Seva Viliuga

Global Analysis of Aggregation Determinants in Small Protein Domains Protein aggregation is an obstacle for engineering effective recombinant proteins for biotechnology and therapeutic applications. Predicting protein aggregation propensity remains challenging due to t...

All data are available on Zenodo and we'd love to see what you can do with it! Cydney also made a nice notebook to run the predictive model (including DMS scan) in Colab!

Preprint: biorxiv.org/content/10.1...
Colab: colab.research.google.com/drive/1KNWvG...
Data: docs.google.com/forms/d/e/1F...

A proper way of commuting through piles of snow 🤣

Exchanging for an authentic Wienerschnitzel 🤠

Biggest achievement of 2025 😂😂 First ever self-made cake seems to be a success!!

Are you interested in integrative structural biology, but feel a bit lost?

Don't worry, we have you covered with our FEBS Advanced Course, Lost in Integration Vol. 2 — probing biomolecules with AI and experiments

probingbiomolecules2026.febsevents.org
network.febs.org/posts/integr...

Folding models learn protein stability only implicitly. Without access to negative data, one can in principle make use of the folding free energy (dG) and the change in the free energy upon mutation (ddG). I believe simple aux losses could help for cases where a mutation is clearly disruptive.

Hi Thomas, that's indeed the case! However, this would also mean that one can reliably refold and score redesigned sequences only of those proteins whose structures were deposited post AF2 training cutoff date. This is a big limitation in our opinion!

Had a great time presenting our work on flexibility-conditioned protein structure design at @embl.org and right after our recent work on generating conformational ensembles of proteins at @euripsconf.bsky.social together with Nico. Follow along - we will be soon releasing more exciting things :)

New preprint out!
We present "Transferable Generative Models Bridge Femtosecond to Nanosecond Time-Step Molecular Dynamics,"

GitHub - richardshuai/fampnn: Sidechain conditioning and modeling for full-atom protein sequence design Sidechain conditioning and modeling for full-atom protein sequence design - richardshuai/fampnn

FAMPNN or PPIpack?

github.com/richardshuai...

colab.research.google.com/github/Kuhlm...

You asked and we listened... @workshopmlsb.bsky.social is excited to be expanding to Copenhagen, DK at @euripsconf.bsky.social 🎉

Two workshops (San Diego & Copenhagen) will run concurrently to support broader attendance. You can indicate your location preference(s) in the submission portal💫

Good luck and enjoy, Seb!

Hope you all had a good summer. I'm very happy to announce the speaker line-up for the falls Chalmers AI4Science seminars! Hope to catch you all there!

If you’re interested in learning more about protein folding and misfolding, I’ve created a convenient reading list with a few essential papers:

scholar.google.com/citations?us...

scholar.google.com/citations?us...

6/6 Amazing team work with great co-authors
@leif-seute.bsky.social, Nicolas Wolf, Simon Wagner, @bioinfo.se, Jan Stühmer and @graeterlab.bsky.social

Try out FliPS and BackFlip yourself, code and Google Colab tutorials are available on GitHub!

5/6 We introduce a framework in which we generate candidate protein structures conditioned on flexibility with FliPS and use BackFlip to select the candidates whose predicted flexibility profile best matches the target before running expensive MD simulations.

4/6 We also introduce BackFlip - an equivariant network that can accurately predict backbone flexibility as derived from MD simulations. Crucially, BackFlip infers flexibility solely from the backbone geometry without requiring evolutionary information, making it useful for de novo protein design.

3/6 In a series of experiments, we demonstrate that FliPS samples novel, realistic proteins with diverse secondary structure composition and a remarkable resemblance to custom target flexibility profiles, as verified in 300ns Molecular Dynamics (MD) simulations of the designed samples.

2/6 Our model FliPS is a conditional flow matching model for protein structure generation. FliPS receives a flexibility profile as conditional input feature and learns how to generate realistic protein structures while respecting target flexibilities.

1/6 One of the key features of functional proteins is their inherent structural flexibility. In our recent work at #ICML, we introduce flexibility to protein structure design! More in a thread below.

Code / Tutorial: github.com/graeter-grou...
Poster: W-109, Thu 17 Jul 11 a.m. PDT — 1:30 p.m. PDT

First time transatlantic and such a view over Greenland 🤯 Wish there were more glaciers :/

Hahaha this is hilarious!! I’m now totally convinced I should play with it as well 🤣

1/4
🚀 Announcing the 2025 Protein Engineering Tournament.

This year’s challenge: design PETase enzymes, which degrade the type of plastic in bottles. Can AI-guided protein design help solve the climate crisis? Let’s find out! ⬇️

#AIforBiology #ClimateTech #ProteinEngineering #OpenScience

This is not happening, right?

New pre-print from PhD student Hang Zou on warm-starting the variational quantum eigensolver using flows: Flow-VQE! Flow-VQE is parameter transfer on steroids: it learns how to solve a family of related problems, dramatically reducing the aggregate compute cost!

"...we investigated the performance of ... inverse folding models when trained on different structural datasets and on held-out set of experimentally determined PDB structures."

It would be interesting if they did it vice versa and ran inference on some FoldSeek or dark clusters of AFDB...

🫠

YouTube video by Science Communication Lab The Most Beautiful Experiment: Meselson and Stahl

Just learned that Frank Stahl (of the Meselson and Stahl DNA replication experiment ("the most beautiful experiment in biology") died at the beginning of April, to no fanfare. Here's a lovely video of them reminiscing: www.youtube.com/watch?v=7-tn...

You can now check out a recording of Julija's excellent talk on the Chalmers AI4Science YouTube channel: www.youtube.com/watch?v=y_7a...

🤹 New blog post!

I write about our recent work on using hierarchical trees to enable sparse attention over irregular data (point clouds, meshes) - Erwin Transformer, accepted to ICML 2025

blog: maxxxzdn.github.io/blog/erwin/
paper: arxiv.org/abs/2502.17019

Compressed version in the thread below: