QE+ that isn't currently on fire or under water is off by that much.

EDIT/Clarification: Proteome Discoverer 1.4 only controls FDR at the PSM level. All later versions of PD were designed to control FDR at multiple levels in order to more accurately control data quality with extremely large numbers of spectra or with high complexity search space.

Sorry, but this seems like an invitation to a list of proteins that are gonna be bad.


If you do your searching of your global and then validate the results with Parallel Reaction Monitoring (PRM) -- and it works -- maybe I should shut up.

Particularly, I guess, if you make heavy isotopic standards as well for your PRMs and this stuff looks legit. And if you go to the supplementary info -- it looks on point --

-I should probably definitely shut up.

In the end -- I'm still confused. WHY ARE THERE DOG PEPTIDES IN HUMAN MILK?!?!?

But the work looks solid.....

Monday, December 10, 2018

Assessing automatic sample prep technologies!

Was this already on here? I forget. And does it really matter anyway? This is cool enough to talk about more than once!

Honestly, in general, all these robots are freaking me out a little. It seems like every time I go to a store or leave a parking lot, there is one more robot thing and one less person working. And it's weird.

However -- in science, where reproducibility is kind of insanely of completely paramount somewhat importance -- automation is amazing!

Max Planck is using the Agilent thing.
This new study used a bunch of things -- and the reproducibility is kind of insanely completely awesome looking -- even on samples that are prepped on separate days!

Sunday, December 9, 2018

OMNITRAP -- Upgrade your QE to do EVERYTHING!!


Apparently, I've been under a rock for a while. Or hiding from my...

...few unanswered emails.... Don't worry. It is mostly because I bought a compost bin from Amazon in 2015 and now obviously I'm a compost bin collector now.

I'm driving home from horrendous holiday shopping craziness and my wife is flipping through the new Science and less cool Scientist and reads off something about Roman Zubarev (Ben is paying attention) and then reads off details about the "omnitrap" (Ben is having trouble keeping a hybrid in it's lane) and then "that it's an upgrade for any Q Exactive! (Ben requires moving from driving to being the passenger).

You add the OmniTrap to the back end of your Q Exactive.

It costs you about $250-$300k

It gives that Q Exactive some new capabilities. Like --

I don't even know what all those things are!  I just know that I don't have them. And I sure would like to.....

Saturday, December 8, 2018

Split resolution control on the QE HF Turbo!

Okay -- am I an idiot? Sure. Yes. Totally. But after a night of broken Orbi XL and-still-not-working after the rainstorm Orbi Elite AND the HF slEasyNano on the fritz again -- maybe being able to FINALLY GET AN MS1 RESOLUTION BETWEEN 60,000 and 120,000 is cause for celebration!


Mark Ivanov and
Elizaveta Solovyva

of the great Gorshov lab for taking a look at my data, finding dumb mistakes that I personally made and showing me awesome new tools I can't talk about quite yet.

To be clear, they weren't trying to help me replicate Jenny's phone number on my Orbitrap. They were being really brilliant and helpful and, as a consequence, I did something stupid while verifying I finally got the stupid nanoLC running again. But -- nowI can run my samples at a resolution right in the goldylocks zone. 60,000 just isn't enough for digging deep and 120,000 takes too long.

86,753 MS1 resolutoin
15,009 MS/MS. YEAH!


Friday, December 7, 2018

MaxQuant.LIVE Workshop in Zurich next week!

I know the first thing you probably thought when you read that subject line. Same thing I thought --"what's a Zurich". I'm happy to tell you that it's a city. And either something is very wrong with my Google search preferences OR the most important image to tell Americans about Zurich is that Rihanna has been there.

Now there is a second reason to go wherever that is!

Is this the world's first ever MaxQuant.Live workshop? I think it might be!

I love ProteomeXchange. Don't download this amazing dataset yet!

Okay --- so how ridiculously cool does this sound...?

Did I download this dataset? 

Will I consider sending grumpy letters to a bunch of journals if I'm not able to read about this ridiculously cool study STAT?

Are these questions rhetorical?

Oh -- and on that topic. This is one really interesting page of text from John Yates, III. that you should check out if you haven't seen it already. 

I think it spun off a lot of Twitter conversations that I need to catch up on from my metabolomics hiatus.

Thursday, December 6, 2018

Plasma profiling -- Real patients -- post gastric bypass surgery!!

I'm going to geek out over every detail of this one later. Unfortunately, in a huge rush right now.  Honestly, I may spare you from it. I'll save it for trying to convince people in my geographic area that mass spectrometry is more than just an experimental toy.

Punch lines?

175 patient PLASMA proteomes

>1,000 proteins when each was ran in triplicate (no depletion)

This goes down as THE largest quantitative dataset on non depleted real patient plasma (prove to me that SomaScan is quantitative and I'll consider changing my mind on this statement -- and comparing your array to GWAS doesn't count, yo.)

