Monday, November 13, 2023

What does an LCMS proteomic run cost? Breakdown for my lab.


This weekend on BlueSky we spent a lot of time talking about what proteomics costs today per sample and I thought I'd do a run down of what it costs our lab to run a sample. 

Further clarifications for the most popular post in a few months (eek)

This should be called. "What is the bare minimum expense for a PI with 20 years of LCMS experience to complete a single shot LCMS shotgun proteomics experiment at a depth of approximately 5,500-7,000 proteins measured per cell." We are not a service center or core. We are not regenerating our costs in any way. This is the minimum cost for us to do this aspect of our research. 

EDITS AT BOTTOM! While I was driving in through the Thunderdome today and should have been paying attention to one of the most dangerous stretches of road on the entire east coast, some critical dislaimers came to mind - then when I saw the popularity of this post I felt like these were important enough to hold up my breakfast. 

Our biggest costs, by far, are our instruments. 

The TIMSTOF Flex we purchased in late 2020 (when I arrived at Hopkins) was a ridiculously good COVID panic deal from the vendor. No joke, I think it was $500k less than what I was originally quoted. 

This year's lease payment on that instrument is $188,000 or something. I might be rounding it one way or the other but that makes me think it is about $500/day for it to be here.

Mike MacCoss estimated that his people's salaries are a lot more than his instrument leases. I suspect people are a lot happier to go to work at University of Washington because I can take my salary and two other people's in my lab and those don't add up to our instrument lease. It's close-ish and below I'll use it in the math to assume the same thing.  

I should know for certain, but I think because of the laser on the Flex our annual service is higher than something like a TIMSTOF Pro. I believe it was $60k. I probably won't be able to afford a service plan in 2024 and the 2023 one has been a poor investment because literally nothing has broken on it the entire year. My laser went over the "maximum number of pulses" limit in 2021 and it still passes QC. 

Since we're going to cross 8,000 LCMS runs (and, obviously a LOT of MALDI runs) any day it is pretty clear we do just about everything with this Flex so I'll just focus on it for now. 

Then software:

SpectroNaut increases in price this year, but last year my key was $7,000 academic

Proteome Discoverer has some really weird pricing structure stuff now that I honestly don't understand, but it was $2,500/year to keep the Protein Center nodes active. This year we had to purchase xxx number of hours of Jackalope even though we can't use it so I'll just round that to $3,000 so I can have software at $10k for the year. 

We prepare every sample with S-Traps. I prefer 96 well plates for prep and the last batch I ordered that ran out to about $3 per S-Trap. 

I just discovered some students in the lab are still using the complete S-Trap kits that come with reagents A/B/C/D and those come to around $20/sample. I generally keep those around to give to collaborators with cool samples who have never prepped a sample before so I need to do something about that. $20 each adds up to a lot. The spin columns are sort of in the middle. Maybe $8-$12 depending on how many you order? I might round it below

We have a paper coming soon where we investigate different trypsins. Honestly, it was a lot cooler on the poster Ahmed had at US HUPO. We found it less interesting as we moved into organs and plasma so it kind of went on a back burner. We estimated our typical trypsin usage to be about $7/sample

We use the EvoSep for every big project. Honestly, I'm getting closer to a point where I don't know why we use anything else. If I've ever said anything remotely negative about the EvoSep, ignore it. We had some early issues with ours that were largely environmental. As one core director I respect a lot asked recently "Are your instruments outside?" Not quite. When it rains in our lab it might not be raining outside, but the temperature inside and outside are typically about the same. It's a tough place to be an instrument. 

EvoTips run us about $2. If we use IonOpticks columns they last about 2-4 weeks. If we use PepSep columns and we never take the column off at the ZDV union (they universally fail at the "NanoViper" junction) one of those can last months. The chromatography isn't nearly as good. However, if we do need to switch those, we can take one off and put it back on about 3 times before it fails. Tough to really estimate, but I'd put the Flex column consumption at about 1 column/month. So $800 for a PepSep or $1200 for an IonOpticks. Let's go with $1k/month. 

Quick math then! 

$500/day for instrument

$500/day for PI, 1 lab specialist, 1 graduate student

Service + software @ $70k/year + nanoLC columns at $12k = $82k lumped = $225/day.

For most things we're using EvoSep 30SPD which is 44 minutes LC gradient. We have a big study coming where we did 128 organs with EvoSep30 and 60SPD and the protein numbers are very close, but the number of missing values with 60SPD was a lot higher. With that many biological replicates, it's sort of a how much do you care at double throughput, but let's keep it at 30SPD.


