Saturday, June 19, 2021

I missed London Proteomics this week, but here are two great studies that were covered!

 

Friday was our first run trying nPoP and which required an early start and missing London Proteomics which was all about super cool drug discovery stuff. Fortunately, the site links to the most relevant papers the speakers covered (though, I assume we missed some top secret new advances.)


This week featured Ilaria Piazza and I've talked about here cool chemoproteomics stuff, the study linked is this recent one




It describes going hunting for small molecule covalent inhibitors that don't just mess with cysteines! Before some stupid virus derailed the world, the hottest drugs in the whole world were covalent inhibitors of the "undruggable target" KRAS. (KRAS is messed up in a huge percentage of tumors, and it's generally really bad when it's involved). If you can covalently inhibit a protein this terrible, what could you do with other drugs of this class? 

Friday, June 18, 2021

ASMS 2021 Vendor Wrapup for Proteomics Stuff?

Okay, this might take some time, because!! WOW!! what an ASMS hardware rollout! 

If I'm not covering your product, I apologize, its been a long year (and ASMS starts on Halloween!)

There were some seriously big vendor rollouts on the hardware front which got me to thinking ---


 From this Tweet, I learned a few thing. 

1) That there are people out there who didn't realize they were being eccentric with their low signal mediocre resolution data. 😇

2) It was kind of a tie between software and your relationship with your sales and support team. Which...probably doesn't bode well for Thermo since they recently got rid of their subject matter expert sales teams for their products in the US. If you weren't aware, they now sell by geography. So...your local rep who got her/his PhD from the famous proteomics lab? He/she might be selling ICP-OES systems now. Strategery from the company that can't seem to find a way to give up market share in proteomics fast enough to keep their executives happy! 

VENDOR LAUNCH 1: BRUKER! 

Bruker's launch was tied for the longest (3 days, thank goodness for how much time off we get in the USA) and was centered on the "TIMSTOF TRUE SINGLE CELL"


This smaller TIMSTOF was the result of a close collaboration with Matthias Mann's lab and based around the preprint from last May that still hasn't been accepted. The fact that it appears to use 800 picograms of peptide as an equivalent for single human cells might be related to that fact in some way (that's around 4 normal human cells -- assuming ZERO loss in digestion, which, of course is completely normal.)

The response was extremely warm from the single cell proteomics community. 


There are some interesting puzzles about this instrument. Such as how the box can't possibly be the hardware shown in the preprints (the ion source doesn't even go into the instrument in the same place). And there hasn't been much in the way of data to support that this thing works, but everyoen doing single cell proteomics is a novice because doing it is super easy, so I'm sure no one noticed these things. 

After "Fake single cell proteomics" was legit trending, Bruker altered the name a little. It turned out to be a translation issue, but the ions are still entering the instrument at the wrong place on the instrument.  

In the sadness of this release was some cells spread out on a plate and someone hitting them with a MALDI laser in some sort of a "single cell" approach which I think it's best we destroy the video evidence of and we never consider it again. I wasn't the only person that worried that Bruker was holding his dog hostage. I'd hit cells with a 20 um laser and sign things saying that it was 5 um laser if someone had my dog. 


One upside of this event was the TIMSTOF Pro X! This is a new instrument and/or field upgrade for TIMSTOF Pros! The price didn't seem absurd for current users and it's always cool to have a field upgrade for a 10 foot tall 2,000 pound box!

Legit, I think both of these things are totally cool, my job is to be a critic, I think 

ASMS Launch 2 and 2.5

The next week was coinciding launches by Waters and the first of 3 days of Thermo! 

Waters! 

I've been asked by multiple people "wow, you seem really pumped about this one" and I am from a theoretical standpoint. I'm extremely clear in my "mass spec physics taught by a dumb person" slide deck that you can't get above 70,000 resolution on a Time of Flight instrument. The vacuum on the flight tube just becomes far too long to make any sense at all.

Waters has a MALDI/DESI-TOF that can hit 200,000 resolution against the whole goshdarned mass range. Speed? About 10 scans/second. 


If you're thinking "my Orbitrap Fusion/Exploris/can get 480,000 resolution" let's stop for a second

1) How much time does that take? I don't feel like doing math, but it's probably about a second. (You can download my calculator here)

2) What about resolution degradation across a mass range? What? 

Okay, so there are a bunch of secret rules in mass spectrometry vendors that they must stick to. I can't tell you all of them, but:

1) The compounds that you are provided for calibration must be the stupidest thing they can come up with. A peptide that will only singly charge and will rapidly oxidize? PERFECT! A polymer that will stick in your system until the fucking end of goddamned time and will always affect your sensitivity? BINGO! 

