iiinova — where your digital self keeps evolving

Reposted by Simon Dürr

Helena Klara Jambor @helenajambor.bsky.social · 10/09/2026

Very happy about this months #PointsOfView in #NatureMethods - The Graphical Abstract. Co-authored with @simonduerr.eu, with whom I share a long-time passion for #Icons #BioIcons and #CC0 figures - and fuelled by our joint workshop for the @sib.swiss at [BC]2! Here is a 🎁 link lnkd.in/e5ew-E2R

Four ways to use a graphical abstract, as table of content icon, on social media, in talks and on posters.

193

Reposted by Simon Dürr

Gina El Nesr @ginaelnesr.bsky.social · 28/04/2026

Excited to report the first zero-shot designed, de novo enzyme catalyzing two of the most energetically demanding reactions in biology—phosphomonoester and phosphodiester hydrolysis—with catalytic efficiencies comparable to natural enzymes! 🚀 /🧵 @stanfordbiosci.bsky.social @simonduerr.eu

1205

Simon Dürr @handle.invalid · 04/02/2026

We could have this for biology if we hadnt collectively decided to spend ~10 Mio dollars per year on BioRender instead 😢

142

Simon Dürr @handle.invalid · 04/02/2026

SVG is much harder to generate for LLMs and there is much less training data of complex images. Current state of the art is this work: starvector.github.io/rlrf/

starvector.github.io

RLRF: Rendering-Aware Reinforcement Learning

RLRF: Rendering-Aware Reinforcement Learning for Vector Graphics Generation

120

Simon Dürr @handle.invalid · 04/02/2026

Curious to try it but it will probably still be unusable if you care about the things you want to illustrate. There is already some services that wrap NanoBananaPro for Figure generation and even the teaser images they show are wrong: For example grabstract.io

Carbons with 5 bonds in AI generated images.

010

Simon Dürr @handle.invalid · 12/12/2025

Zenodo probably isnt too bad, could also set up a small blog on GitHub pages with Rogue Scholar to get DOI or use OpenScienceFramework for non-preprints that can't be reviewed on Prereview

110

Reposted by Simon Dürr

jproney @jproney.bsky.social · 12/12/2025

As a bonus, here's a video of ProteinEBM folding up the fast-folder NTL9, rendered in stunning 2D by py2Dmol from @sokrypton.org! We hope models like ProteinEBM can serve as a step toward solving the "real" protein folding problem.

1365

Simon Dürr @handle.invalid · 12/11/2025

Yeah it can work really well as www.qedscience.com demonstrates (for wet lab biology)

010

Simon Dürr @handle.invalid · 16/10/2025

Great. Would be nice to have a more direct comparison of the same pipeline on different hardware. If one has to specially optimize each model (TensorRT, BF16) its clearly an argument to stick with an x86 machine

000

Simon Dürr @handle.invalid · 15/10/2025

Yes true. Also it's ARM which might create issues with dependencies. But CUDA ecosystem, large memory, low energy consumption are nice... Maybe, someone from the NVidia Bio Team ( cc @kdidi.bsky.social @machine.learning.bio )can enlighten us if they got their hands on it for testing

100

Simon Dürr @handle.invalid · 15/10/2025

Anyone from the Bio community already got their hands on a NVidia DGX Spark for structure prediction and Protein design workloads ?

332

Simon Dürr @handle.invalid · 01/10/2025

And little sidenote that the versioning is becoming quite a mess: Now there is RFDiffusion,RFDiffusionAA, RFDiffusion2, RFDiffusion3 and RFDiffussion2-MI (last two preprinted weeks from each other)

130

Simon Dürr @handle.invalid · 01/10/2025

They saw it is hard to achieve a specific binder that recognizes only the phosphorylated target and not the unphosphorylated. Sequence specificity was easy. I also find notable that the model doesn't use motifs common in natural proteins like SH2

110

Simon Dürr @handle.invalid · 01/10/2025

Cool work to design phospho-tyrosine binder by the Baker lab. They show it works but success rate is low (<0.1%) and binders are not super strong (>500nM, likely related to cost of desolvating the phosphate) . www.biorxiv.org/content/10.1...

1165

Simon Dürr @handle.invalid · 25/09/2025

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...

0105

Simon Dürr @handle.invalid · 24/09/2025

Interesting work to design antibodies by merging AlphaFold hallucination and an antibody specific language model to design de novo CDR regions in a given antibody framework 🖥️ github.com/SantiagoMill... 🗎 www.biorxiv.org/content/10.1...

