Max Fürst

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

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

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

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.

Figure 1

Figure 2

Figure 3

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

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 🤔

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

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

@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

The AlphaFold Database Ages

Figure 1

Table 1

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

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...

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

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

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?

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.

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?

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

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?

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 🧐

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

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

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

Why Everybody Is Losing Money On AI Hello and welcome to another premium newsletter. Thanks as ever for subscribing, and please email me at ez@betteroffline.com to say hello. As I've written again and again, the costs of running genera...

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...

I often wonder whether all the folks flocking to (lured into?) longevity research are thinking about who and what this tech will actually benefit

Bluesky now platform of choice for science community It’s not just you. Survey says: “Twitter sucks now and all the cool kids are moving to Bluesky”…

Good to see @bsky.app has emerged as the social media platform of choice for science stuff
arstechnica.com/science/2025...

I'm essentially unable to log into any websites run by Editorial Manager. I have 20 something different logins saved, with different usernames and passwords, for different journals. My password manager is useless, and I am at the point where every login requires a PW reset.

Thanks a for your work and this article! Though, as a currently part time working father of a young kid, I would have loved to have seen a male perspective included, too. Perhaps in a follow up :)

Then, spend most of the time explaining why it's actually worth their spending the next years of their life on it and be open to (partly) let them find their own ways

You could give them the proposal to read and tell them "this is the idea, but in all likelihood we won't get to do half of it, and that half we'll actually gonna have to do very differently then what's written here"

Congratulations on Your New Academic Position of Assistant Professor of Annual Reviews, Merit Reviews, Peer Reviews, Student Evaluations, Tenure Review, Post-Tenure Review, and [Your Discipline] Welcome to our faculty. We have recently changed our titling norms to more accurately reflect how you will allocate your time. We will conduct the ...

"The goal is to place in the top category for merit pay that year, a range that could be anything from zero to upwards of $75 per academic semester. There is no actual money available for merit this year, but we will still conduct the reviews. The merit is symbolic."

Accelerating Biomolecular Modeling with AtomWorks and RF3 Deep learning methods trained on protein structure databases have revolutionized biomolecular structure prediction, but developing and training new models remains a considerable challenge. To facilita...

(1/7)
Training biomolecular foundation models shouldn't be so hard. And open-source structure prediction is important. So today we're releasing two software packages: AtomWorks and RosettaFold3 (RF3)

[https://www.biorxiv.org/content/10.1101/2025.08.14.670328v2](www.biorxiv.org/content/10.1...)

My biochemistry I class has an ongoing project known as “my enzyme is the coolest” in which the students compete for their enzyme to be selected as the SUNY ESF Enzyme of the Year.

I am putting together my list of nominees for the students to pick from. Please comment with your nominations!