🧶🧬 We present LMi4Boltz:
www.biorxiv.org/content/10.1...
Boltz-2 is an excellent open source alternative to AlphaFold3. However, high VRAM use restricts modeling large complexes. Using careful memory management, we increase the Boltz-2 size limit by >60% while maintaining execution speed.
www.biorxiv.org/content/10.1...
Boltz-2 is an excellent open source alternative to AlphaFold3. However, high VRAM use restricts modeling large complexes. Using careful memory management, we increase the Boltz-2 size limit by >60% while maintaining execution speed.
October 31, 2025 at 8:54 PM
🧶🧬 We present LMi4Boltz:
www.biorxiv.org/content/10.1...
Boltz-2 is an excellent open source alternative to AlphaFold3. However, high VRAM use restricts modeling large complexes. Using careful memory management, we increase the Boltz-2 size limit by >60% while maintaining execution speed.
www.biorxiv.org/content/10.1...
Boltz-2 is an excellent open source alternative to AlphaFold3. However, high VRAM use restricts modeling large complexes. Using careful memory management, we increase the Boltz-2 size limit by >60% while maintaining execution speed.
Looks like OpenFold3 has been formally released in a public "preview". Not quite on parity with AlphaFold3 on a few benchmarks shown, in particular for antibody interactions. All info on the github link. I am sure we will hear more about this from the developers github.com/aqlaboratory...
GitHub - aqlaboratory/openfold-3: OpenFold3: A fully open source biomolecular structure prediction model based on AlphaFold3
OpenFold3: A fully open source biomolecular structure prediction model based on AlphaFold3 - aqlaboratory/openfold-3
github.com
October 28, 2025 at 9:30 AM
Looks like OpenFold3 has been formally released in a public "preview". Not quite on parity with AlphaFold3 on a few benchmarks shown, in particular for antibody interactions. All info on the github link. I am sure we will hear more about this from the developers github.com/aqlaboratory...
AlphaFold3 co-structure predictions were also consistent with entry when this single mutation was introduced, revealing the likely molecular interaction that enables hACE2 compatibility.
October 28, 2025 at 5:07 AM
AlphaFold3 co-structure predictions were also consistent with entry when this single mutation was introduced, revealing the likely molecular interaction that enables hACE2 compatibility.
Preprint: Experiment-guided AlphaFold3 resolves accurate protein ensembles www.biorxiv.org/content/10.1...
Experiment-guided AlphaFold3 resolves accurate protein ensembles
AlphaFold3 predicts highly accurate protein structures from sequence, but tends to collapse to a single dominant conformation, even when the underlying structure is inherently heterogeneous. Moreover,...
www.biorxiv.org
October 23, 2025 at 3:39 PM
Preprint: Experiment-guided AlphaFold3 resolves accurate protein ensembles www.biorxiv.org/content/10.1...
📢 New preprint:
Experiment-guided AlphaFold3 resolves accurate protein ensembles.
doi.org/10.1101/2025...
AlphaFold3 is incredible, but has crucial limitations: it typically collapses to a single conformation, ignoring the inherent dynamics of proteins. And it can be wrong. Here's a solution. 🧵👇
Experiment-guided AlphaFold3 resolves accurate protein ensembles.
doi.org/10.1101/2025...
AlphaFold3 is incredible, but has crucial limitations: it typically collapses to a single conformation, ignoring the inherent dynamics of proteins. And it can be wrong. Here's a solution. 🧵👇
October 18, 2025 at 6:59 PM
📢 New preprint:
Experiment-guided AlphaFold3 resolves accurate protein ensembles.
doi.org/10.1101/2025...
AlphaFold3 is incredible, but has crucial limitations: it typically collapses to a single conformation, ignoring the inherent dynamics of proteins. And it can be wrong. Here's a solution. 🧵👇
Experiment-guided AlphaFold3 resolves accurate protein ensembles.
doi.org/10.1101/2025...
AlphaFold3 is incredible, but has crucial limitations: it typically collapses to a single conformation, ignoring the inherent dynamics of proteins. And it can be wrong. Here's a solution. 🧵👇
Check out our approach for adding covalent bonds to AlphaFold3, which we apply to ubiquitin. Watch this space!
