the metric code is given at:
\nhttps://www.kaggle.com/code/metric/ribonanza-tm-score/notebook
the score code is:
\ndef (: pd.DataFrame, : pd.DataFrame, : str) -> :\n\n \n solution[] = solution[].(lambda : x.()[])\n submission[] = submission[].(lambda : x.()[])\nnow we know that, the submission ID column has the format:
\n,resname,resid,x_1,y_1,z_1,x_2,y_2,z_2,x_3,y_3,z_3,x_4,y_4,z_4,x_5,y_5,z_5\n,G,,-.,.,.,,,,,,,,,,,,\n,G,,-.,.,.,,,,,,,,,,,,\n,C,,-.,-.,.,,,,,,,,,,,,\n,A,,-.,-.,.,,,,,,,,,,,,\n,G,,-.,-.,.,,,,,,,,,,,,\n\n\n = pdb_id + chain_id + residue_id\nthe score code will ignore the chain and combine all chains into one. is this the intended behaviour?
\nif so, our model must align the multiple chains per rna?
the metric code is given at:
\nhttps://www.kaggle.com/code/metric/ribonanza-tm-score/notebook
the score code is:
\ndef (: pd.DataFrame, : pd.DataFrame, : str) -> :\n\n \n solution[] = solution[].(lambda : x.()[])\n submission[] = submission[].(lambda : x.()[])\nnow we know that, the submission ID column has the format:
\n,resname,resid,x_1,y_1,z_1,x_2,y_2,z_2,x_3,y_3,z_3,x_4,y_4,z_4,x_5,y_5,z_5\n,G,,-.,.,.,,,,,,,,,,,,\n,G,,-.,.,.,,,,,,,,,,,,\n,C,,-.,-.,.,,,,,,,,,,,,\n,A,,-.,-.,.,,,,,,,,,,,,\n,G,,-.,-.,.,,,,,,,,,,,,\n\n\n = pdb_id + chain_id + residue_id\nthe score code will ignore the chain and combine all chains into one. is this the intended behaviour?
\nif so, our model must align the multiple chains per rna?
Not sure if is the answer you need but:
\n\"For a few targets, multiple slightly different structures have been captured experimentally; your predictions' scores will be based on the best TM-score compared to each of these reference structures.\"
Good spot - I'm a complete newbie to this field, but it seems that RNA is a single-stranded chain, so maybe there is no ambiguity here @hengck23.
", "rawMarkdown": "Good spot - I'm a complete newbie to this field, but it seems that RNA is a single-stranded chain, so maybe there is no ambiguity here @hengck23.", "votes": null }, { "id": "3147170", "postDate": "03/11/2025 18:24:07", "content": "there are monomers and complex RNA targets in casp. I now presume that we are only interested in monomers case (until the host clarifies). In that case the correct code for getting target id in dataframe should have been
\ndf[] = submission[].apply(lambda x: ,join(x.split()[:-]))\nBruhv is good, I want to be as good as you @hengck23. 😪
", "rawMarkdown": "Bruhv is good, I want to be as good as you @hengck23. 😪", "votes": null }, { "id": "3147215", "postDate": "03/11/2025 19:07:58", "content": "This is not intended and it's because of splitting multi-chain structures into separate monomers for training. For train, you may use something like this instead
\n re\n\n ():\n = re.(, input_str)\n :\n .group(), .group()\n input_str, \n\n\n(split_pdb_chain()) \n(split_pdb_chain()) \n(split_pdb_chain()) \nAll test targets will be monomers.
", "rawMarkdown": "This is not intended and it's because of splitting multi-chain structures into separate monomers for training. For train, you may use something like this instead\n\n```python\nimport re\n\ndef split_pdb_chain(input_str):\n match = re.match(r\"(.+?)_(\\d+)$\", input_str)\n if match:\n return match.group(1), match.group(2)\n return input_str, None # If no match, return the input and None\n\n# Example cases\nprint(split_pdb_chain(\"6SDW_B_1\")) # Expected: ('6SDW_B', '1')\nprint(split_pdb_chain(\"6SDW_1\")) # Expected: ('6SDW', '1')\nprint(split_pdb_chain(\"6SDW_B\")) # Expected: ('6SDW_B', None)\n```\n\n\nAll test targets will be monomers.", "votes": null }, { "id": "3147226", "postDate": "03/11/2025 19:17:59", "content": "thanks for the clarification!
