Welcome to the Kaggle challenge on computational protein function prediction. We're thrilled to have you here, and we can't wait to see what you'll accomplish in the Critical Assessment (protein) Function Annotation or CAFA 6. This is the sixth iteration of the CAFA challenge, which we started in 2010, and we are excited about hosting it for the second time on Kaggle. CAFA 5 was also hosted on Kaggle, and you can learn a lot from looking there.
\nPredicting protein function is a critical area of research that has enormous implications for both basic science and real-world applications. Proteins are the workhorses of the cell, carrying out a vast array of tasks, from catalyzing chemical reactions to transporting molecules across membranes to transmitting signals between cells. Understanding how proteins function and their roles is essential for comprehending the fundamental processes of life, as well as for developing new drugs, designing enzymes for industrial applications, and numerous other applications. Proteins are large molecules built of 20 different types of building blocks called amino acids. And while knowing the amino acid sequence (or just \"sequence\") of a protein is relatively simple, inferring what is does, or its function, is still an open problem in computational molecular biology. Therefore, one of the most pressing challenges is to accurately predict the function of newly discovered proteins based on their amino-acid sequence. This is where computational methods come in. By using algorithms to analyze protein sequences and structures, we can make educated guesses about what a protein might do, based on its similarity to other proteins with known functions.
In this Kaggle challenge, you will be focusing on just that: using computational methods to predict the function of proteins whose functions are unknown, or only partially known. We hope that this challenge will not only help us make progress towards this important goal, but also inspire and challenge you to push the boundaries of what is possible in this exciting field.
\nAt the end of this competition, we will write a scientific article describing this event. If you would like to be considered as a co-author, you can opt-in using this form. The organizers reserve the right to select co-authors based on contribution merit. You can always opt-out later by emailing Iddo Friedberg idoerg at iastate dot edu and you will get a chance to see the manuscript before it is submitted.
\nSo once again, welcome, and let's get started!
\nIddo @iddux1, An @ahphan, Pedja @predragradivojac, Clara @claradepaolis, and Damiano @damianopiovesan
", "messages": [ { "id": 3302615, "postDate": "2025-10-16T06:53:05.870Z", "content": "Welcome to the Kaggle challenge on computational protein function prediction. We're thrilled to have you here, and we can't wait to see what you'll accomplish in the Critical Assessment (protein) Function Annotation or CAFA 6. This is the sixth iteration of the CAFA challenge, which we started in 2010, and we are excited about hosting it for the second time on Kaggle. CAFA 5 was also hosted on Kaggle, and you can learn a lot from looking there.
\nPredicting protein function is a critical area of research that has enormous implications for both basic science and real-world applications. Proteins are the workhorses of the cell, carrying out a vast array of tasks, from catalyzing chemical reactions to transporting molecules across membranes to transmitting signals between cells. Understanding how proteins function and their roles is essential for comprehending the fundamental processes of life, as well as for developing new drugs, designing enzymes for industrial applications, and numerous other applications. Proteins are large molecules built of 20 different types of building blocks called amino acids. And while knowing the amino acid sequence (or just \"sequence\") of a protein is relatively simple, inferring what is does, or its function, is still an open problem in computational molecular biology. Therefore, one of the most pressing challenges is to accurately predict the function of newly discovered proteins based on their amino-acid sequence. This is where computational methods come in. By using algorithms to analyze protein sequences and structures, we can make educated guesses about what a protein might do, based on its similarity to other proteins with known functions.
In this Kaggle challenge, you will be focusing on just that: using computational methods to predict the function of proteins whose functions are unknown, or only partially known. We hope that this challenge will not only help us make progress towards this important goal, but also inspire and challenge you to push the boundaries of what is possible in this exciting field.
\nAt the end of this competition, we will write a scientific article describing this event. If you would like to be considered as a co-author, you can opt-in using this form. The organizers reserve the right to select co-authors based on contribution merit. You can always opt-out later by emailing Iddo Friedberg idoerg at iastate dot edu and you will get a chance to see the manuscript before it is submitted.
\nSo once again, welcome, and let's get started!
