Aftermarket high resolution FAIMS allows separation of MEGADALTON protein complexes!

 

FAIMS gets a bad wrap because most of the commercial systems have a resolving power of something between 5 and 20. They're great systems if you just don't want to fragment (or see) +1 ions and you want your mass spec to only see +2 or +3 ions. Cleans up your spectra so your mass spec doesn't have to work as hard, and everyone is happy at the end.

But....is that a limitation of FAIMS technology itself, or is that what is mass 😁 produced for the general market? Sure sounds like it's the latter. 

In this new study an aftermarket/custom high resolution FAIMS system was coupled to a UHMR (which has an upper mass limit of 80,000 m/z? Is that right? That's huge) and oligomers of antibodies (so a monomer is around 150,000 Da!) were coupled. 

The study is maths heavy and there are a lot of formulas, so I found it hard to get to the effective IMS resolution. However, this 2019 study indicates that the FAIMS is >100 resolution, and I think the two devices are similar. 

And - get this - you can get this FAIMS system for the front of just about any instrument and they can be custom tuned for small molecules, peptides, or intact proteins. And they're a lot less expensive than the 10 resolution units that you can buy for only certain instruments. 

5-plex your data INdependent analysis methods by modifying dimethyl tags!

This has sat with a bookmark on it for quite a while I hoped that I'd remember to ask someone intelligent questions about it. 

Multiplexing DIA sounds like either the best or worst of both worlds. More samples/day but you've increased the complexity of your background so those magical neural network thingies have to think a lot harder. 

There is very little chance I'd consider 2-plexing my DIA. That isn't worth it to me in any way at all. 3-plex? That is enough that I bought reagents so I could eventually try it, but I haven't been anywhere near excited enough to actually do the try part. 

5-plex? That's worth thinking about. 5-plex without fancy expensive labeling kits? That's worth bookmarking.

Disclaimer: I've never dimethyl-labeled. I feel like there is some drawback to it, like the tags shift the retention times just a tiny bit? I forget. Again, meant to do some background research - and didn't.

You can read about it here!

White House document appears to address sharing of proteomics and metabolomics data?

 

I'm not sure what to make of this, but it is making the rounds on social media. It does appear to be in a fact finding mode. If you're interested in policy and how it might affect how and what data we share internationally, maybe you'd enjoy this quick read. 

Multi-study meta-analysis of dog samples with oral diseases!

 

I'm wrapping up a meta-analysis right now that I've bugged just about every proteomic informatics person I know about in one way or another. The insanely beautiful data that I'm working on reanalyzing is from a patient study where they said "you thought CPTAC was thorough? hold my espresso". No joke 33 offline fractions at 2 hrs each on a QE HF on these priceless human samples. The data is perfect for what I'm doing, but during the uploading of data for almost 70 patients, they missed a few here and there. The PI is now retired and the team is dissolved, so I ain't getting patient 36 fraction 32 or patient 51 fraction 7. Whether to keep these patients or not is what I've been bugging people about. 

BTW, this data just provided a spreadsheet of which patient was whom and I find it just fine to work with. 😉. I didn't need a short wave radio or to learn Morse code or anything. 

Crap, that's a joke like 15 people might get, again, right? I need to get out of my house. 

The great people at the Proteomics Standards Initiative came up with a meta-data standard for data in repositories called SRDF and it's a great idea for anyone wanting to automatically pull loads of proteomics data for reanalysis. 

Since my poor sense of direction would never allow me to find the utility closet at HUPO where the PSI meeting is allowed to have their annual meeting, I missed it. When I was told I was supposed to be using this format, Google helpfully told me that it has something to do with condensed short wave radio signals. 

And....now....we're around to the topic of this post....

These researchers successfully drew conclusions from an extremely wide array of proteomic studies deposited over time for the under-studied disease that they are interested in. 

Including ELISA data - old spot cut out MALDI-TOF peptide mapping stuff and some more modern iTRAQ and TMT work.

Now -- do not get this sideways -- I'm going to figure out how to use LesSRDF this summer and be a good participating researcher to allow my data to be re-utilized by real bioinformatician later. We're all going ot need to do this to help our field be taken seriously by the outside world. 

However, I do think this is a good example of studies where this hasn't been done and no one will go back and do it later, where some interesting findings were still made. 