You may have noticed that AlphaFold3 does not allow covalent bonds between/within polymer chains (protein, DNA, RNA) but allows it for ligands. We tried to go around this limitation by simply treating one of residues/nucleic acids as a ligand.
October 17, 2025 at 12:57 AM
Check out our approach for adding covalent bonds to AlphaFold3, which we apply to ubiquitin. Watch this space!
Dumb question: if you have a lead compound you want to dock to a protein and you're in a non-profit institution not working with a commercial entity, can you (legally) use AlphaFold3 and LATER commercialize the lead compound (or its derivative)?
October 10, 2025 at 7:42 AM
Dumb question: if you have a lead compound you want to dock to a protein and you're in a non-profit institution not working with a commercial entity, can you (legally) use AlphaFold3 and LATER commercialize the lead compound (or its derivative)?
🍭 Glycans are biology’s sugar code—super complex, super important. Thanks to AlphaFold3, we can now model their interactions with proteins! 🧪 Our new workflow unlocks fresh ways to explore immunity, infection & disease.
doi.org/10.1093/glyc...
#BioFGREAT #NSF #glycotime
doi.org/10.1093/glyc...
#BioFGREAT #NSF #glycotime
September 18, 2025 at 5:10 PM
🍭 Glycans are biology’s sugar code—super complex, super important. Thanks to AlphaFold3, we can now model their interactions with proteins! 🧪 Our new workflow unlocks fresh ways to explore immunity, infection & disease.
doi.org/10.1093/glyc...
#BioFGREAT #NSF #glycotime
doi.org/10.1093/glyc...
#BioFGREAT #NSF #glycotime
Anecdotal experience with AlphaFold3 vs AlphaFold2:
The pLDDT, ipTM, and pTM scores are constantly scoring worse than the same input in AF2, but the PAE is better with AF3.
Anyone else have the same experience?
The pLDDT, ipTM, and pTM scores are constantly scoring worse than the same input in AF2, but the PAE is better with AF3.
Anyone else have the same experience?
September 21, 2024 at 3:56 AM
Anecdotal experience with AlphaFold3 vs AlphaFold2:
The pLDDT, ipTM, and pTM scores are constantly scoring worse than the same input in AF2, but the PAE is better with AF3.
Anyone else have the same experience?
The pLDDT, ipTM, and pTM scores are constantly scoring worse than the same input in AF2, but the PAE is better with AF3.
Anyone else have the same experience?
If I get it correctly, they use the CA coordinates of the binding pocket to predict binding. It is still impressive but perhaps not directly comparable to RoseTTAfold-AA/AlphaFold3, as they predict the structure simultaneously and do not know the binding pocket.
November 6, 2023 at 1:27 PM
If I get it correctly, they use the CA coordinates of the binding pocket to predict binding. It is still impressive but perhaps not directly comparable to RoseTTAfold-AA/AlphaFold3, as they predict the structure simultaneously and do not know the binding pocket.
#Google libera #AlphaFold3 para uso científico não comercial. Modelo de #IA melhora previsões de proteínas. Antes restrito, agora é #open-source para acelerar pesquisas. Fonte: #Nature.
November 20, 2024 at 12:02 PM
#Google libera #AlphaFold3 para uso científico não comercial. Modelo de #IA melhora previsões de proteínas. Antes restrito, agora é #open-source para acelerar pesquisas. Fonte: #Nature.
Isomorphic Labs та Google DeepMind спільно представили передову систему штучного інтелекту AlphaFold 3, яка використовує нову архітектуру на основі дифузії для точного моделювання структур різних біомолекулярних комплексів.
#AI #AlphaFold3 #Дослідження #Медицина #ШІ
#AI #AlphaFold3 #Дослідження #Медицина #ШІ
AlphaFold 3 дає поштовх для відкриття нових ліків | TheTransmitted
Isomorphic Labs та Google DeepMind спільно представили передову систему штучного інтелекту AlphaFold 3, яка використовує нову архітектуру на основі дифузії для точного моделювання структур різних біом...
thetransmitted.com
May 9, 2024 at 12:53 PM
Isomorphic Labs та Google DeepMind спільно представили передову систему штучного інтелекту AlphaFold 3, яка використовує нову архітектуру на основі дифузії для точного моделювання структур різних біомолекулярних комплексів.