\nIddo @iddux1, An @ahphan, Pedja @predragradivojac, Clara @claradepaolis, and Damiano @damianopiovesan
", "rawMarkdown": "\n\nWelcome to the Kaggle challenge on computational protein function prediction. We're thrilled to have you here, and we can't wait to see what you'll accomplish in the Critical Assessment (protein) Function Annotation or CAFA 6. This is the sixth iteration of the CAFA challenge, which we started in 2010, and we are excited about hosting it for the second time on Kaggle. CAFA 5 was also [hosted on Kaggle](https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/overview/description), and you can learn a lot from looking there. \n\n\n\nPredicting protein function is a critical area of research that has enormous implications for both basic science and real-world applications. Proteins are the workhorses of the cell, carrying out a vast array of tasks, from catalyzing chemical reactions to transporting molecules across membranes to transmitting signals between cells. Understanding how proteins function and their roles is essential for comprehending the fundamental processes of life, as well as for developing new drugs, designing enzymes for industrial applications, and numerous other applications. Proteins are large molecules built of 20 different types of building blocks called amino acids. And while knowing the amino acid sequence (or just \"sequence\") of a protein is relatively simple, inferring what is does, or its function, is still an open problem in computational molecular biology. Therefore, one of the most pressing challenges is to accurately predict the function of newly discovered proteins based on their amino-acid sequence. This is where computational methods come in. By using algorithms to analyze protein sequences and structures, we can make educated guesses about what a protein might do, based on its similarity to other proteins with known functions.\n \nIn this Kaggle challenge, you will be focusing on just that: using computational methods to predict the function of proteins whose functions are unknown, or only partially known. We hope that this challenge will not only help us make progress towards this important goal, but also inspire and challenge you to push the boundaries of what is possible in this exciting field.\n \n\nAt the end of this competition, we will write a scientific article describing this event. If you would like to be considered as a co-author, you can opt-in using [this form](https://forms.gle/LGeYNhwiPj7x1voc6). The organizers reserve the right to select co-authors based on contribution merit. You can always opt-out later by emailing Iddo Friedberg idoerg at iastate dot edu and you will get a chance to see the manuscript before it is submitted.\n\n \nSo once again, welcome, and let's get started!\n \nIddo @iddux1, An @ahphan, Pedja @predragradivojac, Clara @claradepaolis, and Damiano @damianopiovesan", "votes": 19 }, { "id": 3307649, "postDate": "2025-10-27T13:37:48.560Z", "content": "Great news ! Thanks a lot !
\n@ahphan \nWould it possible to share code/create notebook with blast predictions method - you used - to create baseline scored 0.17 - just for future references in publications ?
\n@iddux1\nWould you be so kind to take a look on on previous CAFA5 question:\nhttps://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/613154#3307593
", "rawMarkdown": "Great news ! Thanks a lot ! \n\n@ahphan \nWould it possible to share code/create notebook with blast predictions method - you used - to create baseline scored 0.17 - just for future references in publications ? \n\n@iddux1\nWould you be so kind to take a look on on previous CAFA5 question:\nhttps://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/613154#3307593 ", "votes": 1, "replies": [ { "id": 3307659, "postDate": "2025-10-27T14:03:41.410Z", "content": "I answered the question in CAFA5 @alexandervc . Why are you submitting to CAFA5 though?
", "rawMarkdown": "I answered the question in CAFA5 @alexandervc . Why are you submitting to CAFA5 though?", "votes": 1, "replies": [ { "id": 3307666, "postDate": "2025-10-27T14:15:16.930Z", "content": "Thanks a lot ! Just to have additional test set.
", "rawMarkdown": "Thanks a lot ! Just to have additional test set. \n", "votes": 1 } ] } ] }, { "id": 3303055, "postDate": "2025-10-17T04:35:12.150Z", "content": "Hi,
\nCould you please add a Code Requirements section to the description?
\nAlso, could you specify which resources are available — CPU, GPU, etc.?
Please note that there is no Code Requirements section, as this is a no-code competition. You are expected to submit a submission.tsv file containing your predictions.
Wow, that’s unexpected 🙂 Thanks for the reply!
\nSo, you’re expecting that within the next few months researchers will experimentally annotate some of the proteins from the test set, and that will allow you to evaluate our predictions later on?
Yes, that is correct. Experimentally-validated functional annotations in any subontology between the submission deadline and the time of evaluation will be used to evaluate participants' predictions.
", "rawMarkdown": "Yes, that is correct. Experimentally-validated functional annotations in any subontology between the submission deadline and the time of evaluation will be used to evaluate participants' predictions.", "votes": 4 }, { "id": 3303265, "postDate": "2025-10-17T14:23:11.463Z", "content": "Considering my previous question, alright, I know this issue.
", "rawMarkdown": "Considering my previous question, alright, I know this issue." }, { "id": 3303390, "postDate": "2025-10-17T21:32:01.073Z", "content": "Yes, that's correct. You can see the historic growth of annotations here: https://geneontology.org/stats.html
", "rawMarkdown": "Yes, that's correct. You can see the historic growth of annotations here: [https://geneontology.org/stats.html](https://geneontology.org/stats.html)", "votes": 2 } ] } ] }, { "id": 3303368, "postDate": "2025-10-17T19:38:02.547Z", "rawMarkdown": "", "isDeleted": true }, { "id": 3307675, "postDate": "2025-10-27T14:37:45.380Z", "content": "Great news ! Thanks a lot !
", "rawMarkdown": "Great news ! Thanks a lot !", "votes": 1 } ], "comments": [ { "id": 3307649, "author_name": "Alexander Chervov", "author_url": "", "post_date": "2025-10-27T13:37:48.560000", "content": "Great news ! Thanks a lot !
\n@ahphan \nWould it possible to share code/create notebook with blast predictions method - you used - to create baseline scored 0.17 - just for future references in publications ?
