Welcome to this competition! We are looking forward to seeing your solutions. We have provided ~300M protein/molecule interactions to train on. But even with 100M, chemical space is vast (10^60), so the true challenge is showing that you can generalize to new molecules in the test sets. If you can train a model that generalizes well, that could unlock a new era of medicines! We hope this dataset provides a stepping stone in that direction, opening up this important AI task to the wider ML community.
\nMuch ink has been shed on how to represent molecules. ECFP features, graphs, smiles transformers, 3D meshes, point clouds, spherical projections, and physics simulations, just to name a few. There are also many different ways to bring in information about the proteins. We hope to see a wide diversity of solutions attempted and compared. Please use this forum to bring up and discuss any papers you find interesting on this problem!
\nThe team and I will keep an active eye on this forum, please bring any questions you have.
\nGood Luck!
", "messages": [ { "id": "2735134", "postDate": "04/04/2024 14:49:11", "content": "Welcome to this competition! We are looking forward to seeing your solutions. We have provided ~300M protein/molecule interactions to train on. But even with 100M, chemical space is vast (10^60), so the true challenge is showing that you can generalize to new molecules in the test sets. If you can train a model that generalizes well, that could unlock a new era of medicines! We hope this dataset provides a stepping stone in that direction, opening up this important AI task to the wider ML community.
\nMuch ink has been shed on how to represent molecules. ECFP features, graphs, smiles transformers, 3D meshes, point clouds, spherical projections, and physics simulations, just to name a few. There are also many different ways to bring in information about the proteins. We hope to see a wide diversity of solutions attempted and compared. Please use this forum to bring up and discuss any papers you find interesting on this problem!
\nThe team and I will keep an active eye on this forum, please bring any questions you have.
\nGood Luck!
", "rawMarkdown": "Welcome to this competition! We are looking forward to seeing your solutions. We have provided ~300M protein/molecule interactions to train on. But even with 100M, chemical space is vast (10^60), so the true challenge is showing that you can generalize to new molecules in the test sets. If you can train a model that generalizes well, that could unlock a new era of medicines! We hope this dataset provides a stepping stone in that direction, opening up this important AI task to the wider ML community.\n\nMuch ink has been shed on how to represent molecules. ECFP features, graphs, smiles transformers, 3D meshes, point clouds, spherical projections, and physics simulations, just to name a few. There are also many different ways to bring in information about the proteins. We hope to see a wide diversity of solutions attempted and compared. Please use this forum to bring up and discuss any papers you find interesting on this problem!\n\nThe team and I will keep an active eye on this forum, please bring any questions you have.\n\nGood Luck!", "votes": null }, { "id": "2735151", "postDate": "04/04/2024 15:00:20", "content": "Thank you for joining! We hope the challenge here will prove interesting to you all.
", "rawMarkdown": "Thank you for joining! We hope the challenge here will prove interesting to you all.", "votes": null }, { "id": "2777476", "postDate": "04/26/2024 17:14:18", "content": "Hello, quick question, is the test set the final test set for the competiton, or is there a hold out test set? And if there is, is there a time limit? I had a previous competition where i got nice results, but my featurisation method took to long to parse the hold out test set and I got timed out.
", "rawMarkdown": "Hello, quick question, is the test set the final test set for the competiton, or is there a hold out test set? And if there is, is there a time limit? I had a previous competition where i got nice results, but my featurisation method took to long to parse the hold out test set and I got timed out.", "votes": null }, { "id": "2777833", "postDate": "04/26/2024 20:24:06", "content": "The test set is the final test set for the competition. We did this because we wanted to leave solutions open to docking simulations and other slower methods.
", "rawMarkdown": "The test set is the final test set for the competition. We did this because we wanted to leave solutions open to docking simulations and other slower methods.", "votes": null }, { "id": "2811722", "postDate": "05/13/2024 20:48:16", "content": "Hi! Simple question here.
\nOnce I submit 5 times in a day, the next time I will be able to make another submission will be:
\n24 hours after my last submission or at a specific time in a specific timezone?
Hello, I am Rajiv from Mumbai. I am not able to open the xl input as it is very large. Can you help?
", "rawMarkdown": "Hello, I am Rajiv from Mumbai. I am not able to open the xl input as it is very large. Can you help?", "votes": null } ], "comments": [ { "id": 2735151, "author_name": "ianquigley", "author_url": "", "post_date": "04/04/2024 15:00:20", "content": "Thank you for joining! We hope the challenge here will prove interesting to you all.
", "votes": null, "replies": [] }, { "id": 2777476, "author_name": "ellagale", "author_url": "", "post_date": "04/26/2024 17:14:18", "content": "Hello, quick question, is the test set the final test set for the competiton, or is there a hold out test set? And if there is, is there a time limit? I had a previous competition where i got nice results, but my featurisation method took to long to parse the hold out test set and I got timed out.
", "votes": null, "replies": [ { "id": 2777833, "author_name": "andrewdblevins", "author_url": "", "post_date": "04/26/2024 20:24:06", "content": "The test set is the final test set for the competition. We did this because we wanted to leave solutions open to docking simulations and other slower methods.
", "votes": null, "replies": [] } ] }, { "id": 2811722, "author_name": "rodrigoffreitas", "author_url": "", "post_date": "05/13/2024 20:48:16", "content": "Hi! Simple question here.
\nOnce I submit 5 times in a day, the next time I will be able to make another submission will be:
\n24 hours after my last submission or at a specific time in a specific timezone?
Hello, I am Rajiv from Mumbai. I am not able to open the xl input as it is very large. Can you help?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2735134": "Welcome to this competition! We are looking forward to seeing your solutions. We have provided ~300M protein/molecule interactions to train on. But even with 100M, chemical space is vast (10^60), so the true challenge is showing that you can generalize to new molecules in the test sets. If you can train a model that generalizes well, that could unlock a new era of medicines! We hope this dataset provides a stepping stone in that direction, opening up this important AI task to the wider ML community.\n\nMuch ink has been shed on how to represent molecules. ECFP features, graphs, smiles transformers, 3D meshes, point clouds, spherical projections, and physics simulations, just to name a few. There are also many different ways to bring in information about the proteins. We hope to see a wide diversity of solutions attempted and compared. Please use this forum to bring up and discuss any papers you find interesting on this problem!\n\nThe team and I will keep an active eye on this forum, please bring any questions you have.\n\nGood Luck!", "2735151": "Thank you for joining! We hope the challenge here will prove interesting to you all.", "2777476": "Hello, quick question, is the test set the final test set for the competiton, or is there a hold out test set? And if there is, is there a time limit? I had a previous competition where i got nice results, but my featurisation method took to long to parse the hold out test set and I got timed out.", "2777833": "The test set is the final test set for the competition. We did this because we wanted to leave solutions open to docking simulations and other slower methods.", "2811722": "Hi! Simple question here. \n\nOnce I submit 5 times in a day, the next time I will be able to make another submission will be: \n24 hours after my last submission or at a specific time in a specific timezone?", "2812072": "Hello, I am Rajiv from Mumbai. I am not able to open the xl input as it is very large. Can you help?" }, "source": "meta" }