SEER Proteograph prepped plasma on TIMSTOF Pro2 and HT!

 

There has been lots of excitement (and some scary good data) out of the SEER proteograph system for plasma proteomics.

Here is some more! 

While the authors clearly intended this to be more of a comparison of two very nice recently released instruments in their lab, as you can probably see from the figure at the top, the proteograph steals the show.

Clearly the 14-bit digitizer and higher capacity TIMS improve identifications, but the plasma precursors go up 3x - 4x when moving the prep to a kit that I have absolutely no idea at all how any of you can afford to use. 

The peptide loading plots are also really cool. I've never considered running over 400ng on the TIMSTOF Flex, and we only use that much when we use the EvoSep which runs closer to microflow than nanoflow levels. 

Registration is open for Cold Spring Harbor Proteomics Course 2024!

 

Do you want to buckle down and spend 2 straight weeks learning proteomics? 

There is probably no better way to make that happen than getting accepted at the Cold Spring Harbor proteomics course. You can apply here! 

I'm lucky enough to have been able to participate as a guest instructor twice in the past at this amazing workshop. I did have to take it off of my CV, however, because I'm not listed on the website as an instructor. 

The way I've described it to people is "Sunday night it's a bunch of people in a room with a slide deck that says "What is proteomics". Later that week there are 10 people crowded around an instrument monitor at 1:30 am watching phosphopeptides someone taking the course prepped themselves finally start eluting and fragmenting off an instrument someone taking the course is running themself. Then everyone celebrates!" 

It might have changed I was there a long time ago. 

If you do get to go pay very very close attention to the train station map. There are ZERO sidewalks and a windy back road with New York drivers so you and your roller bag may have to rapidly dive off the road with your roller bag if you walk in from the wrong stop. (Story I heard from someone who is very bad at maps). 

Deep learning - of glycopeptides!?!?!?

 

Okay.....on the surface this was first surprising that the first PTMs we'd see deep learning successfully applied to en masse was going to be glycopeptides. Then I thought....well...the problem is the stupid sugars all have the exact same masses. 

Here is an illustration from an unrelated study for fragmenting the fragments of the fragments to figure out what a glycan chain actually is because the fragments of the fragments of these important glycan chains still have the EXACT SAME MASSES. 

So maybe this isn't the biggest stretch in the world ever (link to this paper and topic of this post). 

Unlike the topic of yesterday's post, this tool is still in the - you need some skill with a computer to actually use it - but if you have such skills you can get DeepGlyco here. It does require installation on GPU for the deep learning magic. 

AI assisted proteomics - for everyone -with Ms2ReScore 3.0!

 

MS2ReScore is a great idea! Let's use Artificial Intelligence things to reanalyze our proteomics data based on various features of confidence to create bigger and higher confidence lists. It's free, too!

Then....you see what you're dreading you'll see. "DOCKER"?? "git"?!?  

Great...another tool for bioinformaticians written for bioinformaticians that I could spend the next 3 weeks learning how to use if I didn't have 11 mass spectrometers that are running suboptimally, a grad student who still hasn't returned from Dubai and 3 grant deadlines. (These are only examples, but you get the point, time is limited). Could I just have a Windows installer for some of this magic? 

You can now! 

Not to be missed here is MSAmanda 3.0 which can be installed in PD 3.1 or newer and has stand alone versions for PC, Mac and Linux and can be ran through PeptideShaker. MSAmanda 3.0 by default outputs the features that Ms2ReScore uses to reassign peptide confidence.

When using MS2ReScore you also get cool HTML QC plots along with a bunch more peptides. When compared to MSAmanda + Percolator, it looks like a solid 20% increase on low abundance peptides. The authors use a nice SCP dataset that I'm going to assume came from Erwin Schoof's group and then they do a bunch of analysis that makes it look like these peptides are real and make sense in relation to the other peptides. 

I'm all for more data analysis tools, and I'm even more for them when I can just download them and run them! 

PRECISE readout of MEK phosphorylation cascades by top down proteomics!

 

.....wow.... I won't lie. I'm stunned. I didn't think we were here yet.... and, to be fair, maybe we arent,but this group is! 