\n@iddux1\nWould you be so kind to take a look on on previous CAFA5 question:\nhttps://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/613154#3307593
", "votes": 1, "replies": [ { "id": 3307659, "author_name": "Iddo Friedberg", "author_url": "", "post_date": "2025-10-27T14:03:41.410000", "content": "I answered the question in CAFA5 @alexandervc . Why are you submitting to CAFA5 though?
", "votes": 1, "replies": [ { "id": 3307666, "author_name": "Alexander Chervov", "author_url": "", "post_date": "2025-10-27T14:15:16.930000", "content": "Thanks a lot ! Just to have additional test set.
", "votes": 1, "replies": [] } ] } ] }, { "id": 3303055, "author_name": "Pavel Orlov", "author_url": "", "post_date": "2025-10-17T04:35:12.150000", "content": "Hi,
\nCould you please add a Code Requirements section to the description?
\nAlso, could you specify which resources are available — CPU, GPU, etc.?
Please note that there is no Code Requirements section, as this is a no-code competition. You are expected to submit a submission.tsv file containing your predictions.
Wow, that’s unexpected 🙂 Thanks for the reply!
\nSo, you’re expecting that within the next few months researchers will experimentally annotate some of the proteins from the test set, and that will allow you to evaluate our predictions later on?
Yes, that is correct. Experimentally-validated functional annotations in any subontology between the submission deadline and the time of evaluation will be used to evaluate participants' predictions.
", "votes": 4, "replies": [] }, { "id": 3303265, "author_name": "Timmy Juicehouse", "author_url": "", "post_date": "2025-10-17T14:23:11.463000", "content": "Considering my previous question, alright, I know this issue.
", "votes": 0, "replies": [] }, { "id": 3303390, "author_name": "Clara De Paolis", "author_url": "", "post_date": "2025-10-17T21:32:01.073000", "content": "Yes, that's correct. You can see the historic growth of annotations here: https://geneontology.org/stats.html
", "votes": 2, "replies": [] } ] } ] }, { "id": 3303368, "author_name": "", "author_url": "", "post_date": "2025-10-17T19:38:02.547000", "content": "", "votes": 0, "replies": [] }, { "id": 3307675, "author_name": "Khushi Yadav", "author_url": "", "post_date": "2025-10-27T14:37:45.380000", "content": "Great news ! Thanks a lot !
", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "3302615": "\n\nWelcome to the Kaggle challenge on computational protein function prediction. We're thrilled to have you here, and we can't wait to see what you'll accomplish in the Critical Assessment (protein) Function Annotation or CAFA 6. This is the sixth iteration of the CAFA challenge, which we started in 2010, and we are excited about hosting it for the second time on Kaggle. CAFA 5 was also [hosted on Kaggle](https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/overview/description), and you can learn a lot from looking there. \n\n\n\nPredicting protein function is a critical area of research that has enormous implications for both basic science and real-world applications. Proteins are the workhorses of the cell, carrying out a vast array of tasks, from catalyzing chemical reactions to transporting molecules across membranes to transmitting signals between cells. Understanding how proteins function and their roles is essential for comprehending the fundamental processes of life, as well as for developing new drugs, designing enzymes for industrial applications, and numerous other applications. Proteins are large molecules built of 20 different types of building blocks called amino acids. And while knowing the amino acid sequence (or just \"sequence\") of a protein is relatively simple, inferring what is does, or its function, is still an open problem in computational molecular biology. Therefore, one of the most pressing challenges is to accurately predict the function of newly discovered proteins based on their amino-acid sequence. This is where computational methods come in. By using algorithms to analyze protein sequences and structures, we can make educated guesses about what a protein might do, based on its similarity to other proteins with known functions.\n \nIn this Kaggle challenge, you will be focusing on just that: using computational methods to predict the function of proteins whose functions are unknown, or only partially known. We hope that this challenge will not only help us make progress towards this important goal, but also inspire and challenge you to push the boundaries of what is possible in this exciting field.\n \n\nAt the end of this competition, we will write a scientific article describing this event. If you would like to be considered as a co-author, you can opt-in using [this form](https://forms.gle/LGeYNhwiPj7x1voc6). The organizers reserve the right to select co-authors based on contribution merit. You can always opt-out later by emailing Iddo Friedberg idoerg at iastate dot edu and you will get a chance to see the manuscript before it is submitted.\n\n \nSo once again, welcome, and let's get started!\n \nIddo @iddux1, An @ahphan, Pedja @predragradivojac, Clara @claradepaolis, and Damiano @damianopiovesan", "3307649": "Great news ! Thanks a lot ! \n\n@ahphan \nWould it possible to share code/create notebook with blast predictions method - you used - to create baseline scored 0.17 - just for future references in publications ? \n\n@iddux1\nWould you be so kind to take a look on on previous CAFA5 question:\nhttps://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/613154#3307593 ", "3303055": "Hi,\nCould you please add a **Code Requirements** section to the description?\nAlso, could you specify which resources are available — CPU, GPU, etc.?\n", "3303368": "", "3307675": "Great news ! Thanks a lot !" } }