IMO academia creates high-quality data and novel bespoke ML architectures, but often doesn't train for enough time and doesn't scale them enough. So big tech can "innovate" by scaling alone, and the field ends up using ESM-IF instead of GVP-GNN and ProGen2 instead of ProtGPT2

New research published! I wanted to write about this, because I think the DE-Score *might* be the thing I'm proudest of so far! All my career I have dreamt of finding something so clean. The DE-Score is a way to tell when amino acid data stops being informative. 1/7 academic.oup.com/sysbio/advan...

Does anyone with familiarity with Claude Skills have a sense of how to reliably test these things? I've seen it succeed by peeking at the code instead of just running the skills as-is, which defeats the purpose of managing the context window

Training antibody language models on evolutionary trajectories inferred from ancestral sequence reconstruction and NGS data, rather than a standard masked objective, leads to improved prediction of fitness and expression

There's an interesting parallel between the in silico affinity maturation presented in the PPIFlow paper, which turns the antibody design problem into a motif scaffolding problem, and what the adaptive immune system does with the so-called germline-encoded GRAB motifs www.biorxiv.org/content/10.6...

Potential fork of PPIFlow with a more permissive license is very exciting: github.com/Mingchenchen... Highly recommend their paper: results are super impressive and it's *not* an AF3 clone

Most benchmark papers are boring but this is a masterpiece. Authors present a comprehensive breakdown of different inference strategies on variant effect prediction (VEP) in viruses. There were two surprising results that IMO deserve more attention🧵

The Twitter algorithm is completely broken but occasionally some gold slips through

New Preprint!! We show that binding entropy can be quantitatively predicted from crystallographic ensemble models, accounting for both protein conformational entropy and solvent entropy! www.biorxiv.org/content/10.6...

Anyone have access to the PPIFlow paper? It’s only hosted on Dropbox for some reason

Is there a Bluesky feed that’s just nothing but random good news? People announcing that they just got married, had a kid, are officially cancer-free, lost 50 pounds etc.

This was fun work and a remarkable effort across the computational and wet-lab teams! Strategies for in-silico filtering and ranking of antibody designs have been under-discussed in the literature, e.g. in most technical reports on antibody design that I've seen. Let's talk about these here! [1/n]

It's exciting to see a de novo antibody design method from industry finally get released with a Methods section!

Random infuriating thing of the day. PDBfixer can't be ported over to pip because the name 'pdbfixer' is blocked by PyPI admins for reasons that nobody seems to be able to articulate, preventing packages using those fxns from being fully conda-free

"We show that this GNN potential, trained on effective energies from ESM3, is robust enough to drive stable, long-timescale molecular dynamics simulations." What on earth? arxiv.org/abs/2601.05388

Done! Can be installed with `pip install graphrelax`

Only have 8 days left on my Claude Max plan so it's time to make the LigandMPNN/OpenMM-based replacement for Rosetta Relax that, for some reason, still doesn't exist

My hot take is that CIF and PDB formats are totally ill-equipped for adequately storing metadata describing how structures get generated and processed, and this problem will get worse as AI agents imbued with structural bioinformatics skills pass these files back and forth

Some more evidence that BioEmu et al have largely rendered clever AF2 template and MSA hacks obsolete for quick conformational sampling of soluble monomeric proteins

Turns out that if miniprotein designs against a specific epitope fail, an effective fallback plan is to design against an adjacent and more tractable surface nearby, and then “grow” or “bulk up” your experimentally validated hits into the epitope you’re interested in www.biorxiv.org/content/10.6...

Please give me your best .pymolrc.py functions. Here's one that implements "fetch_imgt <pdb_file>" and "fetch_imgt <pdb_file>_<chain_id>" for fetching renumbered antibody chains from sabdab

“although current methods reveal the presence of heterogeneity, their stochasticity and potential bias present challenges for their routine use” the methods used include cryoDRGN and Bayesian ensemble reweighting, among others

wtf

So 2025 turned out to be a big year for de novo antibody design! Here are thoughts and predictions on the state of de novo antibody design heading into 2026 🧵

MFW you have an awesome idea and need to pull up a random biorxiv paper for critical details

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?