Bad background by Ben: A whole lot of the central regulatory pathways controlling tons of things in cells are based on some key central phosphorylation cascades. MAP kinase and MTOR are famous ones. They modify proteins by phosphorylating them because the modification can be fast and it reversible. If you go into any oncology centric place there is probably some really really really skilled pathway scientist (or 4, if they can afford them) who can dissect these cascades, probably through western blots and FACs. These people can tell you that if MEK1 S2998 is phosphorylated but T2992 is not - that means something critically important. 

When we do shotgun proteomics, we cut this region into a small piece and peptides phosphorylated once on the same region of that fragment coelute. It's often very hard, if not impossible, to tell which site is phosphorylated - or both. 

Obviously we should do this without digesting them, right? The problem with that is that top down proteomics only really works on small proteins. If MEK1 was 16kDa, it would still be tough, but you could do it. MEK1 is almost 50kDa, though! 

Enter "Individual Ion Mass Spectrometry" IIIMS? (older post explaining what that is here) and - what?? - ETD IIMS? 

I was going to cut in parts of the materials and methods section here, but I don't want to intimidate anyone into thinking this is clearly just a proof of concept. However, I will say that the process to get to these data makes label free single cell proteomics look like a nice fun day (which it is NOT). 

However, we have to start somewhere and being able to confidently localize 4 separate phosphorylation sites on a critically important protein this big - with anything - is a step in the right direction! 

US HUPO 2024 special live episode of THE Proteomics Show is out!

 

No joke at all, this is by far, my favorite episode of the podcast so far. RASR has a great radio voice, asks super smart questions and Dr. Olga Vitek has such a cool perspective and story to share! Such great content that Ben couldn't even ruin it. 100% recommended!  

Asparagine to iso-aspartage conversion in norovirus infection!

                            

Holy cow, y'all, this one was definitely not a fun puzzle for these authors to sort out! 

The punchline here is that there is a spontaneous post-translational modification - get this one - it is a deamidation of asparagine - which makes it the exact same mass as aspartate - that changes both the protein 3D structure AND alters binding partners.

How do you go about even suggesting that is what is happening? It seems like they started with trying to model the norovirus capsid protein by NMR and it was fuzzy and suggested two forms. They ended up doing a lot of NMR and hydrogen deuterium exchange mass spectrometry. 

Deamidations are very easy to mistake for naturally occurring isotopes, so they used inline automated pepsin digestion and used 120,000 resolution on the MS/MS spectra (!!!) to help properly resolve the fragments and this story. Seriously, just a monumental effort to show how a tiny and extremely simple virus can use clever biochemistry (I've never heard of at all until now!) to cause complex things to happen. 

Improve intact protein analysis with outlier rejection borrowed from astronomy!

Apparently mass spectrometers aren't the only thing that generate "spectra"! And we aren't the only people who have to worry about averaging lots of measurements because we have limited signal.

Would it be worth borrowing some tools from one of these other fields for -- a 45% increase in intact protein identifications?!?!? 

Considering how many people in our field will drop $1M for a 20% increase in IDs, I'd say this is worth taking a look at. And - you don't even have to - it looks like you can just upgrade MetaMorpheus and run some topdown proteomics through it (good benchmark dataset here) and just use the new "averaging with rejection" feature. 

 

THUNDER-PASEF - Get those real immunopeptides!

 

I need more time to spend on this, but I'm going to put it here anyway. 

What's this about? 

Based on comments I've received about what I write recently I'm going to be trying to provide what I think is important context in anything I review. Obviously, context is from my perspective and background

Okay, so HLA peptides are super annoying to analyze but they are very very important. These little peptides are on the surface of our cells and they are part of the system that tells our immune system DON'T EAT ME. It's all self-self recognition stuff, sort of. When that system isn't working perfectly it's bad. Cells that are diseased (like cancer cells) aren't being targeted and destroyed. When it is working too well, it is also bad (auto-immune problems). So it's important that we can categorize those cells on the surface. HLA Class II are easier. The peptides are bigger and easier to sequence. Not easy, per se, but easier than Class I.

Class I peptides are much smaller than the ideal targets for mass spectrometry. Also, as they are produced after proteosomal degradation, it is very unlikely that the amino acid on the end is a lysine or arginine. Which is the opposite of useful for mass spec based sequencing.

And - we've known for a long time that just running a standard proteomics method on HLA peptides is suboptimal.  Now that I'm an academic, I am wondering why I didn't publish that..... I probably should....