2) Vendors can pick any one of these stupid inconvenient molecules and use that for the point where they provide their mass accuracy cutoff. As long as that molecule hurts the end user, it's fair game. 

On Orbitrap instruments, the resolution DECREASES as m/z INCREASES. Thermo, therefore specs their instruments on the lowest molecule they can get away with (which, to be fair, they use 200, if they really wanted to cheat, they'd go to 40 m/z where the resolutions is insane!) 

On TOF instruments they generally spec on that polymer you'll never ever get out of your system (acetone helps) and they aim for 1222 or maybe 922) and resolution slips (but not to the same level) as m/z decreases. 

For real, if you're doing stable isotope labeling (heavy glucose/glutamine incorporation, for example) on an Orbitrap) around 800 m/z you're out of luck. At max resolution (if you aren't running the 1M option) your heavy ATPs are all mixed up because your resolution has fallen off a cliff. You can't tell that heavy N from heavy C. That's a huge deal for metabolism researchers. 

So...200,000 resolution is a big deal to anyone in the medium mass molecule range who needs to resolve nitrogen from carbon isotopes. And 10Hz? That's fast. Yes, right now it's just MALDI and DESI, but how does DESI work? The last 3 letters are ESI! I'd be shocked if Waters doesn't have an LCMS super high resolution TOF out soon. I'd volunteer to test it! 

THERMO!

Ummm....okay...so due to the preprinting time of a really cool study from some ultra marathoning dude with who has great taste in high pressure turbochargers in the Seattle area, I think a lot of people were expecting somethig really super ultra cool for ASMS. 

What did we get? An ultra expensive metabolomics unit and a new FAIMS unit! 

(Again, my job is to be a critic. Who hired me, again? That's right! Nobody! Whassssuppp?)

Jokes aside, the FAIMS is super cool. 


It can do high flow without you having to hack it to do high flow yourself! And it's higher resolution or something. I don't have an instrument that it will go on right now, so I didn't pay attention. Again, anything that will add more capabilities to a currently existing instrument is awesome and there are tons of Thermo boxes out there with the next generation ion source that can get much cleaner data by tacking one of these awesome things on. 

And since this is the Proteomics blog and no one has discovered the hundreds of posts I've made on the Metabolomics blog, I'll move fast on this one. The ID-X (which I love) is a Fusion 1 that has been retarded (not in the insensitive way, in the way tractor trailers slow themselves in the Rockies way) do only do metabolomics, unless you've got 2 minutes and 15 seconds and a basic understanding of how to use a Windows computer to unlock it. The new instrument is based on the Eclipse and I bet it takes longer to release the brakes and make it an Eclipse. There is more intelligent acquisition software for both metabolomics and the small molecule biopharma community.  All of which does seem really impressive. 

SCIEX


You'd be fair if you said I've been "uncharitable" or "an absolute jerk" about SCIEX hardware over the years. For real, I have been. I started my career on ABI/SCIEX instruments and until I used a Q Exactive the 3200/4000 QTrap was my favorite instrument I'd ever used. I don't know what happened because from the 4000 QTrap onward all I ever saw was mediocre improvemetns on those instruments in difference colorss and medium resolution crazy expensive QTOFs and an overpaid marketing department (for real, why aren't they doing the NBA playoffs, not selling analytical instruments. As an aside, during this year's playoffs the NBA will interrupt the actual game to show a commercial about how you should be watching the NBA playoffs [for real, that's continuing to happen].) 

But this is something entirely different. The TIMSTOF is exciting because Bruker figured out how to accumulate ions prior to sending them into a TOF. Bruker has had a TOF for years that can hit QE level proteomic coverage, you just crank up your injections 10-fold. With the TIMS acquisition, you could get speed AND sensitivity. And this is where a ZenoTrap is similar. 

The new SCIEX instrument can acquire off the quadrupole (and the quad specs are legit. 0.4Da isolation in high res mode! they do know something about making good quads over there) and the ions can accumulate in the ZenoTrap before being fired into the TOF. Numbers from 4 really good speakers with real data? Looks like 5-20x more signal depending on the molecule.

