From the article:
\n\n\nGiven an input list of molecules, AlphaFold 3 generates their joint 3D structure, revealing how they all fit together. It models large biomolecules such as proteins, DNA and RNA, as well as small molecules, also known as ligands — a category encompassing many drugs.
\n
and
\n\n", "messages": [ { "id": "2802097", "postDate": "05/08/2024 21:28:54", "content": "\nAlphaFold 3 achieves unprecedented accuracy in predicting drug-like interactions, including the binding of proteins with ligands and antibodies with their target proteins.
\n
From the article:
\n\n\nGiven an input list of molecules, AlphaFold 3 generates their joint 3D structure, revealing how they all fit together. It models large biomolecules such as proteins, DNA and RNA, as well as small molecules, also known as ligands — a category encompassing many drugs.
\n
and
\n\n", "rawMarkdown": "[It was released today](https://blog.google/technology/ai/google-deepmind-isomorphic-alphafold-3-ai-model/#life-molecules)\n\nFrom the article:\n\n>Given an input list of molecules, AlphaFold 3 generates their joint 3D structure, revealing how they all fit together. It models large biomolecules such as proteins, DNA and RNA, as well as small molecules, also known as ligands — a category encompassing many drugs. \n\nand\n\n>AlphaFold 3 achieves unprecedented accuracy in predicting drug-like interactions, including the binding of proteins with ligands and antibodies with their target proteins.", "votes": null }, { "id": "2802109", "postDate": "05/08/2024 21:43:34", "content": "AlphaFold 3 achieves unprecedented accuracy in predicting drug-like interactions, including the binding of proteins with ligands and antibodies with their target proteins.
\n
I’m also curious to know. Here’s the link to use it.
\nBut I am not sure if it can be used in this competition as it is just available for non-commercial use:
\n“AlphaFold Server is for non-commercial use only, subject to AlphaFold Server Terms of Service”
I had the same thought when reading this, but looking at the server it says the predictions are limited to \"biologically common ligands\" and to input you need to \"select the desired entity from the list of supported types\" - so I don't think they model the types of molecules we need to predict here?
", "rawMarkdown": "I had the same thought when reading this, but looking at the server it says the predictions are limited to \"biologically common ligands\" and to input you need to \"select the desired entity from the list of supported types\" - so I don't think they model the types of molecules we need to predict here?", "votes": null }, { "id": "2802283", "postDate": "05/09/2024 01:00:02", "content": "\"I am not sure if it can be used in this competition as it is just available for non-commercial use:\"
\nwe have to prove that it works or does not work first.
\nassume it works, then we know that 3d protein structure is important. then we can think of other methods that can work the same way and replace AlphaFold
I think the key question is whether AF3 can predict bonding probability or affinity values, not just bonding posture
", "rawMarkdown": "I think the key question is whether AF3 can predict bonding probability or affinity values, not just bonding posture", "votes": null }, { "id": "2802497", "postDate": "05/09/2024 04:48:19", "content": "If it works, DeepMind/other research institution might be able to support some teams😃, as AF2's success is tied to its marketing strategy.
", "rawMarkdown": "If it works, DeepMind/other research institution might be able to support some teams😃, as AF2's success is tied to its marketing strategy.", "votes": null }, { "id": "2802759", "postDate": "05/09/2024 06:56:00", "content": "Yep, looking at the server we have some ~20 pre-defined ligands to choose from.
\nTo be fair though, I have not explored it too much
", "rawMarkdown": "Yep, looking at the server we have some ~20 pre-defined ligands to choose from.\n\nTo be fair though, I have not explored it too much", "votes": null }, { "id": "2803252", "postDate": "05/09/2024 12:00:09", "content": "In principle, if you can determine the correct pose with one piece of software, you can reasonably do your post-dock scoring with another.
\nI expect that at least someone is trying to apply docking to this competition.
", "rawMarkdown": "In principle, if you can determine the correct pose with one piece of software, you can reasonably do your post-dock scoring with another.\n\nI expect that at least someone is trying to apply docking to this competition.", "votes": null }, { "id": "2811229", "postDate": "05/13/2024 16:56:39", "content": "Can it actually do multiple ligands (SMILES strings)?
