Kyle Swovick

For some reason, I feel like it may have been from the beginning of the proteomics show episode with Dr. Olga Dudchenko (like within the first t minutes). Could be very wrong though

I recently finished "Everything is Tuberculosis" by @johngreensbluesky.bsky.social and wow, he was right. Even my proteomics data is tuberculosis.

Oops. Forgot to give a shout-out to @justinwwalley.bsky.social for a really cool talk as well.

After @proteomicsnews.bsky.social 's SCP session at ABRF yesterday, someone from our genomics core asked me "are proteomics talks usually that interesting?"
How it felt as he was asking (we're slowly turning them to the dark side):

I'm a sucker for wiener dogs so I enjoyed the Instacart one...enough

Can't be Lamar because it's definitely Goedert.

This makes sense. I recently came across a paper comparing DIA methods and was using PSMs as a comparison metric and had me wondering how useful it actually is in these scenarios as an actual metric

And now for the real naivety to shine...is that really a good term for DIA data? Wouldn't each precursor ID would inherently have used several MS2 spectra (assuming you don't have 1 point across the peak)? If I'm dead wrong (probably), I would love for a good explanation.

This may be an incredibly naive mass spec question, but what's a PSM for DIA data, particularly in Spectronaut? Is it simply referring to a confidently ID'd precursor?

But if anyone has other ideas on possible search tools that could be useful to have ~20-30 people running a search at once, or even just to have them visualize their spectra, I'd love to hear them.

Right now though I'm thinking of using last year's runs to make a chromatogram library of the top 20ish proteins in Skyline. Then to make a targeted method and have all of this year's runs uploaded to a project on Panorama using the library from last year.

But last year, we tried doing this for the first time since getting an Astral and learned the hard way that it's just not feasible to use ProteinProspector for this class anymore. So I need to come up with a different way to do this lab.

We used to use ProteinProspector since it was all browser-based (no software to install and no need to worry about Windows/Mac compatibility). Then we could have the students mess with as many different combos of mods + species they wanted. And they could visualize the fragmentation spectra.

We’ve run into a database search problem that maybe #TeamMassSpec could help with. Due to the massive file size of Astral (and timsTOF) runs, we need to change how we do the MS/MS database search section of our undergrad biochem lab (IP of two different fusion proteins expressed in E. coli).

Yeah. MS1 was completely fine. MS2 in the orbi was completely fine. With MS2 in the astral, there was absolutely nothing

Hah! We still have our broken one as well. They said they would take it back when they did the PM. Guess who didn't take it back when they did the PM.

Would it be surprising if I told you that almost all of their 2,000 unique protein IDs are, in fact, one-hit wonders.

Actually, I just did some creeping and I would hazard a guess that he was fixing your board when we called.

We had ours die back in July of this year...about 9 months after install. Our engineer was at Wistar replacing the exact same board when we called support.

I do love seeing these cloud-based pipelines coming out. Unfortunately, my little pea brain cannot comprehend how to even begin setting anything up on AWS. Luckily for my pea brain, our Research IT is pushing university researchers hard to use their NAS's. So they let us store all our files there.

That was the first thing we noticed when we demo'd the HT. Going from lumos files to those was definitely a double-take moment.

I've got many other things to recommend, what to avoid, etc but this will get way too long way too quickly. But if you do have any more questions or just want to browse through our methods, feel free to reach out at any time.

One thing you will definitely need to think about is how you will plan to manage your data....the files are huge and you can generate them so quickly now. So far in 2024, we've generated 9TB of data and we're nowhere near 100% utilization.

I've read some papers where people are loading in the ug range which seems wild to me. We generally inject 250ng unless we're running a longer gradient...we've seen diminishing returns injecting anything more than 100ng and you'll start getting space charge effects with too many ions in the Astral.

But in general, some other things to keep in mind: Astral deflection mirror drift dramatically impacts MS2 signal...we've learned to run a system suitability standard daily and whenever MS2 dips, do a full calibration (usually every few weeks, but it is more dramatic during temperature changes).

The second time, we were essentially losing all MS2 signal if it was analyzed in the Astral. If we did MS2 in the orbi, everything was fine. Turns out it was a dying supply board (this might be a problem part...our engineer was out at another site fixing the exact same thing when we reported ours).

Specifically, I can think of 2 times where we were completely flustered. The first was shortly after it was installed, we just lost all connection to the Astral but the front end was fine. Turns out it was a faulty Ethernet -> Ethernet adapter.

Of course! We've had it for a little more than a year at this point so we've come across several issues along with developing some best practices to prevent some of those. Overall, we really like ours and it's been a workhorse (which seems like it a mixed bag) with a handful of frustrating moments.

With the 72SPD method with the column size Kevin said at 600nl/min, our peaks are generally 7-10 seconds. And at least from our 2 species mixes, both coverage and quantitative precision+accuracy are much better with 2Th windows and 3ms max ITT

🤣 Same with us. Hopefully you don't run into the subsequent issue we had with a bad Ethernet connection that is there because of ✨reasons✨.
Had us down for a week because no one could pinpoint the issue