New Nature Genetics study comparing pQTLs is....worth reading....

 

Ummm.....so...Imma just leave this here and not talk about it any more, maybe. Wait. Maybe just this - if your technology is producing results that can be validated 30% of the time then you could save a lot of time and just pick a gene or protein and flip a coin and go read up on other technologies....

DIA Multiplexed proteomics with off-the-shelf TMTPro reagents!

 

This is obviously interesting - and surprisingly easy to pull off. The data is processed in FragPipe and one of the output sheets is put into these python tools to identify the complementary fragment ions. 

I like the figure above because they use 2 very similar peptides labeled with TMT and demonstrate that they can clearly find clean complementary fragment ion pairs. Oh yeah, here is the paper. 

They really really don't want to do any spectral deconvolution so they only used 3 of the TMTProC tags that give them clusters of complementary ions 4 Da apart. The open suggestion is here the whole time that if you aren't afraid of deconvoluting your complementary ion clusters - you can obviously do more than a 3-plex DIA experiment. 

This is a really nice read with the appropriate controls included as well as a way to dramatically increase the throughput of some DIA proteomics workflows on basically any mass analyzer. Worth a read for sure. 

If you type "TMTc" into the blog search bar you'll find a lot of stuff over the years. This is one old post that goes more into what this is and why it can be valuable. 

Y-MRT - a new prototype TOF with 1 million resolution and 300 Hz?!?

 

Ummmmm......okay so....these specs are amazing....

How do you increase mass resolution? Generally just increase the flight path, right? But you can only go so far before there isn't enough electricity on earth to generate the appropriate vacuum. Reflectrons double the path and the W-TOFs from Pegasus that a big vendor acquired recently can really push those numbers up by multiple reflectrons.

The Y-TOF takes that concept to 11. It's one thing to say "I can make my instrument do 1 million resolution". Give me 45 minutes with your Q-Exactive and I can make it do 1 million resolution. Each scan will take about 8 minutes. (more like 4 seconds, I forget) but it's completely impractical. 

AND you can tune a Time of Flight to get really good mass resolving power at one particular m/z. My Q-TOF gets incredible resolving power in a mass range that isn't exactly where I need it.

The Y-TOF did a 30 minute proteomics run and averaged 600,000 to 800,000 resolution across the usable peptide range!!!  

AND sub-PPM mass accuracy. Parts per BILLION mass accuracy. ON A TOF. 

Obviously a prototype, but more obviously something we should keep our eyes on. Worth noting, they do have to use Astral level loads for bulk proteomics (1 microgram of peptides for the best data) and that this prototype isn't going to smoke your recently purchased $1M instrument, but it's starting in a very nice spot. 

Prosit-PTM! Deep learn modified peptides???

 

We all know other great protein informatics teams are working on the holy grail for DIA proteomics - deep learning and prediction of modified peptides.

Am I extra excited because the team that gave us Prosit is working on it? Yes. Yes, I unfairly am, when I should be evaluating this preprint purely on it's own merits and not the historic success of one of our field's most historically reliable teams.  And not just because of their informatics skills. What makes me excited the most is their long history of making tools that anyone can use. 

Check out this preprint here! 

Breaking through barriers with an Orbitrap-TOF instrument!

Thanks to all the journals allowing Open Peer Review and allowing me to sign about half of the 30 or so papers I've reviewed recently, it's pretty clear to people how unproductive I think things like this title are. Even if, as in here, I really do like the paper. 

I think I'm just old, for real, but I do think that if you've got cool biology in your paper but you've got the instrument front and center you're doing yourself a disservice. 10 years from now that instrument is going to be $100k from second party vendors or $50k on Ebay without an ionization source or accompanying PC and no one is going to look at the biology in your paper. 

However - this is some pretty amazing crosslinking data. And that's my point, I guess. It's a nice study. FAIMS helps a lot with crosslinking on both an Eclipse and Astral, but stepped collision energy - while helpful on the Tribrid, has minimum effects on the Astral. Higher CE helps a lot. There is also a neat toolkit I heard mentioned at iHUPO called "Raw Vegetable" which I assumed someone said but actually meant "Raw Beans" (which I love). You also get a cool step-by-step breakdown on how to optimize crosslink data analysis in Proteome Discoverer. They do some filtering inside the software and break it down at every level. Super helpful for anyone using that toolkit in their lab (I think they use 3.1).

Worth noting, they used the freely available MSAnnika node for the crosslinking, which is pretty cool to see it in use - and optimized through. 

Plasma fractionation increases proteomic coverage!

 

Y'all aren't going to believe this one. For real. 

Everyone out there complaining about the number of proteins you can identify in plasma proteomics and no one has ever tried fractionating it??? What is wrong with us? 