", "rawMarkdown": "thanks for the clarification!", "votes": null } ], "comments": [ { "id": 3145302, "author_name": "sacuscreed", "author_url": "", "post_date": "03/09/2025 17:16:12", "content": "Not sure if is the answer you need but:
\n\"For a few targets, multiple slightly different structures have been captured experimentally; your predictions' scores will be based on the best TM-score compared to each of these reference structures.\"
Good spot - I'm a complete newbie to this field, but it seems that RNA is a single-stranded chain, so maybe there is no ambiguity here @hengck23.
", "votes": null, "replies": [] }, { "id": 3147170, "author_name": "hengck23", "author_url": "", "post_date": "03/11/2025 18:24:07", "content": "there are monomers and complex RNA targets in casp. I now presume that we are only interested in monomers case (until the host clarifies). In that case the correct code for getting target id in dataframe should have been
\ndf[] = submission[].apply(lambda x: ,join(x.split()[:-]))\nBruhv is good, I want to be as good as you @hengck23. 😪
", "votes": null, "replies": [] } ] }, { "id": 3147215, "author_name": "shujun717", "author_url": "", "post_date": "03/11/2025 19:07:58", "content": "This is not intended and it's because of splitting multi-chain structures into separate monomers for training. For train, you may use something like this instead
\n re\n\n ():\n = re.(, input_str)\n :\n .group(), .group()\n input_str, \n\n\n(split_pdb_chain()) \n(split_pdb_chain()) \n(split_pdb_chain()) \nAll test targets will be monomers.
", "votes": null, "replies": [ { "id": 3147226, "author_name": "hengck23", "author_url": "", "post_date": "03/11/2025 19:17:59", "content": "thanks for the clarification!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3145090": "the metric code is given at:\nhttps://www.kaggle.com/code/metric/ribonanza-tm-score/notebook\n\nthe score code is:\n```\ndef score(solution: pd.DataFrame, submission: pd.DataFrame, row_id_column_name: str) -> float:\n\n # Extract target_id from ID (target_resid)\n solution['target_id'] = solution['ID'].apply(lambda x: x.split('_')[0])\n submission['target_id'] = submission['ID'].apply(lambda x: x.split('_')[0])\n```\n\nnow we know that, the submission ID column has the format:\n```\nID,resname,resid,x_1,y_1,z_1,x_2,y_2,z_2,x_3,y_3,z_3,x_4,y_4,z_4,x_5,y_5,z_5\n6SDW_B_1,G,1,-22.83503,6.876176,17.790787,0,0,0,0,0,0,0,0,0,0,0,0\n6SDW_B_2,G,2,-20.73482,1.7411711,17.335567,0,0,0,0,0,0,0,0,0,0,0,0\n6SDW_B_3,C,3,-20.854864,-1.736068,14.085384,0,0,0,0,0,0,0,0,0,0,0,0\n6SDW_B_4,A,4,-21.762539,-4.106943,9.645739,0,0,0,0,0,0,0,0,0,0,0,0\n6SDW_B_5,G,5,-23.610292,-3.844714,4.6971865,0,0,0,0,0,0,0,0,0,0,0,0\n\n\nID = pdb_id + chain_id + residue_id\n```\n\nthe score code will ignore the chain and combine all chains into one. is this the intended behaviour?\nif so, our model must align the multiple chains per rna?", "3145302": "Not sure if is the answer you need but:\n\"For a few targets, multiple slightly different structures have been captured experimentally; your predictions' scores will be based on the best TM-score compared to each of these reference structures.\"", "3145476": "Good spot - I'm a complete newbie to this field, but it seems that RNA is a single-stranded chain, so maybe there is no ambiguity here @hengck23.", "3147170": "there are monomers and complex RNA targets in casp. I now presume that we are only interested in monomers case (until the host clarifies). In that case the correct code for getting target id in dataframe should have been\n\n```\ndf['target_id'] = submission['ID'].apply(lambda x: '_',join(x.split('_')[:-1]))\n\n```", "3147193": "Bruhv is good, I want to be as good as you @hengck23. 😪", "3147215": "This is not intended and it's because of splitting multi-chain structures into separate monomers for training. For train, you may use something like this instead\n\n```python\nimport re\n\ndef split_pdb_chain(input_str):\n match = re.match(r\"(.+?)_(\\d+)$\", input_str)\n if match:\n return match.group(1), match.group(2)\n return input_str, None # If no match, return the input and None\n\n# Example cases\nprint(split_pdb_chain(\"6SDW_B_1\")) # Expected: ('6SDW_B', '1')\nprint(split_pdb_chain(\"6SDW_1\")) # Expected: ('6SDW', '1')\nprint(split_pdb_chain(\"6SDW_B\")) # Expected: ('6SDW_B', None)\n```\n\n\nAll test targets will be monomers.", "3147226": "thanks for the clarification!" }, "source": "meta" }