train_terms.tsv provided by cafa5 organizer comes from 2022-11-17 version of uniprot-goa at http://ftp.ebi.ac.uk/pub/databases/GO/goa/old/UNIPROT/goa_uniprot_all.gaf.212.gz.
\nHowever, I observe several inconsistencies between train_terms.tsv and the upstream gaf file.
Some experimental GO terms in the gaf file is missing in train_terms.tsv. For example, A0A0H3L5C8 has the experimental CC term \"GO:0044228\" with IDA evidence in the gaf file but it has no CC term in train_terms.tsv. A0A024B7W1 has the BP term GO:0039653 in the gaf file but no BP term in train_terms.tsv. For both proteins their MF terms are correctly parsed. One phenomenon I observe for the missing GO annotation in train_terms.tsv is that they generally have some qualifier such as \"involved_in\" in the 4-th column of the GAF file. Should these annotations be excluded?
", "messages": [ { "id": "2284254", "postDate": "06/01/2023 21:00:02", "content": "train_terms.tsv provided by cafa5 organizer comes from 2022-11-17 version of uniprot-goa at http://ftp.ebi.ac.uk/pub/databases/GO/goa/old/UNIPROT/goa_uniprot_all.gaf.212.gz.
\nHowever, I observe several inconsistencies between train_terms.tsv and the upstream gaf file.
Some experimental GO terms in the gaf file is missing in train_terms.tsv. For example, A0A0H3L5C8 has the experimental CC term \"GO:0044228\" with IDA evidence in the gaf file but it has no CC term in train_terms.tsv. A0A024B7W1 has the BP term GO:0039653 in the gaf file but no BP term in train_terms.tsv. For both proteins their MF terms are correctly parsed. One phenomenon I observe for the missing GO annotation in train_terms.tsv is that they generally have some qualifier such as \"involved_in\" in the 4-th column of the GAF file. Should these annotations be excluded?
", "rawMarkdown": "train_terms.tsv provided by cafa5 organizer comes from 2022-11-17 version of uniprot-goa at http://ftp.ebi.ac.uk/pub/databases/GO/goa/old/UNIPROT/goa_uniprot_all.gaf.212.gz.\nHowever, I observe several inconsistencies between train_terms.tsv and the upstream gaf file.\n\nSome experimental GO terms in the gaf file is missing in train_terms.tsv. For example, A0A0H3L5C8 has the experimental CC term \"GO:0044228\" with IDA evidence in the gaf file but it has no CC term in train_terms.tsv. A0A024B7W1 has the BP term GO:0039653 in the gaf file but no BP term in train_terms.tsv. For both proteins their MF terms are correctly parsed. One phenomenon I observe for the missing GO annotation in train_terms.tsv is that they generally have some qualifier such as \"involved_in\" in the 4-th column of the GAF file. Should these annotations be excluded?", "votes": null }, { "id": "2284259", "postDate": "06/01/2023 21:06:09", "content": "Moreover, according to CAFA3 paper \"Protein binding is a highly generalized function description, does not provide more specific information about the actual function of a protein, and in many cases may indicate a non-functional, non-specific binding. If it is the only annotation that a protein has gained, then it is hardly an advance in our understanding of that protein; therefore, we deleted these annotations from our benchmark set.\"
\nTherefore, I assume that we should exclude a protein if its only annotation is GO:0005515 \"protein binding\". However, train_term.tsv includes many proteins whose only leaf term is protein binding, such as A0A024R4Z4.
", "rawMarkdown": "Moreover, according to CAFA3 paper \"Protein binding is a highly generalized function description, does not provide more specific information about the actual function of a protein, and in many cases may indicate a non-functional, non-specific binding. If it is the only annotation that a protein has gained, then it is hardly an advance in our understanding of that protein; therefore, we deleted these annotations from our benchmark set.\"\n\nTherefore, I assume that we should exclude a protein if its only annotation is GO:0005515 \"protein binding\". However, train_term.tsv includes many proteins whose only leaf term is protein binding, such as A0A024R4Z4.", "votes": null }, { "id": "2284260", "postDate": "06/01/2023 21:07:11", "content": "Finally, are proteins with only \"protein binding\" included in the test set for kaggle public score calculation?
", "rawMarkdown": "Finally, are proteins with only \"protein binding\" included in the test set for kaggle public score calculation?", "votes": null } ], "comments": [ { "id": 2284259, "author_name": "gdsyzxzcx", "author_url": "", "post_date": "06/01/2023 21:06:09", "content": "Moreover, according to CAFA3 paper \"Protein binding is a highly generalized function description, does not provide more specific information about the actual function of a protein, and in many cases may indicate a non-functional, non-specific binding. If it is the only annotation that a protein has gained, then it is hardly an advance in our understanding of that protein; therefore, we deleted these annotations from our benchmark set.\"
\nTherefore, I assume that we should exclude a protein if its only annotation is GO:0005515 \"protein binding\". However, train_term.tsv includes many proteins whose only leaf term is protein binding, such as A0A024R4Z4.
", "votes": null, "replies": [] }, { "id": 2284260, "author_name": "gdsyzxzcx", "author_url": "", "post_date": "06/01/2023 21:07:11", "content": "Finally, are proteins with only \"protein binding\" included in the test set for kaggle public score calculation?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2284254": "train_terms.tsv provided by cafa5 organizer comes from 2022-11-17 version of uniprot-goa at http://ftp.ebi.ac.uk/pub/databases/GO/goa/old/UNIPROT/goa_uniprot_all.gaf.212.gz.\nHowever, I observe several inconsistencies between train_terms.tsv and the upstream gaf file.\n\nSome experimental GO terms in the gaf file is missing in train_terms.tsv. For example, A0A0H3L5C8 has the experimental CC term \"GO:0044228\" with IDA evidence in the gaf file but it has no CC term in train_terms.tsv. A0A024B7W1 has the BP term GO:0039653 in the gaf file but no BP term in train_terms.tsv. For both proteins their MF terms are correctly parsed. One phenomenon I observe for the missing GO annotation in train_terms.tsv is that they generally have some qualifier such as \"involved_in\" in the 4-th column of the GAF file. Should these annotations be excluded?", "2284259": "Moreover, according to CAFA3 paper \"Protein binding is a highly generalized function description, does not provide more specific information about the actual function of a protein, and in many cases may indicate a non-functional, non-specific binding. If it is the only annotation that a protein has gained, then it is hardly an advance in our understanding of that protein; therefore, we deleted these annotations from our benchmark set.\"\n\nTherefore, I assume that we should exclude a protein if its only annotation is GO:0005515 \"protein binding\". However, train_term.tsv includes many proteins whose only leaf term is protein binding, such as A0A024R4Z4.", "2284260": "Finally, are proteins with only \"protein binding\" included in the test set for kaggle public score calculation?" }, "source": "meta" }