So...how would you go about doing this if you had a really nice 200 resolution ion mobility instrument on the front of a mediocre mass spectrometer? 

THUNDER! Na na na na na-na-na-na (caution: actually the video link. might be loud)

Okay, so from Figure 2 it doesn't look like Thunder makes a lot of sense, it looks more like you should just not use a TIMS octagon. Which.....in my experience has never ever been a good thing...

Get further into the paper when they model the octagons. It definitely improves things. 

Big takeaway here could be - YOU CAN USE MORE THAN 1 TIMS OCTAGON!  I did know this, but I didn't learn it until a few months ago. The company that had a 2023 revenue of almost $3 BILLION still hasn't found the money to make any instruction manuals for said instruments, but that's understandable, of course.

But the numbers here are solid. The gold standard for immunopeptidomics is around 50 milligrams because that's about the size of a normal biopsy punch sample. If you assume 30% of that is protein you're at 16mg or something. If you start with 100 million human cells, assuming a protein content of around 200 picograms/cell you're at 20-ish mg of material! 

If you work out HLA peptides from 100 million cells, you're at a level where you might actually help find a cell surface marker that could be used to inform an immunotherapeutic therapy - or - design a new personalized therapeutic to target and destroy that tumor. And this is where we've been trying to get to from a sensitivity perspective for a long time. 

I missed the 20th US Human Proteomics Organization meeting!

 

I missed the 20TH MEETING OF THE US HUMAN PROTEOMICS ORGANIZATION thanks to my immune system being a slacker. The committee I chair, the US HUPO Virtual Media Outreach Committee (most of the group is above) sent me pictures like this. 

I just found this and I'm brazenly stealing it -- 

Is this the same meeting that we'd been trying to interview all the invited speakers for? Yes. Well...not  all of them, THE Proteomics Show has to fit into space around our jobs and families and things, but we got a bunch of them. 

The same meeting that the spent months working on so we could have a cool proteomics video competition? Yes. Same one. Super bummed to miss it. 

The winning video was a parody video called "The Real Scientists of Beverly Hills" and came from Cedar Sinai. The concept makes me laugh every time I think about it, even though I didn't see it. 

Because proteomics is now (FINALLY) a thing the outside world thinks about once in a while - legit news outlets sent legit reporters. Not joking. Here is a link! 

The 60 second lightening talks at US HUPO are, of course, legendary. Apparently there was some beatboxing. If you've never seen these rapid fire talks, you're missing out. Again - sooooo bummed I wasn't there. 

THE Proteomics Show still happened, just with a legitimate upgrade of one of the hosts! (Whoa. RASR lab website is seriously cool....) 

I have every intention of listening to this one (it will be the first one) when it goes live. Maybe today? 

Big shoutout to the award winners this year. I am ultra bummed to have missed 

Kelleher's Don Hunt award! 

Parag's Computational Award winning talk! 

Jenny van Eyk's always inspirational talks on why we NEED PROTEOMICS IN THE CLINIC. YESTERDAY. 

And the Cotter Early Investigator Award winning talk by Ying Zhu

I like good award winning talks as much as anyone, but posters are my favorite part of any conference. That's where you get to meet the smart young people who are actually in the trenches running the instruments and pushing the limits of what we can do with any proteomics technology. CRITICALLY ULTRA BUMMED. Next time.

If you haven't heard - US HUPO 2025 is PHILADELPHIA!!

Also - have you ever been to a conference that was really disorganized? Like you can't figure out where things are going on at each point in time or stuff starts too early or too late? You know why this doesn't happen at this one? US HUPO has 

From left to right (ignore Pratik, Lydia, Jennifer and Brianna, ignore Ben) from Conference Solutions. And they make all the things work! 

Separate single cells, extract metabolites and ionize them in one go!

 

I'm going to bore (boor? bour?) anyone unlucky enough to visit this awful blog because this study is on:

Single cells

AND 

Metabolomics

This is a 3D printed ionization source that allows you to put in a population of cells. Then the cells have to go in single file because they are being pushed down something too small for them to bunch up in.

While they are being pushed through the metabolites are extracted out of them!

THEN the metabolites are directly ionized into an Orbitrap (looks like a Classic!)

