1/28 How do you optimize a dynamic protein property that emerges from multiple states? Our finally published paper in @NatureComms takes on one of the hardest problems in protein engineering with phage assisted evolution: evolving allosteric switches🧵

Technical Assistant

Are you interested in working with a fresh team on cutting edge protein design?
We're hiring a Technical Assistant for our protein design team at DKFZ Heidelberg: jobs.dkfz.de/en/jobs/1683...

Please share!

ChatGPT please explain the concept of roundtripping in the financial context like I am five and make a statement on its legality in the general context as well as the specific case shown by this graphic...

Not sure how to reconcile A) bindcraft consistently outperforming RFDiffusion, but B) AF3-gen diffusion-based design outperforming AF3-gen hallucination-based design. My best guess is that there's something about AF2 but not AF3-gen methods that makes them amenable to backprop?

Announcing Rust 1.90.0 | Rust Blog Empowering everyone to build reliable and efficient software.

Rust 1.90.0 is out! blog.rust-lang.org/2025/09/18/R...

Major features:
* lld as the default linker on x86_64-unknown-linux-gnu
* cargo workspace publishing support built into cargo
* more const context APIs, and stabilization of new useful APIs

Exciting to see our protein binder design pipeline BindCraft published in its final form in @Nature ! This has been an amazing collaborative effort with Lennart, Christian, @sokrypton.org, Bruno and many other amazing lab members and collaborators.

www.nature.com/articles/s41...

overheard, with regards to nushell "That is pretty cool that you can just do that kind of stuff with the shell scripting. " 🩵

The Office JS API certainly made some choices...

Rational engineering of allosteric protein switches by in silico prediction of domain insertion sites Nature Methods - ProDomino is a machine leaning-based method, trained on a semisynthetic domain insertion dataset, to guide the engineering of protein domain recombination.

Publication alert: Our paper on domain insertion predictions in proteins is now out in @natmethods.nature.com in its final form: rdcu.be/ey7w3
Also check out the nice perspective by @noahholzleitner.bsky.social and @grunewald.bsky.social : www.nature.com/articles/s41...

Scaling down protein language modeling with MSA Pairformer Recent efforts in protein language modeling have focused on scaling single-sequence models and their training data, requiring vast compute resources that limit accessibility. Although models that use ...

Excited to share work with
Zhidian Zhang, @milot.bsky.social, @martinsteinegger.bsky.social, and @sokrypton.org
biorxiv.org/content/10.1...
TLDR: We introduce MSA Pairformer, a 111M parameter protein language model that challenges the scaling paradigm in self-supervised protein language modeling🧵

MMseqs2 v18 is out
- SIMD FW/BW alignment (preprint soon!)
- Sub. Mat. λ calculator by Eric Dawson
- Faster ARM SW by Alexander Nesterovskiy
- MSA-Pairformer’s proximity-based pairing for multimer prediction (www.biorxiv.org/content/10.1...; avail. in ColabFold API)
💾 github.com/soedinglab/M... & 🐍

Beeswarm plot of the prediction error across different methods of double perturbations showing that all methods (scGPT, scFoundation, UCE, scBERT, Geneformer, GEARS, and CPA) perform worse than the additive baseline.

Line plot of the true positive rate against the false discovery proportion showing that none of the methods is better at finding non additive interactions than simply predicting no change.

Our paper benchmarking foundation models for perturbation effect prediction is finally published 🎉🥳🎉

www.nature.com/articles/s41...

We show that none of the available* models outperform simple linear baselines. Since the original preprint, we added more methods, metrics, and prettier figures!

🧵

Rational engineering of allosteric protein switches by in silico prediction of domain insertion sites - Nature Methods ProDomino is a machine leaning-based method, trained on a semisynthetic domain insertion dataset, to guide the engineering of protein domain recombination.

Our work on rationally engineering allosteric protein switches is now out in Nature Methods: www.nature.com/articles/s41... Thanks a lot to @grunewald.bsky.social and @noahholzleitner.bsky.social for the comprehensive news and views: www.nature.com/articles/s41...

