Hello everyone,
\nI noticed some entries on the leaderboard that appear to be benchmarks or baselines, specifically files like \"goa_predictions_Oct2025.tsv\".
\nI am interested in understanding how these predictions were generated. Could anyone kindly point me to where the methodology or description for these submissions is documented?
\nThank you in advance for your help.
", "messages": [ { "id": 3371613, "postDate": "2025-12-11T15:21:41.183Z", "content": "The \"goa_nonexp_predictions_Oct2025.tsv\" were created using non-experimental evidence codes in the GAF file (release 2025-09-04, downloaded here). You can re-create this baseline submission with the following steps:
\nThank you for the clear instructions! I understand how the \"goa_nonexp_predictions_Oct2025.tsv\" baseline was created now. This information helps me a lot!
\nI also noticed other entries such as \"goa_predictions_Oct2025.tsv\", \"naive_predictions.tsv\", \"goa_nonexp_predictions.tsv, \"blast_predictions.tsv\" on the Leaderboard. Could you please share how these were generated as well?
", "rawMarkdown": "Thank you for the clear instructions! I understand how the \"goa_nonexp_predictions_Oct2025.tsv\" baseline was created now. This information helps me a lot!\n\nI also noticed other entries such as \"goa_predictions_Oct2025.tsv\", \"naive_predictions.tsv\", \"goa_nonexp_predictions.tsv, \"blast_predictions.tsv\" on the Leaderboard. Could you please share how these were generated as well?", "votes": 2, "replies": [ { "id": 3377041, "postDate": "2025-12-15T15:44:52.707Z", "content": "\"goa_nonexp_predictions.tsv\" was similarly created, but with GAF file release on 2025-05-03 (most updated file before competition launches, downloaded here).
\n\"naive_predictions.tsv\" predicts every GO term for every protein in the \"testsuperset.fasta\" file; the score of each prediction is the frequency of the GO term in the GAF file. For example, if \"GO:0000001\" was annotated to 50% of all proteins in the \"train_terms.tsv\" file, then \"naive_predictions.tsv\" will have \"GO:0000001\" predictions for all proteins with score 0.5 for each prediction.
\n\"blast_predictions.tsv\" was created using two steps:
\nHello everyone,
\nI noticed some entries on the leaderboard that appear to be benchmarks or baselines, specifically files like \"goa_predictions_Oct2025.tsv\".
\nI am interested in understanding how these predictions were generated. Could anyone kindly point me to where the methodology or description for these submissions is documented?
\nThank you in advance for your help.
", "rawMarkdown": "Hello everyone,\n\nI noticed some entries on the leaderboard that appear to be benchmarks or baselines, specifically files like \"goa_predictions_Oct2025.tsv\".\n\nI am interested in understanding how these predictions were generated. Could anyone kindly point me to where the methodology or description for these submissions is documented?\n\nThank you in advance for your help.", "votes": 2 }, { "id": 3372540, "postDate": "2025-12-12T04:26:14.553Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 3371613, "author_name": "An Phan", "author_url": "", "post_date": "2025-12-11T15:21:41.183000", "content": "The \"goa_nonexp_predictions_Oct2025.tsv\" were created using non-experimental evidence codes in the GAF file (release 2025-09-04, downloaded here). You can re-create this baseline submission with the following steps:
\nThank you for the clear instructions! I understand how the \"goa_nonexp_predictions_Oct2025.tsv\" baseline was created now. This information helps me a lot!
\nI also noticed other entries such as \"goa_predictions_Oct2025.tsv\", \"naive_predictions.tsv\", \"goa_nonexp_predictions.tsv, \"blast_predictions.tsv\" on the Leaderboard. Could you please share how these were generated as well?
", "votes": 2, "replies": [ { "id": 3377041, "author_name": "An Phan", "author_url": "", "post_date": "2025-12-15T15:44:52.707000", "content": "\"goa_nonexp_predictions.tsv\" was similarly created, but with GAF file release on 2025-05-03 (most updated file before competition launches, downloaded here).
\n\"naive_predictions.tsv\" predicts every GO term for every protein in the \"testsuperset.fasta\" file; the score of each prediction is the frequency of the GO term in the GAF file. For example, if \"GO:0000001\" was annotated to 50% of all proteins in the \"train_terms.tsv\" file, then \"naive_predictions.tsv\" will have \"GO:0000001\" predictions for all proteins with score 0.5 for each prediction.
\n\"blast_predictions.tsv\" was created using two steps:
\n