https://arxiv.org/pdf/2405.01616
\n[paper] GENERATIVE ACTIVE LEARNING FOR THE SEARCH OF SMALL-MOLECULE PROTEIN BINDERS
\nhttps://github.com/pchliu
\nhttps://iclr.cc/virtual/2021/4223
We introduce LAMBDAZERO , a generative active learning approach to search for synthesizable molecule
deep reinforcement learning
We apply LAMBDAZERO with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness.
with only ∼ 1e4 docking simulations, LAMBDAZERO produces synthesizable, drug-like molecules with docking scores that would otherwise require the virtual screening of a hundred billion ( 1e11) molecule
https://arxiv.org/pdf/2405.01616
\n[paper] GENERATIVE ACTIVE LEARNING FOR THE SEARCH OF SMALL-MOLECULE PROTEIN BINDERS
\nhttps://github.com/pchliu
\nhttps://iclr.cc/virtual/2021/4223
We introduce LAMBDAZERO , a generative active learning approach to search for synthesizable molecule
deep reinforcement learning
We apply LAMBDAZERO with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness.
with only ∼ 1e4 docking simulations, LAMBDAZERO produces synthesizable, drug-like molecules with docking scores that would otherwise require the virtual screening of a hundred billion ( 1e11) molecule
Very interesting, thanks for posting! For this competition, obviously we are working with the data we have, which is a very large library of compounds made with a combinatorial chemistry type of approach.
\nThe most relevant aspect of the paper in the limited context of our competition may be the optimisation of what compounds to expend the limited available docking resources on. See also other recent work in that area, which I happened to be reading this week.
\nNonetheless, in a slightly wider context the generative AI search strategy is really interesting, and that's the thing that stands out to me as novel and innovative about this work.
", "rawMarkdown": "Very interesting, thanks for posting! For this competition, obviously we are working with the data we have, which is a very large library of compounds made with a combinatorial chemistry type of approach.\n\nThe most relevant aspect of the paper in the limited context of our competition may be the optimisation of what compounds to expend the limited available docking resources on. See also [other recent work](https://doi.org/10.26434/chemrxiv-2024-qsdd1) in that area, which I happened to be reading this week.\n\nNonetheless, in a slightly wider context the generative AI search strategy is really interesting, and that's the thing that stands out to me as novel and innovative about this work.", "votes": null } ], "comments": [ { "id": 2814670, "author_name": "jbomitchell", "author_url": "", "post_date": "05/15/2024 13:34:56", "content": "Very interesting, thanks for posting! For this competition, obviously we are working with the data we have, which is a very large library of compounds made with a combinatorial chemistry type of approach.
\nThe most relevant aspect of the paper in the limited context of our competition may be the optimisation of what compounds to expend the limited available docking resources on. See also other recent work in that area, which I happened to be reading this week.
\nNonetheless, in a slightly wider context the generative AI search strategy is really interesting, and that's the thing that stands out to me as novel and innovative about this work.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2814031": "https://arxiv.org/pdf/2405.01616 \n[paper] GENERATIVE ACTIVE LEARNING FOR THE SEARCH OF SMALL-MOLECULE PROTEIN BINDERS\nhttps://github.com/pchliu\nhttps://iclr.cc/virtual/2021/4223\n\n- We introduce LAMBDAZERO , a generative active learning approach to search for synthesizable molecule\n\n- deep reinforcement learning\n\n- We apply LAMBDAZERO with molecular docking to design novel small molecules that inhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing constraints on synthesizability and drug-likeliness.\n\n- with only ∼ 1e4 docking simulations, LAMBDAZERO produces synthesizable, drug-like molecules with docking scores that would otherwise require the virtual screening of a hundred billion ( 1e11) molecule", "2814670": "Very interesting, thanks for posting! For this competition, obviously we are working with the data we have, which is a very large library of compounds made with a combinatorial chemistry type of approach.\n\nThe most relevant aspect of the paper in the limited context of our competition may be the optimisation of what compounds to expend the limited available docking resources on. See also [other recent work](https://doi.org/10.26434/chemrxiv-2024-qsdd1) in that area, which I happened to be reading this week.\n\nNonetheless, in a slightly wider context the generative AI search strategy is really interesting, and that's the thing that stands out to me as novel and innovative about this work." }, "source": "meta" }