Based on Ksenia Petrova webinar: https://youtu.be/i6s38LGKtX4 Presentation: https://t.me/sberlogacompete/1166
\nFor Multiome part (i.e. ATAC-seq -> RNA-seq) (RNA is also called GEX = Genes Expressions) )
\n4 Layers NN with with batch normalization.
\nSomewhat tricky part - dimensional reduction with the help of differentially opened DNA regions with respect to cell types.
\nSee: https://github.com/openproblems-bio/neurips2021_multimodal_topmethods/blob/main/src/predict_modality/methods/cajal/train/script.py#L33
\nhttps://youtu.be/i6s38LGKtX4?t=861
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2F5fb399e1fb5021b8469e73c64faee95e%2Fphoto_2022-09-17_14-19-22.jpg?generation=1663417179457718&alt=media\")
For \"CITE-seq\" part (i.e. RNA-seq -> CD-proteins ) (CD-proteins also called ADT)
\nNot a neural network was on top.
\nBut rather some kernel RBF model.
\nWould be nice someone would try it and make public notebook.
\nhttps://youtu.be/i6s38LGKtX4?t=927
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2F1b3e3a462ef09ecec5a14a267b7dae43%2Fphoto_2022-09-17_14-22-56.jpg?generation=1663417391259579&alt=media\")
For some other webinars around the competition see:
\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/348661
\n(Press likes on youtube - if you find videos useful - top liked - will a get a prize )
Based on Ksenia Petrova webinar: https://youtu.be/i6s38LGKtX4 Presentation: https://t.me/sberlogacompete/1166
\nFor Multiome part (i.e. ATAC-seq -> RNA-seq) (RNA is also called GEX = Genes Expressions) )
\n4 Layers NN with with batch normalization.
\nSomewhat tricky part - dimensional reduction with the help of differentially opened DNA regions with respect to cell types.
\nSee: https://github.com/openproblems-bio/neurips2021_multimodal_topmethods/blob/main/src/predict_modality/methods/cajal/train/script.py#L33
\nhttps://youtu.be/i6s38LGKtX4?t=861
\n
For \"CITE-seq\" part (i.e. RNA-seq -> CD-proteins ) (CD-proteins also called ADT)
\nNot a neural network was on top.
\nBut rather some kernel RBF model.
\nWould be nice someone would try it and make public notebook.
\nhttps://youtu.be/i6s38LGKtX4?t=927
\n
For some other webinars around the competition see:
\nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/348661
\n(Press likes on youtube - if you find videos useful - top liked - will a get a prize )
Though the gap at \"CITEseq\" part between the first places was not so big.
\nFor \"Multiome\" (ATAC->RNA) it was bigger
\nhttps://youtu.be/i6s38LGKtX4?t=523
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2F091c4707664e556dc273d8057792a5cd%2Fphoto_2022-09-17_14-32-36.jpg?generation=1663417975953416&alt=media\")
Do you know what is Batch Effect? Please teach me.
", "rawMarkdown": "Do you know what is Batch Effect? Please teach me.", "votes": null }, { "id": "2000581", "postDate": "10/23/2022 12:08:02", "content": "Batch effect - is unexpected difference in the data which is due to technical problems not biological.
\nImagine you have absolutely identical cells, but you will make experiment in TWO different laboratories,
\nwe expect same results.
\nBut it never happens.
\nResults would be somehow different.
Got it！Thank you so much.
", "rawMarkdown": "Got it！Thank you so much.", "votes": null } ], "comments": [ { "id": 1943282, "author_name": "alexandervc", "author_url": "", "post_date": "09/17/2022 12:33:09", "content": "Though the gap at \"CITEseq\" part between the first places was not so big.
\nFor \"Multiome\" (ATAC->RNA) it was bigger
\nhttps://youtu.be/i6s38LGKtX4?t=523
\n
Do you know what is Batch Effect? Please teach me.
", "votes": null, "replies": [ { "id": 2000581, "author_name": "alexandervc", "author_url": "", "post_date": "10/23/2022 12:08:02", "content": "Batch effect - is unexpected difference in the data which is due to technical problems not biological.
\nImagine you have absolutely identical cells, but you will make experiment in TWO different laboratories,
\nwe expect same results.
\nBut it never happens.
\nResults would be somehow different.
Got it！Thank you so much.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1943277": "Based on Ksenia Petrova webinar: https://youtu.be/i6s38LGKtX4 Presentation: https://t.me/sberlogacompete/1166\n\nFor Multiome part (i.e. ATAC-seq -> RNA-seq) (RNA is also called GEX = Genes Expressions) )\n4 Layers NN with with batch normalization.\nSomewhat tricky part - dimensional reduction with the help of differentially opened DNA regions with respect to cell types. \nSee: https://github.com/openproblems-bio/neurips2021_multimodal_topmethods/blob/main/src/predict_modality/methods/cajal/train/script.py#L33\nhttps://youtu.be/i6s38LGKtX4?t=861 \n\n\nFor \"CITE-seq\" part (i.e. RNA-seq -> CD-proteins ) (CD-proteins also called ADT)\nNot a neural network was on top.\nBut rather some kernel RBF model.\nWould be nice someone would try it and make public notebook. \nhttps://youtu.be/i6s38LGKtX4?t=927\n\n\n------------\n\nFor some other webinars around the competition see: \nhttps://www.kaggle.com/competitions/open-problems-multimodal/discussion/348661\n(Press likes on youtube - if you find videos useful - top liked - will a get a prize )", "1943282": "Though the gap at \"CITEseq\" part between the first places was not so big. \n\nFor \"Multiome\" (ATAC->RNA) it was bigger\n\nhttps://youtu.be/i6s38LGKtX4?t=523\n\n", "2000509": "Do you know what is Batch Effect? Please teach me.", "2000581": "Batch effect - is unexpected difference in the data which is due to technical problems not biological.\nImagine you have absolutely identical cells, but you will make experiment in TWO different laboratories,\nwe expect same results.\nBut it never happens. \nResults would be somehow different.", "2004042": "Got it！Thank you so much." }, "source": "meta" }