45 minute runs on an HF

Automated the whole sample prep thing with a robot. Heeeellllooooo fully functional clinic-ready assay!

Umm...not what I was looking for...but too funny to leave out.

Wednesday, December 5, 2018

A comprehensive pipeline for translational top down proteomics!??!?

Okay -- so -- I can't read this yet. But I can see the title and it sounds impossible.

"Translational" is the word we use when we mean "WAIT. WHAT?!? WHAT I'M DOING RIGHT NOW  isn't just a curiosity -- this could actually help someone!!", right?

"Translational top-down proteomics" immediately made me think of this.

But....we've seen some stuff recently where the practicality of applying top-down to serious far-reaching problems has actually seemed realistic. And -- I've gotten close to 100 digested samples across my Fusion in the last 3-4 months on this protein Ntai et al., conquered unequivocally with top-down -- and I think digesting makes it almost impossible to work with. Take a screen capture before I edit this: I think one of the most important proteins in cancer can't be studied effectively with shotgun proteomics.

It still seems like a long shot for helping patients. Most hospitals still haven't embraced the power of them new-fangled triple quad things. But maybe not. And we won't know till we try, right?

Tuesday, December 4, 2018

MS/MS visualization in Python!

I do love this python mass spectrometry community. Keep it coming, y'all! This is simple, the output is nice, the code is open...everything you could want...

Friday, November 30, 2018

The ULTIMATE Proteome Crosslinking Protocol!


EVERY step of the process.

A bunch of stuff I hadn't even thought about, like cell lysis, reaction conditions, offline fractionation (including what fractions to keep and which to toss)


Is it my data processing pipeline? Nope! Cause now I have files where Proteome Wide Crosslinking actually worked!!

Thursday, November 29, 2018

Last QE Turbo post, for real!

This might be the last one. (Continuation of this post, possibly yesterday? I forget)

What if the fill times were exactly the same on a QE HF running the same method and the same fill times (arbitrarily 30ms, cause why not?) and all I changed was the MS/MS resolution? 15k vs 8k?

You'd expect to get more scans when using 8k MS/MS, right? Because you'd definitely get more scans completed when you didn't need the full fill time.  More scans = more peptides!

I had a small opening overnight so I put on 500ng of HeLa and ran it on a 45 minute gradient 4 times.
2x with 60k res MS1/15k res MS/MS with top 30 and 30ms max fill time
2x with same except 8k res MS/MS

With 15k MS/MS -- 46,900 and 46,978 MS/MS scans
With 8k MS/MS -- 54,012 and 53,606 MS/MS scans

14% more MS/MS scans! WOOHOO!!

Then it gets less cool.

Clearly this is pseudo-pscientific, but it doesn't appear that my HF benefits from running MS/MS at 8,000 resolution. Honestly -- this is about what I'm seeing on the Fusion 1, to the point I don't use a transient below 15,000 resolution. I would like to take this apart. Like -- where are those 7,000 extra MS/MS scans going? Does mass accuracy appear to suffer due to peak coalescence at the lower resolution? That seems the most likely explanation, but no time to investigate right now.

This is good for me, probably. There was this increasing temptation to use MaxQuant.Live to "hack" my instrument every time there was a gap in the queue, but with this kinda settled, it's time to get back to the metabolomics!!

Wednesday, November 28, 2018

Proteogenomics fills in not-so-subtle blanks in the B.subtilis proteome!

I love these studies where someone tackles a model organism! Bacillus subtilis? Really? We have to know everything about this one, right!?!? Just what I can remember from long long ago grad school...

It has a genome about the same size as E.coli -- but
It can form nearly indestructible spores, hang out for 20,000 years, and start growing again.
It can make super weird fruiting bodies.
You can knock out every one of it's penicillin binding proteins (required for it to make cell walls) and it will still grow!

Despite hundreds? Thousands? Of studies on this little organism, loads of mysteries still exist? Such as...only 50% of the proteome is accounted for...???

Ravikumar et al., from the Macek lab is on the scene, solving these mysteries in this new Nature Scientific Report!

How do you improve on the understanding of something expected to be this well understood?

You grow normal and knockout bacteria in normal media(s) and in SILAC (&Super-SILAC) ones.
You fire up the Orbi-XL and you

Like --
1) In gel digestion
2) In solution digestion
3) You phospho enrich (4 WAYS)
4) You acetyl-enrich
5) You employ the appropriate multi-enzyme approaches

You process 1,688 (!!!!!!) files through MaxQuant.
You use the canonical databases.
You supplement these with 6 frame translated genomes.
You use Motif-X and other tools to assemble everything
..and finally(!) you do this:
...tried to spell that like 12 times. Screenshot!

What do you get out of this monumental amount of work?