Whoops. Not everyone got that this is a joke. Okay...so...literally for all of time everyone has fractionated plasma in some way to get higher coverage. That makes the title funny. 15 years ago I was running SDS-PAGE gels and cutting fractions and running them separately on an LCMS. 13 years ago I was using something called an OFF-GEL to first fractionate the proteins at the intact level by isoelectric focusing and then digesting the proteins in those fractions and fractionating them at the peptide level to get proteomic depth. The problem with fractionating is that mass spec time is expensive and if each sample takes 144 hours to analyze (example) you only complete 60 samples per year if you never run a QC, a blank, your instrument never needs maintenance and you work every single day of the year. The UK Biobank study 1 would take 833 years. Most people aren't that patient and we're all sort of looking for ways to get a lot of samples analyzed before we retire. 

Hypothetical multiplex tag works in single cells?

 

Did you know there are other tags out there for multiplexing proteomics? Younger people probably don't and I can't tell you where they to get them because I actually truly can't. Let's change the subject entirely.

Did you know lawyers are seriously expensive? Like, for real expensive. If you're struggling with science salaries maybe you should check it out. Okay, let's go back to this paper.

If hypothetical multiplex tags did exist in some places where I couldn't tell you about could those ficitious tags be used for single cell proteomics?  Are you thinking....ummm...yes...? why wouldn't they be? These people found a team of peer reviewers who thought it was useful to check - at Analytical Chemistry! 

And they compared it directly to the commercial reagents that we all know and love. They used the same intrument Orbitrap Fusion Eclipse using 120kDa MS1 and 30kDa MS/MS. The only difference is that the fictitious tags that don't exist and if they did I couldn't tell you where to get them used a slightly lower m/z cutoff. They also optimized at 5:1 tag to peptide. 

For the experimental design here, they sorta mailed it in. 3 tags were used for cells and other tags were used for different controls. They also optimized the carrier to single cells by basically not saying "...why would it be different for one multiplexed reagent when 15 different papers already optimized this on the same Orbitrap hardware...? and said you could go about 100x - 200x to one? 

Then they actually did some interesting stuff by labeling mouse spleen cells with 13 of their available 16 channels. The most interesting part is where they find that if they don't use FDR at all ("set to 100%") they can get 12,543 proteins in mouse spleen cells!!! Someone said "ummm....wtf....you need to use FDR..." and they get 3,991 peptides and 3,602 proteins. So....1.1 peptides/protein on average. Ouch. The FDR calculation scheme is ...nonconventional.... and I almost want to download their data and reprocess it in FragPipe and see if the data is good but the data analysis is unnecessarily strange. Oh. The fun I had when I had free time....

Interestingly, however, the authors get those 3,602 one-hit-wonder proteins to clearly separate the different cells in the mouse spleen into their originating cells, and generate a beautiful T-SNE or U-Map plot, and that's what we came here for anyway, right? The authors suggest some follow-up experiments where they plan to combine both their tagging solution with the amazing commercial one....

GlyCounter - find all those glycopeptides whether you fully sequenced them or not!

 

If you've ever tried to look for a glycopeptide in any type of MS/MS spectra you know how very very rare it is that you get all of the information that you're looking for.

If you want to get full sequence coverage of everything it's probably going to take ETD and 2 different energies of collision dissociation of some kind. The clever combinations of energies certainly help get you more fragments, but they also increase the background complexity. "Is that 8ppm away from the b5 ion or could that actually be the NeuNaC is the third sugar in which case that's the 3 ppm off of the z4 after the loss of a less likely HexNaC at the end? (I possibly made that up because it's equally funny to me if that is a chemical impossibility or if it isn't).

Do you need ALL the info, though? Sometimes I just want to know things like "did this drug increase the number of spectra with glycan related oxonium ions".  I definitely do want to know more than that, but that's what I know how to do with some clever R scripts Conor Jenkins wrote me almost 10 years ago. 

How do you get to real information - and spectra - for glycopeptides in your data in an easy way? 

You don't.

Until now! Hello GlyCounter! 

You're probably assuming "cool, now I just need someone to download some crappy python scripts, fix them and then make me a dummies guide on how to run them." 

NOPE! Check this out! I wouldn't write about it if it was the python thingy, probably.

It's a slick little GUI that takes straight RAW files or mzMLs! Click your options (including whether you used UVPD or ETD(!!!!!!!) and it does the rest, including kick out handy IPSA annotated spectra! 

Important - if you are using a non-Thermo format and convert your data to mzML you don't want to compress them. In MSConvert, turn off that thing. Honestly, that thing messes up a lot of other workflows. If you're converting through FragPipe, it might convert them by default depending on what version of FragPipe you're using. 

Gotta run, but if you need a solid new and approachable toolkit for glycan modifications, you should absolutely check this out. 