030

Reposted by Simon Dürr

adaptyv.bio @adaptyv.bio · 18/09/2025

Today we're dropping the "beta" tag from Adaptyv, launching our new website and announcing our $8M seed round. When we started Adaptyv a few years ago, our core belief was: AI models for biology are only as good as the lab data they're trained on and the hypotheses they can test in the real world.

183

Reposted by Simon Dürr

MLSB (in San Diego + Copenhagen) @workshopmlsb.bsky.social · 17/09/2025

🚨To accommodate the addition of EuroMLSB, we have extended the submission deadline to October 1, 2025 11:59pm AoE. Find information on paper guidelines at mlsb.io. Submissions will be made through CMT.

033

Reposted by Simon Dürr

MLSB (in San Diego + Copenhagen) @workshopmlsb.bsky.social · 12/09/2025

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💫

2106

Reposted by Simon Dürr

Diego del Alamo @delalamo.xyz · 11/09/2025

Time to test their claim that diffusion is competitive with hallucination in generating realistic backbones of huge (>800 AA) proteins...

0112

Simon Dürr @handle.invalid · 11/09/2025

La-Proteina from NVidia to do co-design of fully atomistic protein structures (sequence + side-chains + backbone) for up to 800 residues is now open source with permissive licensing 🖥️ github.com/NVIDIA-Digit... 🗎 arxiv.org/abs/2507.09466

04915

Simon Dürr @handle.invalid · 11/09/2025

Ever since Twitter shut down 3rd party clients I sorely missed a client with "remember where I left off" feature. Finally, found a client for Bluesky to do this: @openvibe.social which also supports RSS.

040

Simon Dürr @handle.invalid · 14/08/2025

Hoping to liberate scientific illustrations from eternal payments to them so that we all can communicate and iterate on ideas faster by building on openly available illustrations.

010

Simon Dürr @handle.invalid · 14/08/2025

It's been quite frustrating to deal with BioRender's unwillingness to support open science (while publicly claiming that they do) and administrators in institutions who don't care while paying them $$$ to continue business as usual (globally >10k Mio $).

110

Simon Dürr @handle.invalid · 14/08/2025

@pracheeac.bsky.social I'm working on creating a free and open alternative to BioRender for scientific illustrations. Many scientists support this effort but I've also encountered quite some fans who don't realize that they're supporting a defacto monopoly.

130

Reposted by Simon Dürr

MLSB (in San Diego + Copenhagen) @workshopmlsb.bsky.social · 28/07/2025

Stay tuned for details on the 6th edition of MLSB, officially happening this December in downtown San Diego, CA!

0157

Reposted by Simon Dürr

Gina El Nesr @ginaelnesr.bsky.social · 07/07/2025

Unfortunately, the MLSB Workshop @ NeurIPS (@workshopmlsb.bsky.social) was rejected this year. Feedback from the deciding committee indicates it was a coin flip decision, with 283 proposals & a number related to “computational biology” More on the future of MLSB soon…

285

Simon Dürr @handle.invalid · 05/05/2025

Exactly. Also for some countries like Germany where people have been complaining about the conditions for research for years (people finishing their thesis with unemployment money and other shenanigans) it's kinda a slap in the face that suddenly there is money to bring in new hires...

000

Simon Dürr @handle.invalid · 03/05/2025

Not as pretty but this should give you a vector you can tune easily in a similar way with reduced color complexity (might be a bit slow for large proteins in contour mode) bioicons.com/pdb2vector/ And uses less CO2 😅

bioicons.com

Bioicons - high quality science illustrations

Submit new Bioicons under permissive CC0, CC BY or MIT License.

170

Simon Dürr @handle.invalid · 03/05/2025

No. The other libraries have different names (e.g RCSB Mol*, PDBe Mol*). Mol* = mol* = molstar

000

Simon Dürr @handle.invalid · 02/05/2025

It would be nice if molviewspec would support trajectories because you could easily set up your scene from python and then just use an iframe viewer. colab.research.google.com/drive/1O2Tld... But seems only pdb/cif and map data so far: molstar.org/mol-view-spe...

colab.research.google.com

Google Colab

110

Simon Dürr @handle.invalid · 02/05/2025

The problem is that there are various libraries packaging mol* into various viewers. I don't think the PDBe one supports xtc files, Mol* supports it though. Here is an example using native mol* taken from the quarto-molstar package. plnkr.co/edit/TjTgopo...

plnkr.co

Molstar XTC

Created on https://plnkr.co: Helping you build the web.

