Our most practical episode yet for people who want to dabble in biotech
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
New episode of Hard Drugs!
What if you could design a protein never seen in nature?
Scientists are using new AI tools like RFDiffusion, AlphaFold & ProteinMPNN to hallucinate novel proteins to solve problems nature hasn't.
@jacobtref.bsky.social & I talk about the art of protein design 🧑🎨
What if you could design a protein never seen in nature?
Scientists are using new AI tools like RFDiffusion, AlphaFold & ProteinMPNN to hallucinate novel proteins to solve problems nature hasn't.
@jacobtref.bsky.social & I talk about the art of protein design 🧑🎨
The art of protein design with AI
YouTube video by Works in Progress
www.youtube.com
October 15, 2025 at 5:02 PM
Our most practical episode yet for people who want to dabble in biotech
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
As of 2022, you can hallucinate protein structures using AI similar to Midjourney (RFDiffusion) -> create amino acid strings for them (ProteinMPNN) -> validate with AlphaFold
Biology becomes engineering…
I am thrilled to release ProteinDJ: a high-performance and modular protein design pipeline. Our open-source workflow incorporates #RFdiffusion, #ProteinMPNN, #FAMPNN, #AlphaFold2 and #Boltz-2. It is a fast, free, and fun way to design proteins (1/5)
doi.org/10.1101/2025.09.24.678028 #proteindesign
doi.org/10.1101/2025.09.24.678028 #proteindesign
September 28, 2025 at 9:16 PM
I am thrilled to release ProteinDJ: a high-performance and modular protein design pipeline. Our open-source workflow incorporates #RFdiffusion, #ProteinMPNN, #FAMPNN, #AlphaFold2 and #Boltz-2. It is a fast, free, and fun way to design proteins (1/5)
doi.org/10.1101/2025.09.24.678028 #proteindesign
doi.org/10.1101/2025.09.24.678028 #proteindesign
A failure Ive noticed w RFdiffusion w big proteins is distorted backbone torsions outside normal Ramachandran space, which ProteinMPNN designs as polyglycines that arent accurately refolded. Curious if others have also seen this & if reducing Gly insertion improves designability
October 25, 2024 at 11:34 AM
A failure Ive noticed w RFdiffusion w big proteins is distorted backbone torsions outside normal Ramachandran space, which ProteinMPNN designs as polyglycines that arent accurately refolded. Curious if others have also seen this & if reducing Gly insertion improves designability
Finally read these 2 x papers from the Baker Lab (@uwproteindesign.bsky.social)🧬 De novo protein design for snake venom neutralisation (binders) 🐍 + multistep reactions (enzymes) 🧪 using #RFDiffusion.
nature.com/articles/s41...
science.org/doi/epdf/10.... #proteindesign #stucturalbiology #PhD
nature.com/articles/s41...
science.org/doi/epdf/10.... #proteindesign #stucturalbiology #PhD
De novo designed proteins neutralize lethal snake venom toxins - Nature
Deep learning methods have been used to design proteins that can neutralize the effects of three-finger toxins found in snake venom, which could lead to the development of safer and more accessible an...
nature.com
April 11, 2025 at 1:17 PM
Finally read these 2 x papers from the Baker Lab (@uwproteindesign.bsky.social)🧬 De novo protein design for snake venom neutralisation (binders) 🐍 + multistep reactions (enzymes) 🧪 using #RFDiffusion.
nature.com/articles/s41...
science.org/doi/epdf/10.... #proteindesign #stucturalbiology #PhD
nature.com/articles/s41...
science.org/doi/epdf/10.... #proteindesign #stucturalbiology #PhD
Thanks to @noahholzleitner.bsky.social for the nice animation! And thanks to @dfgpublic.bsky.social and Aventis Foundation for their generous support. Many thanks to Marcela Maus & J. Keith Joung for their help and to David Baker, his lab, and @sokrypton for making RFdiffusion available to all.