I talk a lot about Rust for building high-perf (& even non-perf critical) software, & scientific software in particular. I often discuss what's interesting to me, but wanted to offer the chance to those interested for me to answer their questions about Rust in science. Fire away with questions!🧬🖥️

Absolutely spoiled by match and if/else being expressions in Rust and Nushell. Really feels pedestrian trying to achieve the same in Typescript or Python.

(No the ternary operator is not "control flow as expressions" we have at home)

#AwesomeTools 1: uv docs.astral.sh/uv/ has transformed how I work with Python projects locally and for development and distribution. No more pip, venv, poetry, etc. seamless package management, virtual environments, development and deployment deliverer FAST & without hassle!

Announcing Rust 1.88.0 | Rust Blog Empowering everyone to build reliable and efficient software.

Rust 1.88.0 has been released! ✨🎊 🦀

This release brings you let chains, naked functions, automatic Cargo cache cleaning, Cell::update, proc_macro span locations, as_chunks methods on slices, and more!

Check out the blog post and release notes for all the details: blog.rust-lang.org/2025/06/26/R...

Congratulations!

So many names that sound perfectly with an epic announcer voice:
- POGO-PANCE!
- RAMPhaGE!
- RECOMBITRON!!!

Excited to share my main PhD work is finally out! 🥳

We built a phage-assisted evolution platform to evolve allosteric protein switches (POGO-PANCE) and introduce cumulative, targeted mutations/Indels (RAMPhaGE)!

See how we evolved ultra-strong, light-switchable AraC variants 🌚🌞

I started tgv to learn Rust and building it has been unbelievably fun! If you wanna pick up a super fast and fun programming language, tgv is open for contribution! There aren't many bioinformatics tools built entirely by the community. If tgv can become one, I'll be so psyched

New paper alert! We introduce the modular allosteric thermo-control of protein activity. Employing the AsLOV2 domain and mutants thereof as thermoreceptors, we engineered diverse hybrid proteins, whose activity can be controlled by small temperature changes (37-40/41 °C).

doi.org/10.1101/2025...

Nushell 0.104 is out: www.nushell.sh/blog/2025-04...

gh's @cosineblast continues to iterate on job support, with experimental support for message passing between jobs

AFESM: a metagenomic guide through the protein structure universe! We clustered 821M structures (AFDB&ESMatlas) into 5.12M groups; revealing biome-specific groups, only 1 new fold even after AlphaFold2 re-prediction & many novel domain combos. 🧵
🌐 afesm.foldseek.com
📄 www.biorxiv.org/content/10.1...

Panel 1: a program's input and outputs are streams of bytes everything you type goes into standard input all the output you see comes from either standard output or standard error your terminal emulator can only communicate with programs by reading/writing bytes panel 2: some inputs/outputs are text and some are special instructions in: mouse position, ctrl+left arrow out: make text green, make cursor invisible panel 3: these special instructions are called "escape codes" they're called "escape codes" because they all start with the ESC character panel 4: example: how colours get set program: “^[[31m” Terminal emulator: “ok, I'll make text red from now on!!” (there are also codes for bold, underline, background colour) panel 5: programs can easily "break" your terminal by printing escape codes program: “oops I made your cursor disappear” It's easy to fix though, run reset to print a special escape code that resets everything It's easy to fix though, run reset to print a special escape code that resets everything

terminal escape codes

wizardzines.com/comics/escap...

When you walk into Blenderman you only hear Arginine....

"If you're used to being able to sit down and bang out code, any time, anywhere, with all kinds of tools (libraries, compilers, virtual machines, what have you) at your fingertips, then yeah, working up a new assay protocol in a cell line is going to seem agonizingly slow" (from 2015)

IDK, just that they won the one fighter contract :P

"we read the preceding work and can confidently predict the results to be similar to a LLM"