Why do protein-function AI models look great on benchmarks but disappoint in practice? Often "leakage": proteins meant for testing slip into training data over time, inflating scores. PLUG is a tool to build cleaner training sets. Read Clay's work: diffuse.science/post/plug-be... #proteinML

Unlocking protein dynamics requires infrastructure we all can access. ICYMI: all of our software packages live on GitHub, updated in the open: github.com/diff-use

Open science is useful science. One effort we are developing are logbooks where we share research as it happens: what works, what doesn't, what we've learned. Led by DiffUSE Scientist Steve Meisburger (@meisborg.bsky.social). See the latest: diffuse.science/logbook/beam... #OpenScience

What's happening in protein dynamics research right now? Jaime & Stephanie break down NIH policy changes, preprints & the conformational ensembles conference. New episode up: diffuse.science/posts/TTPD-15/ #StructuralBiology

Bulk solvent fills ~half a protein crystal. Modeling this is required for a good fit to experimental data. How best to model this remains an open question. New from diffUSE: a Phenix patch to supply your own solvent map (MD, 3D-RISM, EDEN, neural nets). Read more here: thestacks.org/publications...

We share science at The DiffUSE Project in a few ways: blogs to contextualize motivation and intuition, scholarly outputs for methods, data, code, and a citable DOI. All open and shared rapidly for the best science. Our latest: doi.org/10.82153/jkx...

We present sampleworks, an open, modular platform for generating and evaluating biomolecular conformational ensembles from structure predictors guided by experimental data doi.org/10.82153/jkx....

Want to shape the future of structural biology? We are seeking a Project Manager to help lead the DiffUSE Project, guiding ML researchers, computational scientists, national labs, and university partners. Apply: buff.ly/gZlj9AP #StructuralBiology #MachineLearning #Hiring #AcademicSky

Simulations can be valuable but there is a huge need for scalable experiments that probe protein dynamics- of course we're expanding our work here and there's also great work from others @hkws.bsky.social @ginaelnesr.bsky.social Basile Wicky (not on bsky) @diffuseproject.bsky.social

The DiffUSE Project wants examples where structural heterogeneity matters but cannot be encoded in PDBx/mmCIF. X-ray, cryo-EM, time-resolved, multi-map, or fragment screening data welcome. Submit yours: diffuse.science/submit/

ICYMI: The latest Tortured Proteins Department episode. @stephanieaw.bsky.social and @fraserlab.com on: NIH strategic plan comments (due 5/26), AI lab guidelines, two new cryo-EM preprints, training scientists for what's next diffuse.science/posts/TTPD-14/

🎉🎉🎉

Speakers include: @nozomi-ando.bsky.social @hkws.bsky.social, @pilarcossio.bsky.social, @cjjackson.bsky.social, @karsonchrispens.bsky.social, @alisiafadini.bsky.social, @thompson-lab.bsky.social, Joey Davis, Alex Bronstein, Rohith Krishna, Matthew McPartlon, Frederic Poitevin

Hosted in partnership with the DiffUSE Project! 🎉👇

Congrats to DiffUSE lead Stephanie Wankowicz @stephanieaw.bsky.social, on winning the inaugural Sheldrick Award! Her work on entropic flexibility and reproducibility in structural biology is core to what we're building. #crystallography #structuralbiology #openscience aucryst.org/sheldrick-aw...

We observed a tradeoff: weak guidance improves fit to data without wrecking protein geometry. But strong guidance destroys the geometry. Interestingly, models differ a lot in how they handle this despite similar architectures.

Excited to share Sampleworks led by @karsonchrispens.bsky.social with @diffuseproject.bsky.social. It's a modular framework connecting structure predictors to experimental data and guidance methods. Swap predictors or guidance methods. All open and made to be built upon. t.co/fvNHH3gfVu

Radial is live! A new organization at @asterainstitute.bsky.social bringing together structural biologists, engineers, and ML scientists to redesign how we do science. @statnews.com has the story: www.statnews.com/2026/03/11/r...

MD simulations show that diffuse scattering patterns shift depending on whether ligand or a buffer molecule occupies the binding pocket. The differences are quantitatively significant, a promising sign that experiments could detect these changes. Read more: diffuse.science/post/shake-i...

Tomorrow (2/27) 3 pm ET, DiffUSE Project Scientist Steve Meisburger presents "Diffuse Scattering Wars (Episode IV): A New Hope for Structural Biology" Learn how diffuse scattering could reveal what AI structure predictors are missing. Join here: www.chess.cornell.edu/science/ches...

The Diffuse Project is looking for an amazing Project Manager to help us achieve our goal of unlocking protein dynamics! Are you a scientist who naturally loves organization, coordination, problem solving and working with a diverse team of scientists & engineers?

Happy @diffuseproject.bsky.social launch day! Check out our latest effort. We're grappling with how to better understand protein dynamics and capture/share the information that would best allow the community to make use of that data. Take a look!

What’s the next PDB? Maybe it’s …the PDB. We just launched @diffuseproject.bsky.social a structural biology initiative exploring protein motion & rethinking how we generate and use experimental data. Quick 🧵👇 www.diffuse.science

Announcing the Diffuse Project! We're unlocking protein dynamics through diffuse X-ray scattering - the overlooked signal that could revolutionize how we understand protein motion. seemay.substack.com/p/from-syste...