Broadly inhibiting PFEMP1 antibodies sequenced from a single child!!

 Sneaky HUGE PAPER ALERT! 

Direct PNAS link here. 

This is so so so cool. Here is the thing, PFEMP1 is this molecule that covers the surface of malaria parasites and it does something very similar to what our antibodies do. It switches domains around to that it is incredibly variable. For a while it was thought that because it's so huge (I forget but I think it's 600kDa or more) it might be able to switch around more than our antibodies.

Back in the 1990s or something the amazing Michal Fried was getting malaria samples from women in Africa who had gotten malaria while pregnant - multiple times. The first time was generally really bad. Like as bad as you could imagine, but if the adult survived the next time she was pregnant she and her baby were basically immune to all forms of Plasmodium falciparum malaria (maybe the others, I'm not sure). Those data basically proved that a malaria vaccine could one day be engineered and today we do have one. It's insanely absurdly difficult and expensive to produce, but it exists. 

It also shows that the human mAB can outcompete PFEMP1...somehow.... and if we could just exploit that gap in flexibility between our immune system and the protective systems of the parasite - we could have much easier to produce vaccines. 

But you can't sequence polyclonal human antibodies....right...? This team seems to have!

If you're interested in the story I mentioned above -- 

https://www.nature.com/articles/27570

KOINA - Proteomics machine learning for everyone!

 

There have been some bioinformatics initiatives where the goal has been "any dumbass can do this, even you!" and it has turned out that 

(pronounced doom basses, the latter word sounds like the instruments) 

Is KOINA really something we could all use, regardless of what languages we do/don't know and how very long ago we had any formal training in informatics? I'm not sure, but the authors seems to have tried really really hard to make it so. 

The first problem a lot of people run into when integrating tools into other tools is what language is supported? Machine learning is almost always in Python and proteomics people like Java and various versions  of C. 

30 different tools are integrated into KOINA in an attempt to both make this all more open and to show off that they can take data from and to different languages! 

The next thing is that even within the same programs peptide annotations can be completely different. Ever try to make a Venny from a peptide list from 2 different software packages? Chances are zero overlap because someone's like, "let's absolutely for no reason whatsoever provide the c-terminal amino acid with a period after it AND THEN let's make up a non-standard abbreviation for this PTM and put it in italics while underlining a random letter in this non-standard abbreviation" one of those is absolutely real, the other one probably is, I don't know everything.

So the high school sized list of KOINA authors decided to do something no proteomic informatics person in all of history has ever tried. 

THEY CONSULTED THE PROTEOMICS STANDARDS INITIATIVE (we have a proteomics standards initiative????) (PSI)

I just got back from International Human Proteomics and rumor was that this year's secret meeting of our field's least popular group of people was held in the nosebleed section of the Sabrina Carpenter concert, ensuring that they actually annoyed more people this year with their suggestions of how making up new ways of doing the same things is stupid and we should stop, and that no one who could possibly need this lesson heard them. 

KOINA includes steps to convert peptide level annotations to a standardized format that PSI suggested that we all use at some point and we all pointedly ignored. 

I need to get going, but if you read this far (sorry) you should check out KOINA here! 

SCPLine - SCP downstream analysis tool that looks easy to operate!

 

No time to check this, but I want to come back to it later, and this is the easiest place for me to find it again! 

SPIN - Super impressive/fast low contact/low volume single cell proteomics I bet you can buy some day soon!

 

This was a fantastic read this weekend.  Both the technology and the fact it is a cool story overall! 

I should, however, start with - no, you can't buy one yet. This is some advanced prototype of what was a student project a few years ago. You can read and watch a video about the Isolatrix here. 

If it was for sale right this second, I wouldn't jump in line to buy it. The reason it is so amazingly super fast is that it makes a decent number of mistakes, but it's smart enough to identify those wells with zero cells or more than 1 cell, or one funny cell itself. While this sounds simple to fix, it's easier for us to run every "single cell" well rather than skip the ones that are funny. That works when we're getting >85% single cells in our current workflow, but I currently wouldn't trade 100x faster for a whole lot more errors till some other issues were sorted out.

Now, when they do compare it to our current single cell isolation solution in our lab - Isolatrix preps in lower volumes and a ton faster AND gets visual data, so it's neither close nor really fair. Higher coverage per cell and the ability to go back and evaluate the cells that you were working with visually? That's a lot. Now...my solution was $25k when I bought it ($42k list now) and the chips were like $12 a piece (about $50 each now) and so we expect better newer tech to crush it in head-to-head analyses.

If you are a mass spec nerd and want to see what a TIMSTOF SCP vs Ultra2 look like on the same cells. This is another good comparison of the early and newer hardware.