220

Simon Dürr @handle.invalid · 02/05/2025

If you want to customize this viewer it is much more complicated.

110

Simon Dürr @handle.invalid · 02/05/2025

The various flavours of Mol* should be pretty easy to integrate but can be a bit more tricky if you use SSR in a modern webstack. Example: embed.plnkr.co/plunk/eMPsfu... You need to convert the trajectory into e.g a PDB trajectory. Try with a NMR structure like 3GB1.

embed.plnkr.co

PDBe Molstar plugin - Basic - v3.2.0

Created on Plnkr: Helping developers build the web.

150

Simon Dürr @handle.invalid · 22/03/2025

Yes but it requires a certain level of understanding of copyright. So many papers that don't even acknowledge 3rd party material properly... I'm looking for funding to build a fully CC0 db to make this easier... currently we spend ~10-20 Mio $ annually on BioRender globally 😔

020

Simon Dürr @handle.invalid · 19/03/2025

CC BY is great but also requires that we replace some of our tools. Many currently published CC BY papers are actually not fully CC BY because of BioRender www.chemistryworld.com/news/thousan...

chemistryworld.com

Thousands of published studies may contain images with incorrect copyright licences

Questions raised over copyright licence that covers images created using scientific illustration service Biorender

142

Simon Dürr @handle.invalid · 02/03/2025

Might be related to this: bsky.app/profile/raph...

110

Simon Dürr @handle.invalid · 21/02/2025

Installation and running it was a breeze 🤗 Diversity of the generated structures is quite good. Quality is a bit hit and miss. Most samples at default params have at least one or two weird C-N peptide bonds >1.6A/<1.1A and also had a few (<0.5%) samples with quite severe clashes.

010

Simon Dürr @handle.invalid · 17/02/2025

Yes that's true but afaik ZeroGPU might get Docker support as well (although I think there is no timeline). You can also ask them for the old community GPU grant which works with generic docker images but has different set of limitations especially for high traffic applications.

000

Simon Dürr @handle.invalid · 17/02/2025

One can build quite complex interfaces with gr.Blocks() in my opinion. Are you using Blocks? Blocks works basically like flexbox layouts. HF spaces also supports generic Docker containers (just expose a port and you use anything you like huggingface.co/docs/hub/spa...)

huggingface.co

Docker Spaces

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

110

Simon Dürr @handle.invalid · 10/02/2025

In earlier work, we already analyzed how AllMetal3D given a structural input and no information about metal stoichiometry fares against co-folding methods given metal stoichiometry but no structure: doi.org/10.1101/2024...

000

Simon Dürr @handle.invalid · 10/02/2025

We compared to a few other tools that perform either metal identity prediction or location prediction and found that is very important to include negative classes during training and also cluster metal ions by biological similarity.

100

Simon Dürr @handle.invalid · 10/02/2025

We've made available a ChimeraX extension and a Webapp to easily run predictions. AllMetal3D is now also easily installable using pip. ChimeraX: cxtoolshed.rbvi.ucsf.edu/apps/chimera... HuggingFace: huggingface.co/spaces/simon... Docs for Python Package: lcbc-epfl.github.io/allmetal3d/

100

Simon Dürr @handle.invalid · 10/02/2025

Excited to share our new AllMetal3D model building on Metal3D. AllMetal3D predicts the location, identity and geometry of the major biologically relevant metal ions doi.org/10.1101/2025...

1175

Reposted by Simon Dürr

Sri Kosuri @srikosuri.bsky.social · 08/02/2025

It’s been a tough few weeks. My 10yo daughter was diagnosed with a very rare, aggressive cancer called interdigitating dendritic cell sarcoma (IDCS). I’m reaching out to identify clinicians/patients who have encountered pediatric IDCS or other (non-LCH) dendritic or histiocytic sarcomas cases.

841009846

Simon Dürr @handle.invalid · 27/01/2025

Gifs are the least accessible form of interactive content. The best would be interactive HTML figures.

120

Simon Dürr @handle.invalid · 21/01/2025

I've seen cases with open code for the webserver without the actual model or method used for the research published as well ... Code review should be part of peer review.

050