Lightnews — Scholar-powered news

Reposted by Max Fürst

Sikkom @sikkom.bsky.social · 10h

Groen licht voor grote AI-fabriek in Groningen. Europa investeert 71 miljoen euro

De kogel is door de oude tabaksfabriek: Groningen krijgt een grote AI-fabriek. De Europese Commissie gaat 71 miljoen euro investeren. Het benodigde bedrag van 200 miljoen is daardoor bij elkaar.

dlvr.it

3 3

Reposted by Max Fürst

SusanK @susankleinfelter.bsky.social · 1d

Woohoo! Go Foldit!

Structural Biology @actacrystd.iucr.org · 1d

Andreas C. Petrides et al.: Reconstructing biological molecules with help from video gamers ##Foldit##PDBREDO##CitizenScience@umassd...##IUCrhttps://journals.iucr.org/paper?S2059798325008149

Reconstructing biological molecules with help from video gamers

The results of the Foldit citizen science puzzles for crystallographic models and their integration into the PDB-REDO databank are discussed.

journals.iucr.org

1 1

Max Fürst @maxfus.bsky.social · 2d

Maxim could you kindly clarify: did you actually recompute all the structures, or are the models identical to the previous AFDB version? And are the MSAs the ones from back in day, or do these reflect more recent db searches? If new predictions were made - with which AF version?

3

Max Fürst @maxfus.bsky.social · 2d

Great post! In agree that complexity is underestimated, lack of knowledge the bottleneck to design, & intuition key to success. Likely, AI can work work around some, but not all of the issues without explicit data.
Do you mind sharing what pockets you think work better than others with bindcraft?

1 1

Max Fürst @maxfus.bsky.social · 3d

Have you found the entire db MSA data as a dump yet?

1 1

Max Fürst @maxfus.bsky.social · 3d

Ooohh providing the MSAs is wonderful. Will the data be in a new Google cloud bucket? Is it already available?

1 1

Reposted by Max Fürst

EMBL-EBI @ebi.embl.org · 3d

We’re renewing our collaboration with Google DeepMind!

We'll keep developing the AlphaFold Database to support protein science worldwide 🎉

To mark the moment we’ve synchronised the database with UniProtKB release 2025_03.

www.ebi.ac.uk/about/news/t...

🖥️🧬 #AlphaFold
@pdbeurope.bsky.social

16 73

Max Fürst @maxfus.bsky.social · 4d

For sure. I'm not downplaying the breakthrough itself. But I do think the other, dependent achievements would not never have come. How much later - who knows.

1

Reposted by Max Fürst

AI x Bio Discovery @aixbiobot.bsky.social · 4d

Mapping the evolution of computationally designed protein binders [new]
Evolve comp. mini-protein binders *in vivo*. Maps seq-affinity.

2 3

Max Fürst @maxfus.bsky.social · 5d

Agree they should get the Nobel, but if it hadn't been that synthetic route, another one would've eventually come, good enough at least to make primers. Or we would've dugged earlier deeper into primases or protein priming. Oh wait, guess maybe we wouldn't have been able to study those without PCR 🤔

1 1

Max Fürst @maxfus.bsky.social · 5d

Un-frigging-believable. I have waited for something like this for so long!

2 11 45

Max Fürst @maxfus.bsky.social · 6d

Wait but that's inconveniently more complex and (😱) mathy, my 1st years will not be happy

Di Jiang @dijiang319.bsky.social · 28d

@science.org ⤵️challenges prevailing notion—repressor and activator act by distinct mechanisms: repressors block via steric hindrance/activators enable Pol II binding to promoter🧬

E. coli transcription factors regulate promoter activity by a universal, homeostatic mechanism | Science bit.ly/adv2064

1

Reposted by Max Fürst

Di Jiang @dijiang319.bsky.social · 28d

@science.org ⤵️challenges prevailing notion—repressor and activator act by distinct mechanisms: repressors block via steric hindrance/activators enable Pol II binding to promoter🧬

E. coli transcription factors regulate promoter activity by a universal, homeostatic mechanism | Science bit.ly/adv2064

13 31

Reposted by Max Fürst

AI x Bio Discovery @aixbiobot.bsky.social · 12d

The AlphaFold Database Ages [updated]
AlphaFold DB:
...needs updating as entries conflict with newer UniProt releases, showing bioinformatics databases' rapid obsolescence.