November 26, 2024 at 2:14 PM
Thanks to @noahholzleitner.bsky.social for the nice animation! And thanks to @dfgpublic.bsky.social and Aventis Foundation for their generous support. Many thanks to Marcela Maus & J. Keith Joung for their help and to David Baker, his lab, and @sokrypton for making RFdiffusion available to all.
AAAS: "AI could help develop cheaper, faster, and more effective snake antivenoms." Yet today, "venomous snakebites kill upward of 100,000 people every year around the world and maim countless more." Yet antivenoms, the most effective treatments, are still produced much as they were a century ago.
January 16, 2025 at 2:04 PM
AAAS: "AI could help develop cheaper, faster, and more effective snake antivenoms." Yet today, "venomous snakebites kill upward of 100,000 people every year around the world and maim countless more." Yet antivenoms, the most effective treatments, are still produced much as they were a century ago.
RFDiffusion is great for de novo design but it can be a bit frustrating to use. Check our our new GUI for making de novo binder design fast and easy! www.youtube.com/watch?v=rwpL...
Using Levitate Bench for RFDiffusion based de novo Binder Design
YouTube video by Levitate Bio
www.youtube.com
January 29, 2025 at 5:17 PM
RFDiffusion is great for de novo design but it can be a bit frustrating to use. Check our our new GUI for making de novo binder design fast and easy! www.youtube.com/watch?v=rwpL...
“Improved motif-scaffolding with SE(3) flow matching”
Competitive results with RFDiffusion
arxiv.org/abs/2401.04082
github.com/microsoft/fr...
Competitive results with RFDiffusion
arxiv.org/abs/2401.04082
github.com/microsoft/fr...
January 9, 2024 at 6:26 AM
“Improved motif-scaffolding with SE(3) flow matching”
Competitive results with RFDiffusion
arxiv.org/abs/2401.04082
github.com/microsoft/fr...
Competitive results with RFDiffusion
arxiv.org/abs/2401.04082
github.com/microsoft/fr...
Having fun trying protein design
Rosetta: 20 designs -> 0 hits
RFDiffusion/ProteinMPPN: 6 designs -> 4 hits
But maybe the best hit (40 μM Kd) doesn't bind where we intended it to! And maybe the 2nd best hit too. Not a "failure" mode I was expecting to see much
Rosetta: 20 designs -> 0 hits
RFDiffusion/ProteinMPPN: 6 designs -> 4 hits
But maybe the best hit (40 μM Kd) doesn't bind where we intended it to! And maybe the 2nd best hit too. Not a "failure" mode I was expecting to see much
November 26, 2024 at 5:00 PM
Having fun trying protein design
Rosetta: 20 designs -> 0 hits
RFDiffusion/ProteinMPPN: 6 designs -> 4 hits
But maybe the best hit (40 μM Kd) doesn't bind where we intended it to! And maybe the 2nd best hit too. Not a "failure" mode I was expecting to see much
Rosetta: 20 designs -> 0 hits
RFDiffusion/ProteinMPPN: 6 designs -> 4 hits
But maybe the best hit (40 μM Kd) doesn't bind where we intended it to! And maybe the 2nd best hit too. Not a "failure" mode I was expecting to see much
RFdiffusion: Beta Documentation Now Available
We’re excited to announce that RFdiffusion now has official, Rosetta Commons-curated documentation! Read more: rosettacommons.org/2025/07/16/r...
We’re excited to announce that RFdiffusion now has official, Rosetta Commons-curated documentation! Read more: rosettacommons.org/2025/07/16/r...
rosettacommons.org
July 16, 2025 at 3:16 AM
RFdiffusion: Beta Documentation Now Available
We’re excited to announce that RFdiffusion now has official, Rosetta Commons-curated documentation! Read more: rosettacommons.org/2025/07/16/r...