In addition, there is a collision cell that is two 1 tesla magnets opposed. This (somehow! wtf?) induces a very democratic fragmentation similar to ETD and ECD...with very little decrease in scan speed! Birgit Schilling showed some great PTM fragment spectra to back this up. Think about how slow your ETD or ECD fragmentation is. You've got to inject your reagent and then allow the reaction to proceed, THEN scan your fragments. 50-100 ms reaction time is common. Not including the scan acquisition time, you're at tops 10-20 scans/second. Realistically about 5-10, max. This box can get around 100 EAD fragmentation scans.

I'm out of time so I'm going to just leave this here. I didn't catch the Agilent launch live and I will follow up on that one later, but I think it was small molecule IMS focused. And there were some other launches, but with this out of the way -- HALLOWEEN ASMS IN PHILLY should be just research focused, right?!??

Thursday, June 17, 2021

A couple hours until the final ASMS vendor reveal!

 


I don't know anything about anything, of course, but the final big ASMS hardware reveal (that I don't know about) is at 11am EST and....umm....you may not want to miss it. Whatever this top secret thing is might be the thing I'm most excited about. 

Don't be too grossed out about how much they spent on the video editing. There is real substance behind it. 

You can log in here at 11am EST


Wednesday, June 16, 2021

Serum proteomics of a lot of old dudes over a 12 year period!

 


This stats heavy paper takes a look at serum proteome changes correlated with different negative effects of getting older in a population of men 64 and older. 


Stats are ridiculously critical when you're looking at so many variables.

...how are you tossing these awesome outliers...?

I'm not qualified to really evaluate the downstream interpretation, but it seems solid and thoughtful. The upstream, however, was performed by depletion columns and using one of those giant Agilent time of flight instruments that have the ion mobility cell in the middle and the depth of coverage looks solid. 

I'm assuming they'll follow up with an analysis of elderly females shortly. 

Tuesday, June 15, 2021

Constant levels of KPR4 (at least in Arabidopsis) maintains cellular size!

 


I spend a lot of time these days thinking about how large human cells are and what their relative protein contents are. A great reference, of course, is the original Proteomics Ruler paper

I'll be honest, it never occurred to me to ask questions like.why is a human cheek cell 10x wider than a human red blood cell? This group did

Even if you hate arabidopsis, I recommend just looking at the pretty pictures in this study. They find a really interesting system where a protein called KPR4 maintains a constant concentration from a cell before and after it divides. It does this by some relationship with histones that I don't fully understand. Being arabidopsis poeple, they of course mess with this gene and can back up their hypotheses! 


Monday, June 14, 2021

glycosylated RNA (glycoRNAs) and tools to help identify them in LCMS data!

 



I was kicking around the idea of submitting a formal review on this topic, but I'm not sure I have enough data for more than a really cool blog post. If you do take this and write up a paper feel free to acknowledge me. I won't object. 

If you haven't seen this paper, you should.  This study presents compelling evidence of an entirely new class of molecules that we had zero idea were even around. The ramifications of this are at the level of text-book-altering. Glycosylated short RNAs sticking on the surface of cells doing -- presumably -- super important things. The initial study only identifies a few and they appear to be annotated as: 

"Non coding RNA" -- which means, they don't appear to make proteins. Why are they there? No idea. Till now. (Kinda).

I just realized that picture above from the PDF cuts off 5 of the author names. Pedro, Benjamin, Alex, Benson and Karim deserve to have their names shown in the PDF. Get your act together, ElfSeverer. Fixed it. 


Without running this out forever, these authors did some really innovative labeling and RNA stuff that I'm sure makes sense to RNA people. And these cool molecules disappear when you treat with something that cleaves RNA or cleaves sugars. They also did some top notch LCMS work. You can find all their RNA data on public repositories, but I can't find the LCMS files. It was obviously something that was secondary to their goals. 

It is not, however, secondary to my goals. And I bet that there are a whole bunch of people weird enough to read a proteomics blog that don't know that there are great tools out there for looking at RNA data by LCMS. There are, in fact, even SEARCH ENGINES and a couple of them are in formats that you already know how to use. 

A couple of things real fast, though. 

1) Nucleotides don't like to ionize in positive mode. Neither do glycans. However at a certain size of a molecule you can pretty much stuff a proton on it somewhere, but the signal might not be great. If you haven't calibrated your instrument in negative mode since installation, you might be operating at a handicap. 

2) Nucleotides LOVE to fragment. No joke. Check this out. 


I highlighted it but it's from this amazing recent study 




See what I highlighted above?  You know what that means, right?!?!?


No, not popcorn time....geez.....it just rhymes....

What it means is that the tools are in the friendly, open, ultrapowerful, free...