#AI #AlphaFold3 #Дослідження #Медицина #ШІ
#AI #AlphaFold3 #Дослідження #Медицина #ШІ
Structural biology remains useful!
December 10, 2024 at 1:31 AM
Structural biology remains useful!
📜 AlphaFold3 — why did Nature publish it without its code?
🧑🔬 / 📔 @natureportfolio.bsky.social
🔗 www.nature.com/articles/d41...
With commentary by @mbeisen.bsky.social here: x.com/mbeisen/stat...
#AlphaFold #ProteinStructures #StructuralBiology #ScientificPublishing
🧑🔬 / 📔 @natureportfolio.bsky.social
🔗 www.nature.com/articles/d41...
With commentary by @mbeisen.bsky.social here: x.com/mbeisen/stat...
#AlphaFold #ProteinStructures #StructuralBiology #ScientificPublishing
AlphaFold3 — why did Nature publish it without its code?
Criticism of our decision to publish AlphaFold3 raises important questions. We welcome readers’ views. Criticism of our decision to publish AlphaFold3 raises important questions. We welcome readers’ v...
www.nature.com
May 27, 2024 at 8:50 AM
📜 AlphaFold3 — why did Nature publish it without its code?
🧑🔬 / 📔 @natureportfolio.bsky.social
🔗 www.nature.com/articles/d41...
With commentary by @mbeisen.bsky.social here: x.com/mbeisen/stat...
#AlphaFold #ProteinStructures #StructuralBiology #ScientificPublishing
🧑🔬 / 📔 @natureportfolio.bsky.social
🔗 www.nature.com/articles/d41...
With commentary by @mbeisen.bsky.social here: x.com/mbeisen/stat...
#AlphaFold #ProteinStructures #StructuralBiology #ScientificPublishing
1. We introduce BoltzDesign1, which inverts the Boltz-1 model—an open-source reproduction of AlphaFold3—to enable the design of protein binders for diverse molecular targets without requiring model fine-tuning.
April 8, 2025 at 12:17 PM
1. We introduce BoltzDesign1, which inverts the Boltz-1 model—an open-source reproduction of AlphaFold3—to enable the design of protein binders for diverse molecular targets without requiring model fine-tuning.
💡 Using AlphaFold3, RoseTTAFold All-Atom and Chai-1, we generated more than 14000 models with over 6000 of ligand-related PTMs. We discovered that phosphorylation of NADPH-Cytochrome Reductase leads to substantial structural disruption in the binding pocket, potentially hindering protein function
January 28, 2025 at 6:07 PM
💡 Using AlphaFold3, RoseTTAFold All-Atom and Chai-1, we generated more than 14000 models with over 6000 of ligand-related PTMs. We discovered that phosphorylation of NADPH-Cytochrome Reductase leads to substantial structural disruption in the binding pocket, potentially hindering protein function
Well.... Have to say that what will probably be AlphaFold3 broke me. Highly accurate modeling of protein structures with NA, ligands, any partners.... It is a game changing!
I feel like I have to learn a whole new way of generating hypothesis... A new way of doing research after that. Anyone else?
I feel like I have to learn a whole new way of generating hypothesis... A new way of doing research after that. Anyone else?
November 2, 2023 at 6:16 PM
Well.... Have to say that what will probably be AlphaFold3 broke me. Highly accurate modeling of protein structures with NA, ligands, any partners.... It is a game changing!
I feel like I have to learn a whole new way of generating hypothesis... A new way of doing research after that. Anyone else?
