July 12, 2025 at 1:52 PM
AF3 kevin redesign i guess
January 13, 2025 at 5:09 PM
AF3 kevin redesign i guess
ColabFold 1.6.3 is out! 2.5x faster, pip-installable, ipSAE+pDockQ2 scores. Thanks Choonghwan Lee, Marielle Russo, Gyuri Kim
🐍pip install colabfold[alphafold]
CF2 Sneak Peak with AF3/Boltz/Protenix/ESMFold2…
🐍pip install "colabfold[alphafold3]@git+https://github.com/sokrypton/ColabFold@af3-preview"
🐍pip install colabfold[alphafold]
CF2 Sneak Peak with AF3/Boltz/Protenix/ESMFold2…
🐍pip install "colabfold[alphafold3]@git+https://github.com/sokrypton/ColabFold@af3-preview"
September 16, 2026 at 5:14 PM
ColabFold 1.6.3 is out! 2.5x faster, pip-installable, ipSAE+pDockQ2 scores. Thanks Choonghwan Lee, Marielle Russo, Gyuri Kim
🐍pip install colabfold[alphafold]
CF2 Sneak Peak with AF3/Boltz/Protenix/ESMFold2…
🐍pip install "colabfold[alphafold3]@git+https://github.com/sokrypton/ColabFold@af3-preview"
🐍pip install colabfold[alphafold]
CF2 Sneak Peak with AF3/Boltz/Protenix/ESMFold2…
🐍pip install "colabfold[alphafold3]@git+https://github.com/sokrypton/ColabFold@af3-preview"
OpenFold3-preview (OF3p) is out: a sneak peek of our AF3-based structure prediction model. Our aim for OF3 is full AF3-parity for every modality. We now believe we have a clear path towards this goal and are releasing OF3p to enable building in the OF3 ecosystem. More👇
October 28, 2025 at 6:30 PM
OpenFold3-preview (OF3p) is out: a sneak peek of our AF3-based structure prediction model. Our aim for OF3 is full AF3-parity for every modality. We now believe we have a clear path towards this goal and are releasing OF3p to enable building in the OF3 ecosystem. More👇
"Hey Bob, just dock the AF3 prediction into density and deposit"
August 7, 2025 at 5:04 PM
"Hey Bob, just dock the AF3 prediction into density and deposit"
A few py2Dmol updates 🧬
py2dmol.solab.org
Integration with AlphaFoldDB (will auto fetch results). Drag and drop results from AF3-server or ColabFold for interactive experience! (1/4)
py2dmol.solab.org
Integration with AlphaFoldDB (will auto fetch results). Drag and drop results from AF3-server or ColabFold for interactive experience! (1/4)
November 19, 2025 at 8:15 AM
A few py2Dmol updates 🧬
py2dmol.solab.org
Integration with AlphaFoldDB (will auto fetch results). Drag and drop results from AF3-server or ColabFold for interactive experience! (1/4)
py2dmol.solab.org
Integration with AlphaFoldDB (will auto fetch results). Drag and drop results from AF3-server or ColabFold for interactive experience! (1/4)
July 12, 2025 at 11:12 AM
This is the first fully open source release of a AF3 like model. Awesome! Congratulations to the team.
Thrilled to announce Boltz-1, the first open-source and commercially available model to achieve AlphaFold3-level accuracy on biomolecular structure prediction! An exciting collaboration with Jeremy, Saro, and an amazing team at MIT and Genesis Therapeutics. A thread!
November 17, 2024 at 5:13 PM
This is the first fully open source release of a AF3 like model. Awesome! Congratulations to the team.
New OpenFold3 preview out! (OF3p2)
It closes the gap to AlphaFold3 for most modalities.
Most critically, we're releasing everything, including training sets & configs, making OF3p2 the only current AF3-based model that is functionally trainable & reproducible from scratch🧵1/9
It closes the gap to AlphaFold3 for most modalities.
Most critically, we're releasing everything, including training sets & configs, making OF3p2 the only current AF3-based model that is functionally trainable & reproducible from scratch🧵1/9
March 13, 2026 at 3:00 PM
New OpenFold3 preview out! (OF3p2)
It closes the gap to AlphaFold3 for most modalities.
Most critically, we're releasing everything, including training sets & configs, making OF3p2 the only current AF3-based model that is functionally trainable & reproducible from scratch🧵1/9
It closes the gap to AlphaFold3 for most modalities.
Most critically, we're releasing everything, including training sets & configs, making OF3p2 the only current AF3-based model that is functionally trainable & reproducible from scratch🧵1/9
AF3 Inspector: In-Browser 3D Model Visualization and Confidence Analytics for AlphaFold 3 https://www.biorxiv.org/content/10.64898/2026.09.21.753240v1
September 28, 2026 at 8:47 PM
AF3 Inspector: In-Browser 3D Model Visualization and Confidence Analytics for AlphaFold 3 https://www.biorxiv.org/content/10.64898/2026.09.21.753240v1
July 31, 2026 at 7:49 AM
AlphaFold3 still tends to overfit regions that AF2 multimer predicts to be disordered. Almost all disordered regions become helices in AF3.
December 15, 2024 at 2:03 AM
AlphaFold3 still tends to overfit regions that AF2 multimer predicts to be disordered. Almost all disordered regions become helices in AF3.
AF3 BEST METHOD followed by cluspro but also some conversion errors
December 3, 2024 at 5:07 PM
AF3 BEST METHOD followed by cluspro but also some conversion errors
It is indeed striking how much AF3-based methods struggle to generalize to novel protein-ligand interfaces. It is also interesting that there remains a meaningful delta between AF3 and other reproductions at the most difficult generalization task (~25% vs ~13% success rate, so roughly double).
