I noticed that in train_terms, in the beginning of the file,ther ar are Q3EC77 GO:0000138 C\nQ3EC77 GO:0005794 C, but I have searched the UniProt, which is a different result. Anyone can exlain it to me, thanks!
", "messages": [ { "id": "3323494", "postDate": "11/14/2025 08:42:48", "content": "I noticed that in train_terms, in the beginning of the file,ther ar are Q3EC77 GO:0000138 C\nQ3EC77 GO:0005794 C, but I have searched the UniProt, which is a different result. Anyone can exlain it to me, thanks!
", "rawMarkdown": "I noticed that in train_terms, in the beginning of the file,ther ar are Q3EC77\tGO:0000138\tC\nQ3EC77\tGO:0005794\tC, but I have searched the UniProt, which is a different result. Anyone can exlain it to me, thanks!", "votes": null }, { "id": "3323868", "postDate": "11/14/2025 14:27:10", "content": "Which UniProt file did you check? Also see this discussion (bullet point 3), if you have not read it.
", "rawMarkdown": "Which UniProt file did you check? Also see [this discussion](https://www.kaggle.com/competitions/cafa-6-protein-function-prediction/discussion/614270#3310379) (bullet point 3), if you have not read it.", "votes": null }, { "id": "3324959", "postDate": "11/15/2025 03:54:03", "content": "Thanks for your reply, the GO terms of Q3EC77 in the web https://www.uniprot.org/uniprotkb/Q3EC77/entry is different from Q3EC77 GO:0000138 C Q3EC77 GO:0005794 C. And in my opinion, it is strange for a protein that only have CC GO.
", "rawMarkdown": "Thanks for your reply, the GO terms of Q3EC77 in the web https://www.uniprot.org/uniprotkb/Q3EC77/entry is different from Q3EC77 GO:0000138 C Q3EC77 GO:0005794 C. And in my opinion, it is strange for a protein that only have CC GO.", "votes": null }, { "id": "3332281", "postDate": "11/16/2025 16:50:55", "content": "Did you check the GOA file on UniProt FTP to see if Q3EC77 has experimentally annotated terms in CCO, since that is where we obtained the training data from
", "rawMarkdown": "Did you check the [GOA file on UniProt FTP](https://ftp.ebi.ac.uk/pub/databases/GO/goa/old/UNIPROT/goa_uniprot_all.gaf.226.gz) to see if Q3EC77 has experimentally annotated terms in CCO, since that is where we obtained the training data from", "votes": null }, { "id": "3362408", "postDate": "12/04/2025 16:53:45", "content": "@ahphan But why would we use the old version of GOA? Only to reproduce your train dataset, right? If I want to use GOA annotation now for training my model (and don't rely on your compiled training dataset), it makes sense to use the ones which are up to date.
", "rawMarkdown": "ahphan But why would we use the old version of GOA? Only to reproduce your train dataset, right? If I want to use GOA annotation now for training my model (and don't rely on your compiled training dataset), it makes sense to use the ones which are up to date.", "votes": null }, { "id": "3362468", "postDate": "12/04/2025 18:17:17", "content": "At the time of data preparation for release, the UniProt GOA version in June 2025 was the most recent version so the training data was extracted from that version. Right after CAFA6 launches, UniProt GOA released a new version in October 2025.
\nYes, you should use current GO annotations now to train your model, since anything that is available now (until submission closes) will be treated as known information, and excluded from the final evaluation in June 2026. Only new experimental/curated annotations accumulated between submission closes (Feb 2026) and June 2026 will be used for the final evaluation.
", "rawMarkdown": "At the time of data preparation for release, the UniProt GOA version in June 2025 was the most recent version so the training data was extracted from that version. Right after CAFA6 launches, UniProt GOA released a new version in October 2025. \n\nYes, you should use current GO annotations now to train your model, since anything that is available now (until submission closes) will be treated as known information, and excluded from the final evaluation in June 2026. Only new experimental/curated annotations accumulated between submission closes (Feb 2026) and June 2026 will be used for the final evaluation.", "votes": null } ], "comments": [ { "id": 3323868, "author_name": "ahphan", "author_url": "", "post_date": "11/14/2025 14:27:10", "content": "Which UniProt file did you check? Also see this discussion (bullet point 3), if you have not read it.