", "rawMarkdown": "Can it actually do multiple ligands (SMILES strings)?", "votes": null } ], "comments": [ { "id": 2802109, "author_name": "vicentjborrs", "author_url": "", "post_date": "05/08/2024 21:43:34", "content": "I’m also curious to know. Here’s the link to use it.
\nBut I am not sure if it can be used in this competition as it is just available for non-commercial use:
\n“AlphaFold Server is for non-commercial use only, subject to AlphaFold Server Terms of Service”
\"I am not sure if it can be used in this competition as it is just available for non-commercial use:\"
\nwe have to prove that it works or does not work first.
\nassume it works, then we know that 3d protein structure is important. then we can think of other methods that can work the same way and replace AlphaFold
If it works, DeepMind/other research institution might be able to support some teams😃, as AF2's success is tied to its marketing strategy.
", "votes": null, "replies": [] } ] } ] }, { "id": 2802117, "author_name": "thedrcat", "author_url": "", "post_date": "05/08/2024 21:47:54", "content": "I had the same thought when reading this, but looking at the server it says the predictions are limited to \"biologically common ligands\" and to input you need to \"select the desired entity from the list of supported types\" - so I don't think they model the types of molecules we need to predict here?
", "votes": null, "replies": [ { "id": 2802759, "author_name": "narsil", "author_url": "", "post_date": "05/09/2024 06:56:00", "content": "Yep, looking at the server we have some ~20 pre-defined ligands to choose from.
\nTo be fair though, I have not explored it too much
", "votes": null, "replies": [] } ] }, { "id": 2802489, "author_name": "zuolongzhang", "author_url": "", "post_date": "05/09/2024 04:41:17", "content": "I think the key question is whether AF3 can predict bonding probability or affinity values, not just bonding posture
", "votes": null, "replies": [ { "id": 2803252, "author_name": "jbomitchell", "author_url": "", "post_date": "05/09/2024 12:00:09", "content": "In principle, if you can determine the correct pose with one piece of software, you can reasonably do your post-dock scoring with another.
\nI expect that at least someone is trying to apply docking to this competition.
", "votes": null, "replies": [] } ] }, { "id": 2811229, "author_name": "diallobakary4", "author_url": "", "post_date": "05/13/2024 16:56:39", "content": "Can it actually do multiple ligands (SMILES strings)?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2802097": "[It was released today](https://blog.google/technology/ai/google-deepmind-isomorphic-alphafold-3-ai-model/#life-molecules)\n\nFrom the article:\n\n>Given an input list of molecules, AlphaFold 3 generates their joint 3D structure, revealing how they all fit together. It models large biomolecules such as proteins, DNA and RNA, as well as small molecules, also known as ligands — a category encompassing many drugs. \n\nand\n\n>AlphaFold 3 achieves unprecedented accuracy in predicting drug-like interactions, including the binding of proteins with ligands and antibodies with their target proteins.", "2802109": "I’m also curious to know. Here’s the [link](https://golgi.sandbox.google.com/about) to use it.\nBut I am not sure if it can be used in this competition as it is just available for non-commercial use:\n“AlphaFold Server is for non-commercial use only, subject to AlphaFold Server Terms of Service”", "2802117": "I had the same thought when reading this, but looking at the server it says the predictions are limited to \"biologically common ligands\" and to input you need to \"select the desired entity from the list of supported types\" - so I don't think they model the types of molecules we need to predict here?", "2802283": "\"I am not sure if it can be used in this competition as it is just available for non-commercial use:\"\n\nwe have to prove that it works or does not work first.\nassume it works, then we know that 3d protein structure is important. then we can think of other methods that can work the same way and replace AlphaFold", "2802489": "I think the key question is whether AF3 can predict bonding probability or affinity values, not just bonding posture", "2802497": "If it works, DeepMind/other research institution might be able to support some teams😃, as AF2's success is tied to its marketing strategy.", "2802759": "Yep, looking at the server we have some ~20 pre-defined ligands to choose from.\n\nTo be fair though, I have not explored it too much", "2803252": "In principle, if you can determine the correct pose with one piece of software, you can reasonably do your post-dock scoring with another.\n\nI expect that at least someone is trying to apply docking to this competition.", "2811229": "Can it actually do multiple ligands (SMILES strings)?" }, "source": "meta" }