I feel like I have to learn a whole new way of generating hypothesis... A new way of doing research after that. Anyone else?
contrast to the later end-to-end architectures of AlphaFold2 and AlphaFold3. While this potential was originally justified by referring to physical potentials of mean force (PMFs), we reinterpret AlphaFold1's potential as an instance of probability [2/7 of https://arxiv.org/abs/2505.19763v1]
May 27, 2025 at 6:51 AM
contrast to the later end-to-end architectures of AlphaFold2 and AlphaFold3. While this potential was originally justified by referring to physical potentials of mean force (PMFs), we reinterpret AlphaFold1's potential as an instance of probability [2/7 of https://arxiv.org/abs/2505.19763v1]
Accurate Site-specific Folding via Conditional Diffusion Based on Alphafold3 https://www.biorxiv.org/content/10.1101/2025.07.06.663385v1
July 10, 2025 at 9:46 PM
Accurate Site-specific Folding via Conditional Diffusion Based on Alphafold3 https://www.biorxiv.org/content/10.1101/2025.07.06.663385v1
The results show that AlphaFold3 slightly outperforms AlphaFold2 for protein complexes, but the difference disappears when extensive sampling is applied to AlphaFold2. The AF3-server performs better than AlphaFold2 for easier targets, but not for harder ones.
September 23, 2025 at 4:45 PM
The results show that AlphaFold3 slightly outperforms AlphaFold2 for protein complexes, but the difference disappears when extensive sampling is applied to AlphaFold2. The AF3-server performs better than AlphaFold2 for easier targets, but not for harder ones.
Impressively, AlphaFold3 predictions of RBD-ACE2 pairs were very consistent with pseudovirus results! RBDs with ACE2 entry had consistent binding footprints while RBDs without entry were predicted to bind inaccessible parts of the ACE2s.
October 28, 2025 at 5:07 AM
Impressively, AlphaFold3 predictions of RBD-ACE2 pairs were very consistent with pseudovirus results! RBDs with ACE2 entry had consistent binding footprints while RBDs without entry were predicted to bind inaccessible parts of the ACE2s.
Shreya Gopalan, Sundar Narayanan: Hallucinations in AlphaFold3 for Intrinsically Disordered Proteins with disorder in Biological Process Residues https://arxiv.org/abs/2510.15939 https://arxiv.org/pdf/2510.15939 https://arxiv.org/html/2510.15939
October 21, 2025 at 6:50 AM
Shreya Gopalan, Sundar Narayanan: Hallucinations in AlphaFold3 for Intrinsically Disordered Proteins with disorder in Biological Process Residues https://arxiv.org/abs/2510.15939 https://arxiv.org/pdf/2510.15939 https://arxiv.org/html/2510.15939
OpenFold3 Challenges AlphaFold3 in Protein Folding
https://www.byteseu.com/1494026/
A new open-source AI model, OpenFold3, is poised to rival Google DeepMind's AlphaFold3 in predicting 3D protein structures. https://techlife.blog/posts/openfold3-ai-model/
https://www.byteseu.com/1494026/
A new open-source AI model, OpenFold3, is poised to rival Google DeepMind's AlphaFold3 in predicting 3D protein structures. https://techlife.blog/posts/openfold3-ai-model/
OpenFold3 Challenges AlphaFold3 in Protein Folding - Bytes Europe
A new open-source AI model, OpenFold3, is poised to rival Google DeepMind's AlphaFold3 in predicting 3D protein structures.
www.byteseu.com
October 29, 2025 at 2:13 PM
OpenFold3 Challenges AlphaFold3 in Protein Folding
https://www.byteseu.com/1494026/
A new open-source AI model, OpenFold3, is poised to rival Google DeepMind's AlphaFold3 in predicting 3D protein structures. https://techlife.blog/posts/openfold3-ai-model/
https://www.byteseu.com/1494026/
A new open-source AI model, OpenFold3, is poised to rival Google DeepMind's AlphaFold3 in predicting 3D protein structures. https://techlife.blog/posts/openfold3-ai-model/
A new open-source protein structure AI, OpenFold3, aims for parity with AlphaFold3. The OpenFold Consortium has released a preview, inviting feedback. This model is fully open for both academic and commercial use, unlike its competitor. 🧬💻
Open-source protein structure AI aims to match AlphaFold
The developers of OpenFold3 have released an early version of the tool, which they hope will one day perform on par with DeepMind’s protein-structure model.
www.nature.com
October 29, 2025 at 4:15 PM
A new open-source protein structure AI, OpenFold3, aims for parity with AlphaFold3. The OpenFold Consortium has released a preview, inviting feedback. This model is fully open for both academic and commercial use, unlike its competitor. 🧬💻
November 13, 2024 at 6:29 AM