", "votes": null, "replies": [ { "id": 3324959, "author_name": "barristanyi", "author_url": "", "post_date": "11/15/2025 03:54:03", "content": "Thanks for your reply, the GO terms of Q3EC77 in the web https://www.uniprot.org/uniprotkb/Q3EC77/entry is different from Q3EC77 GO:0000138 C Q3EC77 GO:0005794 C. And in my opinion, it is strange for a protein that only have CC GO.
", "votes": null, "replies": [ { "id": 3332281, "author_name": "ahphan", "author_url": "", "post_date": "11/16/2025 16:50:55", "content": "Did you check the GOA file on UniProt FTP to see if Q3EC77 has experimentally annotated terms in CCO, since that is where we obtained the training data from
", "votes": null, "replies": [ { "id": 3362408, "author_name": "quittend", "author_url": "", "post_date": "12/04/2025 16:53:45", "content": "@ahphan But why would we use the old version of GOA? Only to reproduce your train dataset, right? If I want to use GOA annotation now for training my model (and don't rely on your compiled training dataset), it makes sense to use the ones which are up to date.
", "votes": null, "replies": [ { "id": 3362468, "author_name": "ahphan", "author_url": "", "post_date": "12/04/2025 18:17:17", "content": "At the time of data preparation for release, the UniProt GOA version in June 2025 was the most recent version so the training data was extracted from that version. Right after CAFA6 launches, UniProt GOA released a new version in October 2025.
\nYes, you should use current GO annotations now to train your model, since anything that is available now (until submission closes) will be treated as known information, and excluded from the final evaluation in June 2026. Only new experimental/curated annotations accumulated between submission closes (Feb 2026) and June 2026 will be used for the final evaluation.
", "votes": null, "replies": [] } ] } ] } ] } ] } ], "raw_markdown_by_id": { "3323494": "I noticed that in train_terms, in the beginning of the file,ther ar are Q3EC77\tGO:0000138\tC\nQ3EC77\tGO:0005794\tC, but I have searched the UniProt, which is a different result. Anyone can exlain it to me, thanks!", "3323868": "Which UniProt file did you check? Also see [this discussion](https://www.kaggle.com/competitions/cafa-6-protein-function-prediction/discussion/614270#3310379) (bullet point 3), if you have not read it.", "3324959": "Thanks for your reply, the GO terms of Q3EC77 in the web https://www.uniprot.org/uniprotkb/Q3EC77/entry is different from Q3EC77 GO:0000138 C Q3EC77 GO:0005794 C. And in my opinion, it is strange for a protein that only have CC GO.", "3332281": "Did you check the [GOA file on UniProt FTP](https://ftp.ebi.ac.uk/pub/databases/GO/goa/old/UNIPROT/goa_uniprot_all.gaf.226.gz) to see if Q3EC77 has experimentally annotated terms in CCO, since that is where we obtained the training data from", "3362408": "ahphan But why would we use the old version of GOA? Only to reproduce your train dataset, right? If I want to use GOA annotation now for training my model (and don't rely on your compiled training dataset), it makes sense to use the ones which are up to date.", "3362468": "At the time of data preparation for release, the UniProt GOA version in June 2025 was the most recent version so the training data was extracted from that version. Right after CAFA6 launches, UniProt GOA released a new version in October 2025. \n\nYes, you should use current GO annotations now to train your model, since anything that is available now (until submission closes) will be treated as known information, and excluded from the final evaluation in June 2026. Only new experimental/curated annotations accumulated between submission closes (Feb 2026) and June 2026 will be used for the final evaluation." }, "source": "meta" }