Ha. Nature is dragging hard.
His obit said retired after 38 years and it had old names for things, similar to calling NIST the NBS, so I’m not sure when he retired. Figured there’d be reports. Need a real computer to look.
Kinda science: on vacation at Airbnb in Valle Crucis, owned by Parley Rasmussen [a biochemist in the Standards Division, Department of Agriculture (from obit)], who gifted 500 acres to state for park. When sold house in ‘22 they left some of his old stuff. But haven’t found his publications yet. 🤷♂️
Here is the paper ref btw. tl:dr most tools aren’t maintained and don’t work anymore. I also keep thinking this title is more Jay Z (99 problems) than Disney.
pubs.acs.org/doi/10.1021/...
Not sure why @jradoff.bsky.social hasn’t posted here yet, but they did over on reddit. Pretty awesome use case, imagine resurrecting and maintaining defunct proteomics tools @magnuspalmblad.bsky.social.
Spoiler, you can play it now here lofp.metavert.io
meditations.metavert.io/p/resurrecti...
They definitely knew better than to release it last week since it’s getting too close to our satire (or our satire has been too close to reality).
Aside: each new Timmy I misremember Kenny as Timmy.
I mean, I’m past amateur if only for getting paid to science, but not sure I like the connotation of “the biz” of science since I do think that is [often] at odds with the pursuit science.
Building small apps to fulfill your own needs feels super empowering and fun, we're truly in a new age of software hannahilea.com/blog/housepl...
Ha. As a person who does a lot with plants (another axiom to include is “never say thank you for a plant until lotsa time has passed and it lives”), this analogy, including bouquet, is quite apt. Thanks.
I can’t tell if this amalgamation is the best thing ever and pure brilliance, or just super weird. But if you’re ever in the south side of Soda City, this exit has it all.
Watch out Ron and Phil, I’m now thinking about JAMstack pages.
I have discovered this last year that I lean heavy on gallows humor.
Yeah, maybe being a bit facetious with might. It’s 100% an issue.
MS proteomics skywalkers... can I again iterate that we might have a problem. This 👇 is "proteomics space".
... and this is one of my fav proteomics applications. The spider paper from me and @proteomicsnews.bsky.social was kind of this, and I can say spider genomes are scarce, but fungi dbs are another thing. You know what the mushroom said as he got kicked out of the bar... "but why, I am a fun guy?!"
Now my addendum is there are a crap ton of fungi (like 14 M sequences last I played), and so you would need to think unique peptides to discriminate, but again, possible for sure.
Not crazy and arguably legit estabilished field. This one just got posted (h/t @kkmurray.bsky.social), but there are plenty (see CompOmics and @magnuspalmblad.bsky.social). With an actual mass spec (no offense ZooMS people) and a well constructed db you can do it.
www.biorxiv.org/content/10.6...
A consistent message in both Dario’s and now Sam’s, aside from doomsday, is they realize if we don’t get a little socialism/UBI, that it will break the system. I can’t tell if it’s cause the “break” is revolution or just they need people to have money to buy things.
openai.com/index/indust...
I’m now realizing some of these projects are things I was building analog, but static html is easier. Like was trying to create a family hike journal for trails we’ve completed around Deals Gap, NC, with pics, map and elevation change… scrapbook was hard to find time. Digital scrapbook is workable.
I will admit I’ve got a couple Gemini gems that run weekly and are like a Google reader 2.0 for me personally, but I didn’t feed them *all* the preprints at once (though told them to monitor preprints).
What’s a few good use cases you are using unpeer for?
Another thing is if I can spark the kids’ imagination I want to work on “vibe coding” their ideas with them, which annoys me on one hand but on the other seems smart.
My lightning app idea was c2008 sketched on paper, w/ me not having skills. Now, 5 minutes with Claude and it’s a reality.
The big realization was when I could make a better workout program bespoke for me, plus a dashboard that reads and writes to a Google sheet. And even better is I just feed it a pdf and it’ll make a program.
But yeah, there’s a lot and I’m fine being 16. My coding at 16 was on TI-82.
Ha. You mean too many personal apps? Maybe.
It really is silly all the little static html “webapps” you can make. This one is for my son since he never writes things down but he’s trying to keep track in his brain. The import/export is a little json file since I’m using persistent memory which can get wiped.
apps.benneely.com/media-tracker
I’ll save people a click, FTL is not Faster Than Light protein but is ferritin light chain. Whomp whomp. Also transcriptomics not proteomics.
Rats aren’t giant mice and humans aren’t giant rats, but still fun I’m sure for aging research colleagues.
www.nature.com/articles/s43...
There is a need.
Not science: soft shell crab season (it’s like a week tbh) is about to be over and it’s magical. Strong local feels about where to eat or not eat it. James island has the winners btw. B-)
www.instagram.com/reel/DWtbn2l...
Ugh, too early.
1) actual species fasta
2) close fasta + open search (think SAAVs)
3) de novo
4) maybe combine and search unmatched in 2 with 3.
Pretty sure I recommended that in the non model proteomics perspective paper from 2020 or whenever.
I mean v de novo. So 3 tests. Easy to demo and benchmark. Even can qualify how distant the “close” fasta is v performance. @proteomicsnews.bsky.social and me kinda did this with the spider paper wrt related fasta.
1) Proteomics has a severe problem.
We keep pretending database search is “good enough”, while systematically missing everything that isn’t already known.
No genome/proteome → no ID
Variant peptide → invisible
Novel biology → filtered out
I mean I think there’s probably some value in using a closely related fasta but running it for SAAV discovery. One could probably take a known species and compete real data v close fasta + open search + de novo. (De novo in my hands is like 50% ish wrong)