1 3

Reposted by Max Fürst

Simon Dürr @simonduerr.eu · 15d

We are hiring a PhD student 🎓

Work with me and Matthias Rüdt in our department on an exciting industry collaboration with Lonza combining high-throughput experiments and molecular modelling using AlphaFold and MolecularDynamics

📍 Sion, Valais, Switzerland

More info: www.jobup.ch/fr/emplois/d...

5 9

Reposted by Max Fürst

Danielle Navarro @djnavarro.net · 19d

Against my better instincts, I have written some notes on how human probability judgements work and what you should expect from surveys that ask people to guess what proportion of the population is transgender. I hope never to speak of this matter again

Some notes on probability judgement – Notes from a data witch

For the love of fuck, literally nobody thinks that 20% of the population is transgender. Please stop sharing that ridiculous YouGov statistic

blog.djnavarro.net

17 84 200

Max Fürst @maxfus.bsky.social · 19d

Finding enzyme homologs with beneficial properties, may often be relatively close in sequence space. Can largely address problem by plugging a sequence search after foldseek, but makes mining pipelines more complex.
If UI limitation, maybe as an a3m dump-only option for full AFDB or higher % clust?

3

Max Fürst @maxfus.bsky.social · 19d

Given the performance improvements, is a foldseek server searching the full AFDB conceivable? Would be very useful for e.g. for enzyme mining and hard to implement locally.

1 1

Max Fürst @maxfus.bsky.social · 19d

Maybe. Or, there is now such a fierce competition driven by commercial interests that pushes publications to happen at that rate. But is there still time for anyone to seriously scrutinize & compare the models and is able to keep track on what constitutes a major advance and what's a lot of smoke?

1

Max Fürst @maxfus.bsky.social · 20d

They should just strike a deal with nature, until version 10, all models get directly uploaded to nature.com instead of biorxiv.org

2

Max Fürst @maxfus.bsky.social · 20d

And how is Baker lab organising this internally? "OK, you guys work on RFd2. Meanwhile, you others guys enjoy the long weekend, when you're back you start work on RFd3"
Or are we now in a state where we can conceptualise, code, train, benchmark, test, and write on a new frontier model in 5 months?

2 3

Max Fürst @maxfus.bsky.social · 20d

Must be fun for the editors currently juggling the Rfdiffusion2 peer review. "Uh, authors have you seen there is a very similar, supposedly better model out there? Oh wait, it's the same authors"
Seriously, what's the point of reviewing papers anymore if we're now at <6 months update cycles 🧐

Alexis Verger 🧬🧫🧪 @alexis-verger.cpesr.fr · 21d

De novo design of All-atom biomolecular interactions with RFdiffusion3 www.biorxiv.org/content/10.1...

De novo Design of All-atom Biomolecular Interactions with RFdiffusion3

Deep learning has accelerated protein design, but most existing methods are restricted to generating protein backbone coordinates and often neglect interactions with other biomolecules. We present RFd...

www.biorxiv.org

2 1 11

Reposted by Max Fürst

David Bauer @davidbauer.ch · 21d

In the first seven days since I launched this new game, almost ten thousand people from 82 countries have played it. (Top cities so far: New York City, Zurich, Toronto, Chicago, Berlin, Vienna, Stockholm)

What topics would you like me to add next?

dataguessr.com

Dataguessr

Update your knowledge of the world. One quiz at a time.

dataguessr.com

3 20 43

Max Fürst @maxfus.bsky.social · 24d

The problem is: compared to these traditional seq analysis tools that were built and have been used for decades, ML protein design sees vastly faster update cycles and a massive delay / absence of (experimental) verification. A "vetted workflow" today is outdated in 6 months

1 1

Reposted by Max Fürst

Ed Zitron @edzitron.com · Sep 5

Premium newsletter: A 8500 word analysis of how everybody is losing money on gen AI, with some startups spending 100%+ of their revenue on compute. Even GPU providers are struggling, with years-long time horizons to break even, if profit is even possible.
www.wheresyoured.at/why-everybod...

Why Everybody Is Losing Money On AI

Hello and welcome to another premium newsletter. Thanks as ever for subscribing, and please email me at [email protected] to say hello. As I've written again and again, the costs of running genera...

www.wheresyoured.at

18 170 800