OpenMS! I was using 2.4(?) for intact protein work and I had to upgrade to 2.5 in order to use the RNA sequencing tool that this team used in this study (NASE). Honestly, this is probably the best tool for the job today from an automated standpoint, but it definitely isn't the ONLY tool for the job. If you use OpenMS, you're set, though! 

My bias toward using Proteome Discoverer is well established by now. So the first thing I tried was messing with the settings in the OpenMS community nodes RNxPL! You can get those here


They're from an older study by the OpenMS team where they used UV to crosslink RNA to proteins and then tryptically digested them to figure out what RNA is interacting with what proteins. Super cool workflow that I have always kept in my back pocket for conversations with people interested in RNA-protein interactions. It's hard to see in the screenshot above, but there are these extra yellow/green lines. What are those? They're nucleotide diagnostic ions!! No joke, this tool is super cool and really easy to use, and it's got amazing visualization capabilities. My only issue with it at all is that the diagnostic ion mass accuracy is a little wide and I can't figure out how to change it.

Please note, it only works in PD 2.0 and PD 2.1. You can always go to the FlexNet and get an older version. They're easy to install and run. 

Want a stand-alone executable with a bunch of power? You should check out RAMM! The RNA Mod Mapper. It is one of many tools for studying RNA by mass spectrometry that you can get here.  You can go down a complete rabbit hole of nucleotide mass spec work from the Limbach group that will run you back to some of the biggest names in mass spec and chemistry that he's worked with over the years. He postdoc'ed with some guy at Utah who's name comes up in undergrad chemistry.  This site is an absolute gold mine. 


What if you just want to do some manual exploration? For example.....wait.....we did some weird shearing experiments of growing cells one time...right...?...I can't remember exactly why but we wanted stuff off the surface of cells without killing them. Still have those RAW files? Just curious if you see an MS/MS spectra with a HexNaC oxonium ion and a nucleotide fragment ion or 6? 

Or...just entirely new to RNAs and what they are or how they might fragment? ARIADNE time! 





If you're using Xcalibur, I will warn you that it is a little frustrating looking for multiple diagnostic fragment ions. 

My recommendation is to make multiple overlapping ranges that are just filtered on MS2. (See below)


To do this, activate any MS2 spectra and then delete out the information. Then hit OKAY. It will then activate the filter that it doesn't want to that is only for MS2 spectra. Then you can put in your target ion masses. Once you have one done, you can highlight your first BasePeak (right below "Type") and then if you checkmark another box, then it will copy those settings. All you have to do is repeat and type in multiple diagnostic ion masses. In the example above, the 306.0491 fragment mass is pretty rare, the second ion (I think I used a HexNaC fragment) is much more common, but at 17.01 minutes I've got an MS/MS spectra with each within a 10ppm mass tolerance.

Is it a glycoRNA? No idea, but it might be worth trying copying that spectra out as an MGF and putting it into the RNA ModMapper! 

Ariadne isn't the only web based resource for modified RNA masses. It's just got the best color scheme. Another amazing resource is....



I can keep going, probably, but I should probably work on other stuff. 

One thing I'm currently hunting is SOS. If nothing else because the interface looks super cool. 



See, I found a ton of cool stuff, but probably not enough for more than a fun 30 minute blog post. And this really does just look like the tip of an iceberg. 

OH! I almost forgot! 

Why am I even thinking about these to begin with? RNA is totally amenable to separation with C-18. Waters has a 90+ page application manual for studying oligonucleotides by LC and LCMS. They use a slightly different BEH C-18 column, and negative mode for most things. A direct download for the Waters PDF manual is here

What I'm thinking is that there are lots of situation where we may have already accidentally have acquired data on glycoRNAs or RNAglycans and they're just hanging around in the background of our data. How much of the "proteome dark matter" could be linked to this? A bunch? Who knows?  I don't use nucleases when I prep my shotgun proteomics samples. 

Friday, June 11, 2021

Quantitative analysis of L-DOPA incorporation into proteins!?!?

 

Okay. Great study on a super depressing topic. L-DOPA is, in the short term, often a miracle drug for people suffering from Parkinson's. There are some unbelievable videos that you can find pretty easily of someone having symptoms and then -- boom -- they're gone. But the effects don't last forever. 

This new study shows that it's actually worse than just the development of resistance


(As an aside, I would like to point out how cool most of these people's names are.)

