Not new, but a new to us update:
The first preprint out of my lab! We joined forces with @kinasekid.bsky.social @jasonzxzhang.bsky.social and David Baker to study protein phosphorylation! Congrats to Isabella from my lab on her first first author paper! tinyurl.com/43jwwfua
It was such a fun journey working with Krishna’s lab and @kinasekid.bsky.social! Really excited to see where this phospho-binder technology goes!
Thrilled to announce our new preprint, “Protein Hunter: Exploiting Structure Hallucination within Diffusion for Protein Design,” in collaboration with @Griffin, @GBhardwaj8 and @sokrypton.org
🧬Code and notebooks will be released by the end of this week.
🎧Golden- Kpop Demon Hunters
We’d love to hear ideas for where phosphorylation - or PTM-specific binders could make a difference. If you’re interested in collaborating, please reach out! (8/8)
This was only possible in a team effort with great people! Thanks @jasonzxzhang.bsky.social, Kejia Wu, Brian Coventry, Gyu Rie Lee, Kody Klupt, Jiuhan Shi, Rafael Brent and the entire @uwproteindesign.bsky.social, with guidance from David Baker and many collaborators. 🙏 (7/8)
Specificity matters. An all-by-all assay showed strong diagonal binding - each binder recognized only its cognate phosphopeptide, with negligible cross-reactivity. (6/8)
Crystal structures of CD3ε and EGFR complexes matched design models within ~2 Å. The intended phosphate-binding motifs were reproduced with atomic accuracy, validating the approach. (5/8)
We tested four sites: CD3ε (TCR), EGFR pY1068 & pY1173, INSR pY1361. In each case, RFD2-MI produced compact proteins that selectively bound the phosphorylated peptide, confirmed by yeast display & BLI with comparable affinities to native phosphorylation binding domains. (4/8)
We built RFD2-MI, an all-atom diffusion model for molecular interfaces based on RFD2. It co-designs binder + peptide and uses 1D conditioning features like hotspots, secondary structure, and solvent exposure to steer phosphate-pocket formation. (3/8)
Existing tools fall short: antibodies can’t always tell sites apart, and natural domains like SH2 struggle with specificity. We needed a way to design de novo binders that are both phosphorylation- and sequence-specific. (2/8)
Phosphorylation on tyrosines control key pathways in immunity, cancer, and metabolism. For the first time, we can now design proteins that specifically recognize individual phosphotyrosines, even in disordered regions. (1/8)
Preprint: www.biorxiv.org/content/10.1...
(1/7)
Training biomolecular foundation models shouldn't be so hard. And open-source structure prediction is important. So today we're releasing two software packages: AtomWorks and RosettaFold3 (RF3)
[www.biorxiv.org/content/10.1101/2025.08....
Who knew a Nobel Prize win could unlock an entire city? Join us live on YouTube as we celebrate 2024 Nobel Laureate David Baker together with the Mayor of Seattle and many others on March 10th starting at 5 pm (PT)! 🥇🔑🌇
www.youtube.com/live/z8NO4Bg...
A weekend project from a while back -- this little package (with no dependencies) allows you to interact with pymol remotely.
I use it a lot for my protein design workflows together with @biotite.bsky.social.
Just `pip install pymol-remote`
#CompChemSky 🧶🖥️🧬🧪
I wonder, is there any computational approach, that would find protein in PDB according to any arbitrary shape?
Sticker on a laptop saying 'I <3 Biophysics' with all the letters taken form protein structures from the PDB
New sticker design saying 'I <3 Biophysics' with a different heart from SARS-CoV-2 NSP13
I would also be very interested in find some alternative letters, for example a B and P for a new ‘I ❤️ Biophysics’ sticker. Currently its hard to read for some people even with the Howarth and Chroma alphabet. SARS-CoV-2 NSP13 provided an amazing heart btw!
The Praetorius lab for Biomolecular Design at the Institute of Science and Technology Austria (ISTA) is looking for grad students in 2024. If you are interested in protein design at a great institute near Vienna reach out to me!
www.dropbox.com/scl/fi/6iny2...
Update on the Chroma vs RfDiffusion analysis.
ProteinMPNN just doesn't like Chroma's backbones (poor prediction of proteinMPNN generated sequences by ESMFold). Interestingly, Chroma's own sequence design method (which was trained in the context of partially noise backbones) loves it! (1/3)
"Performance and structural coverage of the
latest, in-development AlphaFold model" 🧪🧶🧬
DeepMind & Isomorphic Labs sharing some updates (but no code) on what is presumably alphafold 3, capable of modeling ligands, nucleic acids, antibody-antigen complexes etc
storage.googleapis.com/deepmind-med...