1) By far, the most complete proteome of this important organism ever done.
2) A really intimidating folder on ProteomeXchange/PRIDE (here)  For real, someone deserves an award for just uploading this many files.
3) An amazingly thorough understanding of how this organism tightly regulates itself through a complex combination of acetylation and phosphorylation events.
4) And some useful phylostrafibrastabhic data?
5) B.subtilis finally

(...come on...I had to use that somewhere....)

Tuesday, November 27, 2018

Metaproteomics of medieval teeth tells us about the health of ancient people!!!

Okay -- if the title of this doesn't make you want to read this (open access!) paper we probably can't be friends....

This is like Emily and Antonius's mummy study if you weren't quite as sample limited and you moved forward from an Orbitrap XL to today's fastest Orbitraps!!

They can ID the bacteria that these ancient people had been exposed and infected with -- from their teeth!!

An Encino Man gif is mandatory here, right?

(thought I could find a better one...)

MWAHAHAHAHAAA!!!! Much better one! 

.....told you it was raining a lot in Maryland....

...turns out it was a failed condenser thingamajing.....

Fingers crossed it turns out okay. This Elite is a total rockstar. We've got a bunch of newer stuff around here, but this one is still getting slammed 24/7 for good reason. Hopefully an intense shower of unknown duration isn't the end of the tour!

Monday, November 26, 2018

Clinical mass spectrometry is more than pain management.

I recently went to demo some mass specs. Boring triple quad routine analysis stuff.

I realized on the second day that every time I said the word "clinical" all the people there heard was "pain management" (is this a US exclusive thing?)

Inspired, I just wanted to take a look at clinical mass spec today and some of the proof that we can do more than just quantify oxycodone in plasma.

<begin rant>

Great new study #1

Great new study #2 (You get HDL, LDL measurements in the clinic all the time -- but this is just a tiny clue to the puzzle of what is wrong. We have the tools to flesh out these differences NOW.)

Great newish study #3 (We don't even need anything special. The tubes that the hospitals are using for collecting samples for DNA analysis? Those are perfect for clinical proteomics analysis!!)

Similar and great newish study #4 (Look, we don't need anything special, Mr. Hospital administrator. We're not asking you to spend an extra $0.06 per patient on a new fancy tube. You're already making more FFPE slides than you will use. We just want one of them....

Great newish study #5 (Personalized medicine for bladder cancer. We can track biomarkers to determine what chemotherapeutic to use and when. Today. With the technology we have now. "personalized medicine" is more than just words for politicians to throw around. This is stuff we can do right now to make patient's lives better and improve the chance they go home.)

Okay -- I should wrap this up. And I'm going to end it with the study that is the first image. And -- yes -- this study has appeared on this blog multiple times, with me focusing on different aspects of it, but --- what a fucking awesome piece of work. I'm thrilled that if you Google image search "clinical proteomics" that is now the second picture that pops up.



<end rant>

Sunday, November 25, 2018

What do you get when you run an HFat full X transient speed?

If you can't tell, that's a picture of the racing stripes I'm considering requesting a local auto-body place cut and install on my HF. Now that I'm looking at it, it seems a little dumb. It looks more like suspenders.

Much better!

Okay -- so you should probably check this out on your own, but in case you aren't crazy and want to see how it turns out -- these are my results so far. (Oh. And this is an extension of this post a few weeks ago)

Some results:

500ng of human cell digest at 60k MS1 15k MS/MS (normal HF speed) -- 45 minute gradient;

Okay -- pretty normal.

What if I change nothing, but drop the MS/MS to 8,000 resolution (and the fill time down to 11ms to make sure transient is the limiting factor)?

Loads more MS/MS scans.

Less of everything else. Let's look at why:

Even with the normal method, we're maxing out our fill time the majority (66.85%) of the time.

Cutting that number in more than half? That's totally dumb. Makes me sound as dumb as all these "peer reviewers" keep implying that I am, right?

"Hey Ben, why did you choose 11ms?" Because I'm dumb. Or because I wanted to verify that I really was increasing the number of MS/MS scans and boosting the number of ions sampled per cycle.  And -- I'm increasing the number of MS/MS scans and the number of ions sampled per cycle.

And I wanted to see if I could get anything with 11ms of fill time. Almost 1900 protein groups in 45 minutes with a massive fill time limitation? That's AWESOME. Less than what I get with a normal method an nLC -- but, in case you've never visited this blog before, I HATE nanoLC. Necessary evil and all that, but I'll have a huge party when it's a dead technology we laugh about suffering through.

We've been evaluating options to nanoLC. And I've got a solution that shows some promise, but the peaks are waaaaaay too sharp. Like 4 seconds wide sharp. At the top of a 4 second peak, there is a TON of signal. But if your cycle time for only 20 MS/MS events is a full second? You've got high odds of missing that peak apex, and all that signal is gone.

(In case you're wondering 1ug of digest doesn't look much better)

So....more optimization is necessary...but for specific applications, maybe the HF deserves to get those flames!