We’re excited to announce that RFdiffusion now has official, Rosetta Commons-curated documentation! Read more: rosettacommons.org/2025/07/16/r...
it is about a new global fund for digital sequence information, and about the need to update our legal ontologies to match biology, which almost certainly is only going to get more complicated with AlphaFold2 and RFdiffusion undark.org/2025/03/05/d...
March 5, 2025 at 4:18 PM
it is about a new global fund for digital sequence information, and about the need to update our legal ontologies to match biology, which almost certainly is only going to get more complicated with AlphaFold2 and RFdiffusion undark.org/2025/03/05/d...
This paper achieves results substantially better than RFDiffusion for de novo antibody design (7 mAbs with Kd < 25nM, good stability / Tm), but by using docking & force fields (no deep learning). Like the RFDiffusion paper, however, success rate is low.
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
www.biorxiv.org
May 3, 2024 at 5:04 PM
This paper achieves results substantially better than RFDiffusion for de novo antibody design (7 mAbs with Kd < 25nM, good stability / Tm), but by using docking & force fields (no deep learning). Like the RFDiffusion paper, however, success rate is low.
www.biorxiv.org/content/10.1...
www.biorxiv.org/content/10.1...
RFdiffusion Exhibits Low Success Rate in De Novo Design of Functional Protein Binders for Biochemical Detection https://www.biorxiv.org/content/10.1101/2025.02.07.636769v1
February 9, 2025 at 3:46 AM
RFdiffusion Exhibits Low Success Rate in De Novo Design of Functional Protein Binders for Biochemical Detection https://www.biorxiv.org/content/10.1101/2025.02.07.636769v1
I would appreciate any advice on analyzing a protein structure using #AlphaFold, Colab, or #RFDiffusion models to identify new interactors. If anyone is interested, I would love to connect via Teams and collaborate. 🧬🖥️ #deeplearning
November 22, 2024 at 8:05 AM
I would appreciate any advice on analyzing a protein structure using #AlphaFold, Colab, or #RFDiffusion models to identify new interactors. If anyone is interested, I would love to connect via Teams and collaborate. 🧬🖥️ #deeplearning
Surprise: A modern baseline (RFdiffusion) fails on motifs scaffolded into de novo enzymes 15 years ago 🤯
This suggests modern deep learning methods aren’t always better than past methods!
This suggests modern deep learning methods aren’t always better than past methods!
February 19, 2025 at 8:50 PM
Surprise: A modern baseline (RFdiffusion) fails on motifs scaffolded into de novo enzymes 15 years ago 🤯
This suggests modern deep learning methods aren’t always better than past methods!
This suggests modern deep learning methods aren’t always better than past methods!
RFdiffusion Exhibits Low Success Rate in De Novo Design of Functional Protein Binders for Biochemical Detection https://www.biorxiv.org/content/10.1101/2025.02.07.636769v1
February 9, 2025 at 3:46 AM
RFdiffusion Exhibits Low Success Rate in De Novo Design of Functional Protein Binders for Biochemical Detection https://www.biorxiv.org/content/10.1101/2025.02.07.636769v1
Excited to announce Vilya’s exclusive license for RFpeptides, a groundbreaking RFdiffusion-based model for macrocyclic peptide design from the UW Institute for Protein Design and School of Pharmacy! RFpeptides enable de novo design of potent, selective macrocycles to target undruggable proteins.
Vilya Announces Exclusive License Agreement of RFdiffusion-based Generative Model from the University of Washington for the Design of Macrocyclic Peptides
– License of proprietary RFdiffusion-based model, RFpeptides, provides Vilya with another powerful structure prediction framework for rapid and unique...
www.globenewswire.com
November 21, 2024 at 6:43 PM
Excited to announce Vilya’s exclusive license for RFpeptides, a groundbreaking RFdiffusion-based model for macrocyclic peptide design from the UW Institute for Protein Design and School of Pharmacy! RFpeptides enable de novo design of potent, selective macrocycles to target undruggable proteins.