In this work, we explored how training data similarity impacts protein-ligand prediction accuracy—an overlooked aspect in recent benchmarks. Our analysis shows that the current co-folding methods struggle to generalize beyond ligand poses in their training data.(2/n)
February 8, 2025 at 2:08 PM
It is indeed striking how much AF3-based methods struggle to generalize to novel protein-ligand interfaces. It is also interesting that there remains a meaningful delta between AF3 and other reproductions at the most difficult generalization task (~25% vs ~13% success rate, so roughly double).
Left: Experimentally validated PPIs
Right: AF3-predicted PPIs
Coevolution is all we've got
Source: threadreaderapp.com/thread/18564...
Right: AF3-predicted PPIs
Coevolution is all we've got
Source: threadreaderapp.com/thread/18564...
November 13, 2024 at 12:37 PM
Left: Experimentally validated PPIs
Right: AF3-predicted PPIs
Coevolution is all we've got
Source: threadreaderapp.com/thread/18564...
Right: AF3-predicted PPIs
Coevolution is all we've got
Source: threadreaderapp.com/thread/18564...
Its 2025 and two atoms STILL can exist in the same space according to AF3
June 3, 2025 at 1:33 PM
Its 2025 and two atoms STILL can exist in the same space according to AF3
Former German-German border crossing Helmstedt/Marienborn on film
📷Nikon AF3
🎞️Kodak Gold 200
#kodak #nikon #analog #analogphotography #filmphotography #35mm #filmisnotdead
📷Nikon AF3
🎞️Kodak Gold 200
#kodak #nikon #analog #analogphotography #filmphotography #35mm #filmisnotdead
July 6, 2025 at 6:18 AM
Former German-German border crossing Helmstedt/Marienborn on film
📷Nikon AF3
🎞️Kodak Gold 200
#kodak #nikon #analog #analogphotography #filmphotography #35mm #filmisnotdead
📷Nikon AF3
🎞️Kodak Gold 200
#kodak #nikon #analog #analogphotography #filmphotography #35mm #filmisnotdead
Stay tuned! Will compare MR for #Xray to compare AF2 and AF3 results. This is due to AF3 models after refinement have an insane level of clashes.
April 10, 2025 at 6:56 AM
Stay tuned! Will compare MR for #Xray to compare AF2 and AF3 results. This is due to AF3 models after refinement have an insane level of clashes.
We have to appreciate the authors for sharing that their bespoke structure predictor wasn't effective as a filter, in that antibodies filtered by whether they were predicted to bind in the design arrangement were no more likely to bind than unfiltered antibodies. They suggest AF3 going forward
March 3, 2025 at 9:22 AM
We have to appreciate the authors for sharing that their bespoke structure predictor wasn't effective as a filter, in that antibodies filtered by whether they were predicted to bind in the design arrangement were no more likely to bind than unfiltered antibodies. They suggest AF3 going forward
AF3 has clearly not solved the antibody-antigen prediction problem seeing the CASP16 results…
December 2, 2024 at 2:59 PM
AF3 has clearly not solved the antibody-antigen prediction problem seeing the CASP16 results…
Introducing highly experimental localfold.org
Building on @martinsteinegger.bsky.social af2 webgpu port, @milot.bsky.social optimizations & jax ports of af3-like models by @marielle.bsky.social, Choonghwan Lee, Julia Buhmann.
WARNING: runs directly on your 💻, may drain 🪫 & eat data📱 & overheat 🔥💻
Building on @martinsteinegger.bsky.social af2 webgpu port, @milot.bsky.social optimizations & jax ports of af3-like models by @marielle.bsky.social, Choonghwan Lee, Julia Buhmann.
WARNING: runs directly on your 💻, may drain 🪫 & eat data📱 & overheat 🔥💻
September 17, 2026 at 3:12 AM
Introducing highly experimental localfold.org
Building on @martinsteinegger.bsky.social af2 webgpu port, @milot.bsky.social optimizations & jax ports of af3-like models by @marielle.bsky.social, Choonghwan Lee, Julia Buhmann.
WARNING: runs directly on your 💻, may drain 🪫 & eat data📱 & overheat 🔥💻
Building on @martinsteinegger.bsky.social af2 webgpu port, @milot.bsky.social optimizations & jax ports of af3-like models by @marielle.bsky.social, Choonghwan Lee, Julia Buhmann.
WARNING: runs directly on your 💻, may drain 🪫 & eat data📱 & overheat 🔥💻
Adding fatty acids is indeed a great AF3 trick for many membrane proteins
Amazingly, this gave us a high-confidence transmembrane channel for both mouse and human NS3, with oleic acids forming a bilayer-like pattern.
Our group now call this the “fatty acid hack.”
The #alphafold predictions with and without fatty acids shown below 👇
Our group now call this the “fatty acid hack.”
The #alphafold predictions with and without fatty acids shown below 👇
June 18, 2025 at 8:32 PM
Adding fatty acids is indeed a great AF3 trick for many membrane proteins
An AF3-like model available with an MIT license
“Our model and code are released under MIT License, and can be freely used for both academic and commercial purposes.”
“Our model and code are released under MIT License, and can be freely used for both academic and commercial purposes.”
Thrilled to announce Boltz-1, the first open-source and commercially available model to achieve AlphaFold3-level accuracy on biomolecular structure prediction! An exciting collaboration with Jeremy, Saro, and an amazing team at MIT and Genesis Therapeutics. A thread!
November 17, 2024 at 7:52 PM
An AF3-like model available with an MIT license
“Our model and code are released under MIT License, and can be freely used for both academic and commercial purposes.”
“Our model and code are released under MIT License, and can be freely used for both academic and commercial purposes.”
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?
December 19, 2025 at 2:24 PM
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?