And...no joke, it actually looks like it works.

If you're following along in the statistics world, someone recently realized that there are problems with some of the advanced statistics used in single cells. We'll ignore it right now 'cause the dust may settle one day, but according to these plots, they can tell different cancer cells apart using this crazy source!

FUNCTIONAL spatial proteomics is live!

In a very early entry for paper of the year....

Quick overview - one of the goals of any sort of bioenergy production has been to figure out how the actual fuck microorganisms break down cellulose into usable energy sources without literally catching it on fire. If you can release the base sugar molecules trapped in WOOD you could do easy things like convert that to ethanol or butanol or other handy things. 

The problem is that these things are not happening in a homogeneous way. There are typically diverse populations of organisms widely dispersed throughout the material that are doing these things. 

How do you decide where to actually focus your efforts? 

How do you know that where you are analyzing is actually where the cool stuff is happening? 

Do what this groundbreaking work at PNNL is doing! You first slice your material and you use mass spectrometry imaging (they're using MALDI-FITCR I think) to locate the breakdown molecules that you are interested in.

THEN you take that same material off and you cut tiny tiny tiny tiny sections out of the area where those molecules came from. They're using NanoPots / MicroPots and then doing proteomics / metaproteomics. 

So now they know what enzymes are directly linked to the small spatial area where the lignocellulose breakdown products are coming from! Killer, right? Okay, but get this, why can't you just use this exact same system and apply it to other biological materials? We got a glimpse of what is coming down the pipeline from PNNL when we interviewed Dr. Burnum-Johnson on THE Proteomics Show earlier this year. There is something coming that is definitely worth setting a google scholar alert on this PI's name. 

Single particle analysis enabled by 25 second transient times!

Okay....y'all. Are you ready for ABSOLUTE SINGLE MOLECULE SENSITIVITY?

Are you willing to accept 2 SCANS/MINUTE to get it?

No?!?!? 

Why? Oh. You have multiple studies you'd like to complete before you retire? 

Well.....crap....here it is anyway! 

Look, there isn't enough helium left on earth for unimportant things like cooling large magnetic drivn FTMS systems which...could...well.... do this...ummm....same thing....?...at higher resolution and 50x faster....? We need the remaining helium on earth for MRI scanners, which are actually important. Now....how the "Everything costs $1" store my kid loves can seemingly fill seven thousand Helium balloons a day...for $1....is still a complete and total mystery to me. 

Obviously, measuring a "single-ion" of anything is really really cool and not everyone's first motivation is how many chromatographic peaks across a scan they can get. Honestly, the most impressive part is probably the fact that "singe ion" didn't degrade in 25 seconds while making 400 zillion rotations around the inside of that little tiny vacuum chamber! 

Need more surfactants for proteomics? Here is a big table!

 

Okay, well, that looks grainy and useless, but if you are one of those "low sample input" people, you might also be looking for surfactant ideas! 

We've all clustered around DDM because (so far, knock on wood) it doesn't seem to destroy everything. Longitudinal data would be nice and those of us using it to coat a lot of things will probably have some of that soon.

BUT THERE ARE MORE! And you can read about them in MCP here!

They seem to have one they like better than DDM, but they have to remove it from the peptides afterward, which....is less immediately interesting to me personally....

but I sure am glad to have a whole table of things to try! 

Is it time for cheaper, faster alternative enzymes for shotgun proteomics?

 

Moving fast, but subtilysin ripping through intact proteins in seconds isn't the worst idea I've ever heard.

Strongly recommend you check this one out! 

My alma mater ended a deal with Google - blog functionality impacted!

 

So.... short story is that Google and Virginia Tech had a long standing deal to work together. However, Google wants money for things these days and my alma mater was like - ummm....we'll go to the amazing people at Microsoft if we have to pay money.... cause.... outlook is clearly almost 1% as good as Gmail....

So....I need to migrate 140GB of data by tomorrow night to somewhere.....

Many blog links leading to my Google Drive are already down (ugh). This includes any historic methods on www.LCMSmethods.org that haven't been migrated successfully to protocols.io

I'm unclear what happens to the other 138GB of pictures and .gifs that are linked to this site. LMAO. DO YOU KNOW HOW LONG IT TOOK ME TO DOWNLOAD 138 GB OF GIFS?