Atomically accurate de novo design of antibodies with RFdiffusion
Atomically accurate de novo design of antibodies with RFdiffusion
Nature, Published online: 05 November 2025; doi:10.1038/s41586-025-09721-5 The combination of computational design, laboratory-based screening and biophysical validation enables the de novo generation of variable heavy-chain antibody fragments and antibodies that precisely target chosen disease-related molecules.
www.nature.com
November 5, 2025 at 5:35 PM
Atomically accurate de novo design of antibodies with RFdiffusion
Beautiful to see this out with validation! I've been chatting with Stephen about it for some time and I'm eager to try it. But do I understand correctly that unlike RFDiffusion (BSD licensed), all commercial use of this code will be exclusive to Vilya? www.globenewswire.com/news-release...
Vilya Announces Exclusive License Agreement of RFdiffusion-based Generative Model from the University of Washington for the Design of Macrocyclic Peptides
– License of proprietary RFdiffusion-based model, RFpeptides, provides Vilya with another powerful structure prediction framework for rapid and unique...
www.globenewswire.com
December 2, 2024 at 3:39 AM
Beautiful to see this out with validation! I've been chatting with Stephen about it for some time and I'm eager to try it. But do I understand correctly that unlike RFDiffusion (BSD licensed), all commercial use of this code will be exclusive to Vilya? www.globenewswire.com/news-release...
Can't help but think that anyone outside the field who watched David Bakers's Nobel lecture must have gotten the impression that David & Brian brainstormed about de novo protein design for a bit & the next thing that happened was RFdiffusion came out?! Kind of a bummer for the Rosetta community
December 8, 2024 at 1:02 PM
Can't help but think that anyone outside the field who watched David Bakers's Nobel lecture must have gotten the impression that David & Brian brainstormed about de novo protein design for a bit & the next thing that happened was RFdiffusion came out?! Kind of a bummer for the Rosetta community
Really cool study showing how AI can empower drug discovery! The team used a RFdiffusion-ProteinMPNN-AlphaFold pipeline to design proteins that bind cancer associated peptides w/ high specificity and have the potential to serve as precision immunotherapies
www.science.org/doi/10.1126/...
www.science.org/doi/10.1126/...
De novo-designed pMHC binders facilitate T cell–mediated cytotoxicity toward cancer cells
The recognition of intracellular antigens by CD8+ T cells through T cell receptors (TCRs) is central for adaptive immunity against infections and cancer. However, the identification of TCRs from patie...
www.science.org
July 26, 2025 at 11:58 PM
Really cool study showing how AI can empower drug discovery! The team used a RFdiffusion-ProteinMPNN-AlphaFold pipeline to design proteins that bind cancer associated peptides w/ high specificity and have the potential to serve as precision immunotherapies
www.science.org/doi/10.1126/...
www.science.org/doi/10.1126/...
The deep learning technique that made the latest generation of AI image generators possible has also been used to make an open source novel protein design generator called RFdiffusion
What if all the world's biggest problems have the same invisible solution?
What if all the world's biggest problems have the same solution?
YouTube video by Veritasium
www.youtube.com
February 14, 2025 at 12:21 PM
The deep learning technique that made the latest generation of AI image generators possible has also been used to make an open source novel protein design generator called RFdiffusion
RFdiffusion seems to be incapable of sampling protein topologies that cannot be accurately modeled by its base model RosettaFold. This potential blind spot can (allegedly) be overcome by instead using protein backbone diffusion models trained from scratch www.nature.com/articles/s41...
October 16, 2024 at 12:14 PM
RFdiffusion seems to be incapable of sampling protein topologies that cannot be accurately modeled by its base model RosettaFold. This potential blind spot can (allegedly) be overcome by instead using protein backbone diffusion models trained from scratch www.nature.com/articles/s41...