$60 for EvoTips ($2/sample)

$300 for S-Traps (or $100 if we use 96-well plates, let's assume this even though it is actually higher for me right now) so $3 or so

$210 for Trypsin (or $7/sample)

And if we take the $1225 for lease,service,software columns and salaries for 3 of us that adds $1225/30 ~ $40/sample + $2 EvoSep + $3 S-Trap + $7 trypsin = 

$52/proteomic sample cost to me....

sort of assuming 24-7 round the clock nonstop operation with no QCs and no downtime. We're productive and ambitious, but that is absolute fiction. 

8,000 LCMS injections on the Flex in the approximately 1,000 days we've owned it says we're doing maybe 8 runs/day on average? (Minus MALDI time -- those imaging runs are typically like 1 day per slice, I don't want to do that math) 

At 8 samples/day we're $153/sample but the % MALDI vs LCMS is tough to break down. I'd like to think we're closer to the $52/sample. 

Or about $1,500/day when we're running full speed. 

What's missing? Biognosys QC standards (iRT), K562 from Promega. The big data processing PC that we bought when I got here for $12k? That is full and sorta slow by today's standards and could use a replacement in the near future. My monthly payment to IT to protect everything. The air filters for the captivespray source (clean labs go through them a few times a year, we have to change them every 2 weeks or so). Calibration mix. Adds up, but it is honestly sorta trivial compared to the other things. 

Obviously this gets a lot better when the Flex is paid off next year, but that's where we are. So if someone offers us $2/sample on a grant to do a project.....I might not want to be on that project....

(Not my example - someone else was offered $2/sample....not joking...)

New disclaimers! 

Okay, so - please keep in mind that the National Institute on Aging and National Instutes of Health Center for Genetic Medicine pay for me and the people in my lab. One of the things they also pay is for about 64% of every dollar to go to lab indirects. The samples that we're running are for these projects. Some other funds trickle in other places and for a pending collaboration with a company, a little over 76% of every dollar will go to overhead/indirects...honestly I've tried to talk them into working with someone else, but they like me. That's a big deal because the day these overheads stop coming into my university, I'm unemployed. So is my team. 

As such, we're doing research here on what we found interesting and important enough to write some grants on and we're hugely enormously forever grateful that someone said - WHOA, you're right, that is important, buy some stuff and hire some people to do it. (Thank you thank you thank you, you know who you are and you're on every slide deck from the lab). 

As such, even if we were running 24/7 (which we aren't, but we try really really hard to use every minute) any project that isn't - you know - actively keeping me and these cool young people here employed - is more than detracting from these amazing gifts from the government and corporate collaborators provide so we can work on stuff we hope will make the world a better place.

When I say a minimum of whatever above - that is for me to do work on these projects that I'm genuinely privileged to get to work on. If you offer me $52 to run a sample for you that I don't think is interesting, I'll show you that our wet lab has 3 separate doors (we're in a corner, so there are plenty of ways to leave!)  On a good day I'll provide you some guidance on good cores. The one 3 floors up from me is awesome. They sure as hell won't do anything for you for $52 because they have to make up all their costs and - I hope - they pay their people a lot better than I'm allowed to pay mine. 

That's a rant, maybe. But if you're reading this and thinking you'll leverage this to get some cheap proteomics samples ran because you know what one lab's rough math is for what it costs them to run stuff on their own instruments, you aren't very good at reading.

Someone at Harvard (hi Amanda!) brought up the training aspects. Oh yeah! Okay, so I've been running instruments for 20 years. And we still pay for instrument and software training and apps support. Vendors might argue I beg some freebies, but we've definitely cut POs for training.  And -- I had a student rotate through the lab a couple years ago who killed 3 nanoLC columns on her first batch of samples and her rotation resulted in one figure that might make a paper when I write it mid-2024. Not calling him/her out, but training can be expensive. Another example - we had the TIMSTOF Flex from just before Xmas of 2020 and the first data to make a publication didn't come off the instrument until March. It takes a lot of time to learn how to use one of these things. The big NIA project I mentioned that pays for most of this has over 1,000 separate LCMS runs associated with it. We probably will only upload to MASSIVE around 400 because we just got way way better at diaPASEF in the last year and nothing new comes from the old samples, just making us look like we didn't know what we were doing (mostly me).

Since this was a lot of fun, maybe I'll do the metabolomics one later. I don't know that math. I do know we spent like $6k? $8k in pure standards? It was a lot, but I'll look it up. 

No comments:

Post a Comment