New latent labs report on de novo antibody design closely follows what we’ve seen from Chai and Nabla. Promising results, and I respect that they admit that half of the targets were unsuccessful with their approach www.latentlabs.com/wp-content/t...

Here is Claude Opus's attempt at one-shot refactoring the PyTorch implementation of ProteinMPNN. Impressive work, but it could be better github.com/delalamo/Pro...

Folding funnels are back - now we just need to convince scientists to resume using P_near, a funnel-ness metric that deserved far more use than it received

has anyone found a workaround for dealing with unrealistically overstructured loops being generated by Boltz/chai/etc? It’s an annoying problem that forces me to default to prior-gen NNs for a lot of human proteins

Happy to share our cryoEM Ensemble Optimization method! 🎉It infers optimal structures and their weights directly from individual cryoEM images - no maps just physics! with David Sliva-Sánchez, @erikhthiede.bsky.social & Roy Lederman www.biorxiv.org/content/10.6...

Another advertisement for an antibody design company masquerading as a paper with opaque, non-reproducible methodology

A randomly valuable thing that the simplefold authors did was fine-tune all six structure prediction models on MD-ATLAS data, and honestly I can't tell if there's any benefit to scaling these things if the goal is ensemble prediction

Has this idea of sampling neighbors from an inverse cubic distribution been used in protein structure GNNs at all since the idea was first presented in the Chroma paper ~3 years ago? I thought it was incredibly clever and expected it to catch on, but haven't seen it since

Biomolecular diffusion models aren't able to reproduce the equilibrium dynamics of the simulations they're trained on arxiv.org/pdf/2506.17139

Very cool paper that samples unique conformations by adding a guidance term to diffusion-based protein structure prediction to steer away from already-sampled states. Important caveat: doesn't work with mAbs because it only looks at helices and sheets openreview.net/pdf/dd8aa97e...

An absolutely fascinating edge case from BoltzGen: apparently it loves to design ubiquitins, but only when the prompted sequence is 73-76 AAs long! Reminds of a method from Possu Huang's group (forget which) that designed TIM barrels with lots of enthusiasm

Why do inverse folding methods hate aromatics so much? From the BoltzGen paper

Boltz is now a separate institution? ...

Excited to release BoltzGen which brings SOTA folding performance to binder design! The best part of this project is collaborating with a broad network of leading wetlabs that test BoltzGen at an unprecedented scale, showing success on many novel targets and pushing the model to its limits!

A class like no other!! From #AI structure hallucination 🤖 to #CryoEM structure reality 🔬 by @biozentrum.unibas.ch @unibas.ch undergraduate students👩‍🔬👩‍🎓 in just a few weeks! One of their creations is this beautiful flower-shaped tetrameric pore 🌼, a brand new member of the #ProteinCosmos 🧶🧬 🧪

Hmmmm..... wonder what @opig.stats.ox.ac.uk is up to... github.com/oxpig/HeavyB...

Where can I find thousands of bovine antibody sequences with ultra-long CDR3s? Preferably paired 🧶🧬

Love this preprint and glad to see the field converge on a specific, highly descriptive term for this phenomenon

New job! Senior scientist in AI/ML at Takeda. The plan is to do some fun stuff with biomolecular foundation models

If you've been including PDB 9DMQ chain F with ANARCI/AbNum-derived numbering in your training set for learning something like CDR design, then you'll be surprised to learn that the CDRH3 is only 7 AAs long and encompasses exclusively the region in green opig.stats.ox.ac.uk/webapps/sabd...

Does anyone know how IMGT numbers residues in the DE loop (residues 80-84)? I can't figure out the pattern, seems to be more or less entirely random and not following the logic of CDR1-3

The entire J-gene of PDB 7czv (3.3 Å, shown in salmon) seems to have fallen victim to a register/misassignment error, with the C-terminal valines (right) and tryptophan (center-left) on the wrong side of the beta sheet compared to where they would be expected to be

I've put together a tool to renumber antibody structures using only main chain atomic coords. No sequence info is used at all. In the process I found a few numbering mistakes in Sabdab/PDB, which is numbered by sequence & thus sensitive to gaps/register errors (deleted prev post due to bugs)