There is a lot of work in this manuscript, some MALDI was done, as well as the creation of new ways of generating controls(?) on a microscale level so they could even do proteomics in the first place. A QE Plus was used for a lot of the proteomics data analysis along with PEAKS and MaxQuant. 

What they find? Convincing evidence that L-DOPA can incorporate into the proteome of a human neuron cell line as well as a whole bunch of proteins that get all distorted in their levels of expression (top figure) due to L-DOPA related stresses. 


Tuesday, June 8, 2021

ASMS 2021 Hospitality Suites Week 2 starts in 3 hrs!

 


Just a reminder that there is more stretched out early morning probably alcohol free (no judgment, you do you, yo) ASMS hospitality suite stuff starting soon! 

Today should be Waters 

(Which is relatively easy to sign up for

I don't know nothin' about nothin' but if you have lecture slides on mass spectrometers, you're gonna want to pay attention to this one. I've removed slides from my mass spec physics lecture because of what they're releasing today. I had a bunch of simple equations that show why what they've got isn't possible on this planet (or at least very energetically unfavorable).

AND

Today is also Thermo's! If you get to page 4 of the sign up and you wonder if you'd accidentally clicked the wrong thing, don't worry, you're in the right place. I think this is tied to clues about their hardware releases because the site is called "La Broots". 

I put on my investigative journalism hat (that ASMS never lets me wear. No press pass for me. Maybe this year! We'll see!) this morning and discovered -- 


And! 


Hmmm...more riddles....maybe an inelegant response like an Exploris with an ion funnel big enough to pull in a cicada? 

We'll see! 

Wait. Where is the Hospitality Suites week 1 wrapup? There should be one somewhere. 


Monday, June 7, 2021

Preparing cells for single cell proteomics -- even more ways!

 


This is a continuation of last week's post on single cell sample prep. I meant to put a link to this paper in and just couldn't find it. 


Similar to the methods described last week, this one uses the CellenOne system. Dissimilar to those, this uses the NanoPots things (have you noticed how the West Coast of the US often does it's own thing with proteomics? The Trans Protoemic Pipeline doesn't appear to cross the Mississippi, for example.) 

NanoPots are of course super smart and have been in use for years now and the CellenOne can prep a ton of individual cells on single microscope sized slides! 

One more just showed up this week as well and this one is also very creative and gets great results from traditional 384 well plates.


This great new study also provides a new tool for downstream (Proteome Discoverer, woo!) analysis where the proteomics output integrates the sorting data! (What?? Yeah!) 

In the preprint from Karl Mechtler's team, they find around a 30% loss in signal from using the 384 well plates. Doesn't seem to stop this group, though, these results (Exploris 480 + FAIMS + ultra low flow nanoLC 3 hour gradients) because these results are fantastic! 



Saturday, June 5, 2021

Multiplexed Data INDEPENDENT Acquisition!


 WOOOOOOO! About time someone solved this one! 100% recommend this new paper.


They had to synthesize this themselves and it clearly has some limitations but now that someone solved a strategy for it, I bet other smart synthesizers can expand on it.

The tag neutral losses completely leaving behind DIA MS/MS spectra that don't change. (DIA spectra is already complicated, we don't want to multiply the number of fragments by 3 or 10). 


Friday, June 4, 2021

BIO-ID Proximity labeling of 192 markers!


What a great resource and...wow...what a ton of work


I'm not 100% clear on the differences between Bio-ID and APEX (except that the latter is faster) but they're both proximity based labeling techniques. I'm not really clear on them because in order to perform the technique you have to mutate or fuse a protein in your cell. Or...in this case....192 of them. Then under the right conditions your fused protein will biotinylate everything that is close to it. And just those things that are close to it. Violin! (Viola?) You pull down the biotinylated proteins and now you know all the things in the proximity of your protein of interest.

If you don't know how to fuse or mutate a protein you need to use another technique (or find a smart gene editing collaborator). 

This group appears to have some people who can do it. And maybe 192 mutants made in a human cell line isn't that big of a deal. I'm going with...it probably is as hard as I think, or it wouldn't have went in that Nature thing. 

After the mutations are made you need to do at least 192 experiments (one for each fusion protein) and they did this using an Orbitrap Elite operating in high/low mode using 60 minute gradients. Minor bit of interest for you serious mass spec nerds (weirdos.) is that the instrument didn't leverage ppm level dynamic exclusion widths, the instrument excluded everything within 0.6 Da. On something as simple as a Bio-ID experiment and a gradient of 60 minutes, I doubt that this has any real effect on the data, where in a global experiment it would cause you to completely exclude some coisolating ions of similar m/z. 

