{ "id": 350900, "title": "BioQuestion 03: Feature importance for CITEseq ", "url": "/competitions/open-problems-multimodal/discussion/350900", "author_name": "Alexander Chervov", "post_date": "2022-09-07T14:49:18.356000", "votes": 1, "comment_count": 0, "views": 0, "content": "

Disclaimer. That might improve score, might not, but any outcome would be of interest for research community.
\nSo everyone is welcome to collaborate - hopefully produce a paper - see Discussion1, Discussion2, Discussion3\n

\n

For CITEseq task - we need to predict the protein levels.
\nThe basic biological fact to understand - for each protein there is a corresponding gene which give rise to that protein.

\n

Thus for each target - there is directly corresponding feature !
\nSee correspondence in the post: https://www.kaggle.com/competitions/open-problems-multimodal/discussion/349242

\n

Research question 1: analyze how correlated are the target with its own feature. (I.e. the protein and RNA levels corresponding the SAME gene). Expected correlation is NOT that much high - may be 0.3 or like that. (It is different correlation from what is used for scoring. For scoring - correlation along proteins, but here the question - protein is fixed and correlated along cells).

\n

Research question 2: analyze feature importance for each protein target and try to give biological interpretation - some pathway or G0-group (via gene enrichment analysis).

\n

PS
\nPrevious questions:
\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856
\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856

", "messages": [ { "id": 1930089, "postDate": "2022-09-07T14:49:18.357Z", "content": "

Disclaimer. That might improve score, might not, but any outcome would be of interest for research community.
\nSo everyone is welcome to collaborate - hopefully produce a paper - see Discussion1, Discussion2, Discussion3\n

\n

For CITEseq task - we need to predict the protein levels.
\nThe basic biological fact to understand - for each protein there is a corresponding gene which give rise to that protein.

\n

Thus for each target - there is directly corresponding feature !
\nSee correspondence in the post: https://www.kaggle.com/competitions/open-problems-multimodal/discussion/349242

\n

Research question 1: analyze how correlated are the target with its own feature. (I.e. the protein and RNA levels corresponding the SAME gene). Expected correlation is NOT that much high - may be 0.3 or like that. (It is different correlation from what is used for scoring. For scoring - correlation along proteins, but here the question - protein is fixed and correlated along cells).

\n

Research question 2: analyze feature importance for each protein target and try to give biological interpretation - some pathway or G0-group (via gene enrichment analysis).

\n

PS
\nPrevious questions:
\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856
\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856

", "rawMarkdown": "**Disclaimer.** That might improve score, might not, but any outcome would be of interest for research community. \nSo everyone is welcome to collaborate - hopefully produce a paper - see [Discussion1](https://www.kaggle.com/competitions/open-problems-multimodal/discussion/348293), [Discussion2](https://www.kaggle.com/competitions/open-problems-multimodal/discussion/348293), [Discussion3\n](https://www.kaggle.com/competitions/open-problems-multimodal/discussion/348661)\n\nFor CITEseq task - we need to predict the protein levels. \nThe basic biological fact to understand - for each protein there is a corresponding gene which give rise to that protein.\n\nThus for each target - there is directly corresponding feature ! \nSee correspondence in the post: https://www.kaggle.com/competitions/open-problems-multimodal/discussion/349242\n\n**Research question 1:** analyze how correlated are the target with its own feature. (I.e. the protein and RNA levels corresponding the SAME gene). Expected correlation is NOT that much high - may be 0.3 or like that. (It is different correlation from what is used for scoring. For scoring - correlation along proteins, but here the question - protein is fixed and correlated along cells). \n\n**Research question 2:** analyze feature importance for each protein target and try to give biological interpretation - some pathway or G0-group (via gene enrichment analysis). \n\nPS\nPrevious questions: \nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856\n\n\n", "votes": 1 } ], "comments": [], "raw_markdown_by_id": { "1930089": "**Disclaimer.** That might improve score, might not, but any outcome would be of interest for research community. \nSo everyone is welcome to collaborate - hopefully produce a paper - see [Discussion1](https://www.kaggle.com/competitions/open-problems-multimodal/discussion/348293), [Discussion2](https://www.kaggle.com/competitions/open-problems-multimodal/discussion/348293), [Discussion3\n](https://www.kaggle.com/competitions/open-problems-multimodal/discussion/348661)\n\nFor CITEseq task - we need to predict the protein levels. \nThe basic biological fact to understand - for each protein there is a corresponding gene which give rise to that protein.\n\nThus for each target - there is directly corresponding feature ! \nSee correspondence in the post: https://www.kaggle.com/competitions/open-problems-multimodal/discussion/349242\n\n**Research question 1:** analyze how correlated are the target with its own feature. (I.e. the protein and RNA levels corresponding the SAME gene). Expected correlation is NOT that much high - may be 0.3 or like that. (It is different correlation from what is used for scoring. For scoring - correlation along proteins, but here the question - protein is fixed and correlated along cells). \n\n**Research question 2:** analyze feature importance for each protein target and try to give biological interpretation - some pathway or G0-group (via gene enrichment analysis). \n\nPS\nPrevious questions: \nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/350856\n\n\n" } }