The downstream analysis was performed using Mascot and Comet through the Trans Proteomic Pipeline. Results from that were sent through SAINT. 

What did they get for all that work? 

Check it out for yourself at www.humancellmap.org

Thursday, June 3, 2021

Preparing single cells for proteomics -- count all the ways!

Single cell proteomics is kind of a hot topic right now, I guess! If you're one of the many people who seem interested in this, I'd like to throw just a tiny bit of caution your direction. You should be prepared for the worst sample prep you've ever experienced. I started doing SCoPE-MS at a company (that shall not be named) a have a couple hundred garbage RAW files to show for it before I got my first files that I'd ever consider showing anyone. Actually, after a fun conversation with a friend yesterday where we discussed her sending single cell requests my way and I'll send MHC stuff her way, I might back up and say -- MHC peptides might actually be worse, depending on how you do it, but single cell prep still sucks. But solutions are coming! 

Want the best possible single cell data? Get out that checkbook because you're going to need to drop some serious cash. I bought an LCMS, GCMS, and ICPMS bundle last year (COVID discount special) and that cost just about the same as the ---


Which was used recently in these two amazing new preprints! 

and 

What's cool here (maybe?) is that these both use very different strategies on the same expensive robot system. The first preprint from Hartlmayer and Ctortecka et al., shows the development of the ProteoChip which is so insanely smart that of course it is a commercial product that we'll probably see in July. 

Briefly, the cells are placed under an oil layer. The lysis buffer, trypsin, (if applicable) TMT reagent, and quenching are all dropped onto the oil and they sink to below the oil to do their respective jobs. The oil keeps these tiny amounts of solutions from evaporating. And...get this...the oil selected solidifies at autosampler temperatures, so you don't have to worry about pulling it up, because you can load directly from the ProteoChip into your autosampler. The weakness (in my hands at least) is that it does require a trapping step and if I run a trap column for single cell (especially a PepMap...gross...what they use here...[only shouting it out because this team in Vienna inspired me to ditch PepMap when I visited a few years ago]) my signal falls off a cliff. 

It's less of a problem here, I guess, because the signal is INCREDIBLE. The instrument in use is a FAIMS equipped Exploris 480. With no bioinformagics at all (a lot should be applied to single cell) I'm getting 700+ protein IDs from 80 minute runs when I reprocess their individual data files. That's crazy good. The US West Coast single cell crowd (Kelly and Payne labs) have gotten comparable, if not better numbers, but using much longer gradients and sample prep methods that...well....



There is a clearly an X factor in this method. CellenOne has realized that mass spec people are used to writing justifications of....



...so these chips might cost a Toyota Corolla each. Who knows? And they aren't out yet.

As usual, Slavov lab has something open and approachable... on the same platform! 


It turns out that DMSO is harder to evaporate at room temperature! And we know it doesn't interfere with us downstream too much (just please please please use LCMS grade DMSO, Pierce sells it. Just about every case of "DMSO in my running buffer on one instrument caused so much contamination that instruments 3 floors down were also contaminated" that I've been involved in can be traced back to cell culture grade DMSO on a mass spec. I'm definitely guilty of this whole "it's only 5% of the running buffer...oh...no....") 

It is worth noting, y'all, that the CellenOne is not the only option. Most of the great data we've seen out of NorthEastern and their Q Exactive Classics before April's preprint has come off of a little liquid handler called a MANTIS. And 



...if you google it, you'll find that thing.....still cool...but what you want is something else...probably..? I mean...whatever that is looks kind of awesome. But here is a link to what you're actually looking for. And my quote is closer to an average NanoLC (about half an EvoSep). EDIT: Someone was trying to talk me out of buying an EvoSep, I guess. It isn't more expensive than any other NanoLC from a major vendor if you buy it outside of an LCMS instrument bundle, it's just about the price of the Porsche GT4

What we have at Hopkins is half a CellenOne. Our department has a SciFlex Arrayer S3, which is the liquid handler side of the CellenOne (same company) without the sorter. Our single cell RNASeq core is under a 1 minute walk from where our robot is, and sorting 2,000 cells costs us $120 when performed by a guy with 20+ years of experience sorting out single cells, so we can go with a half priced setup. There are options out there! You might also have some famous sorting wizard within easy reach! 

I'm aware I just briefly touched on NanoPots and those things and they're clearly options. Kelly lab recently did some super smart stuff by modifying an OpenTrons and you should definitely check that out, but I've got lab meeting coming up!