Hi all, thanks for coming to our first info session yesterday. You can find a recording of the session on YouTube here: https://youtu.be/MaHQkT59oEs The recording isn't trimmed yet, apparently that can take a few hours after upload, but I wanted to share it for everyone's reference.
\nSome notes from the session:
\nTo find the notebooks and scripts for running the DE analysis, please check the GitHub repo here: https://github.com/openproblems-bio/neurips-2023-scripts. This is also downloaded into new Saturn Cloud resources started with the recipe linked to in the Saturn Cloud thread.
For details about how the p-value is calculated in Limma, please consult the user guide https://bioconductor.org/packages/devel/bioc/vignettes/limma/inst/doc/usersguide.pdf
For a guide to multiome data analysis, please consult the Single-Cell Best Practices book or the Muon Tutorials. For more details on the technology and biological context, consult the 10x Genomics training module for Multiome.
Is there anything else I'm missing here? Please feel free to leave a note in the comments.
", "messages": [ { "id": "2467481", "postDate": "10/04/2023 15:37:40", "content": "Hi all, thanks for coming to our first info session yesterday. You can find a recording of the session on YouTube here: https://youtu.be/MaHQkT59oEs The recording isn't trimmed yet, apparently that can take a few hours after upload, but I wanted to share it for everyone's reference.
\nSome notes from the session:
\nTo find the notebooks and scripts for running the DE analysis, please check the GitHub repo here: https://github.com/openproblems-bio/neurips-2023-scripts. This is also downloaded into new Saturn Cloud resources started with the recipe linked to in the Saturn Cloud thread.
For details about how the p-value is calculated in Limma, please consult the user guide https://bioconductor.org/packages/devel/bioc/vignettes/limma/inst/doc/usersguide.pdf
For a guide to multiome data analysis, please consult the Single-Cell Best Practices book or the Muon Tutorials. For more details on the technology and biological context, consult the 10x Genomics training module for Multiome.
Is there anything else I'm missing here? Please feel free to leave a note in the comments.
", "rawMarkdown": "Hi all, thanks for coming to our first info session yesterday. You can find a recording of the session on YouTube here: https://youtu.be/MaHQkT59oEs The recording isn't trimmed yet, apparently that can take a few hours after upload, but I wanted to share it for everyone's reference.\n\nSome notes from the session:\n\n1. To find the notebooks and scripts for running the DE analysis, please check the GitHub repo here: https://github.com/openproblems-bio/neurips-2023-scripts. This is also downloaded into new Saturn Cloud resources started with the recipe linked to in the Saturn Cloud thread.\n\n2. For details about how the p-value is calculated in Limma, please consult the user guide https://bioconductor.org/packages/devel/bioc/vignettes/limma/inst/doc/usersguide.pdf\n\n3. For a guide to multiome data analysis, please consult [the Single-Cell Best Practices book](https://www.sc-best-practices.org/chromatin_accessibility/introduction.html) or the [Muon Tutorials](https://github.com/scverse/muon-tutorials/tree/master/single-cell-rna-atac/pbmc10k). For more details on the technology and biological context, consult [the 10x Genomics training module for Multiome](https://www.10xgenomics.com/videos/cloiq73tpm).\n\nIs there anything else I'm missing here? Please feel free to leave a note in the comments.", "votes": null }, { "id": "2469600", "postDate": "10/06/2023 13:18:13", "content": "Thanks for the YouTube of the first info session.
\nI think I understand that the goal of the competition is to predict how confident you can be that a compound has an effect on a gene. Your experience has indicated to you that the p-value is the most robust indicator that you have found that indicates that two distributions are different.
", "rawMarkdown": "Thanks for the YouTube of the first info session. \n\nI think I understand that the goal of the competition is to predict how confident you can be that a compound has an effect on a gene. Your experience has indicated to you that the p-value is the most robust indicator that you have found that indicates that two distributions are different.", "votes": null }, { "id": "2471831", "postDate": "10/06/2023 16:57:41", "content": "My understanding is similar,
\nhowever I am still confused - since LIMMA is linear model its p-values basically the same as measuring log-fold change,
\nthey cannot capture real differences between the distributions - as KS or chi2 or any other concordance criteria would in theory be able.
\nSo despite , Daniel, comments here https://youtu.be/MaHQkT59oEs?si=591xRPvEGuRaeovQ&t=956
\nare quite inspiring and clarifying,
\nI am still confused whether the p-values from linear model like LIMMA are really able to catch difference in distributions, not just means.
Hi @danielburkhardt! When will the fixed RNA expression AnnData be uploaded to Kaggle?
", "rawMarkdown": "Hi @danielburkhardt! When will the fixed RNA expression AnnData be uploaded to Kaggle?", "votes": null }, { "id": "2476365", "postDate": "10/10/2023 14:45:20", "content": "To the best of my knowledge, in limma, the LFC estimates are only adjusted by shrinkage. The p-values are adjusted accounting for other covariates, like library or donor.
", "rawMarkdown": "To the best of my knowledge, in limma, the LFC estimates are only adjusted by shrinkage. The p-values are adjusted accounting for other covariates, like library or donor.", "votes": null }, { "id": "2478796", "postDate": "10/12/2023 08:04:01", "content": "thank you for these invaluable resources. I am gaining more insights about the data
", "rawMarkdown": "thank you for these invaluable resources. I am gaining more insights about the data", "votes": null } ], "comments": [ { "id": 2469600, "author_name": "pcjimmmy", "author_url": "", "post_date": "10/06/2023 13:18:13", "content": "Thanks for the YouTube of the first info session.
\nI think I understand that the goal of the competition is to predict how confident you can be that a compound has an effect on a gene. Your experience has indicated to you that the p-value is the most robust indicator that you have found that indicates that two distributions are different.
", "votes": null, "replies": [ { "id": 2471831, "author_name": "alexandervc", "author_url": "", "post_date": "10/06/2023 16:57:41", "content": "My understanding is similar,
\nhowever I am still confused - since LIMMA is linear model its p-values basically the same as measuring log-fold change,
\nthey cannot capture real differences between the distributions - as KS or chi2 or any other concordance criteria would in theory be able.
\nSo despite , Daniel, comments here https://youtu.be/MaHQkT59oEs?si=591xRPvEGuRaeovQ&t=956
\nare quite inspiring and clarifying,
\nI am still confused whether the p-values from linear model like LIMMA are really able to catch difference in distributions, not just means.
To the best of my knowledge, in limma, the LFC estimates are only adjusted by shrinkage. The p-values are adjusted accounting for other covariates, like library or donor.
", "votes": null, "replies": [] } ] } ] }, { "id": 2475559, "author_name": "songqizhou", "author_url": "", "post_date": "10/10/2023 02:26:44", "content": "Hi @danielburkhardt! When will the fixed RNA expression AnnData be uploaded to Kaggle?
", "votes": null, "replies": [] }, { "id": 2478796, "author_name": "bonfaceonyango", "author_url": "", "post_date": "10/12/2023 08:04:01", "content": "thank you for these invaluable resources. I am gaining more insights about the data
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2467481": "Hi all, thanks for coming to our first info session yesterday. You can find a recording of the session on YouTube here: https://youtu.be/MaHQkT59oEs The recording isn't trimmed yet, apparently that can take a few hours after upload, but I wanted to share it for everyone's reference.\n\nSome notes from the session:\n\n1. To find the notebooks and scripts for running the DE analysis, please check the GitHub repo here: https://github.com/openproblems-bio/neurips-2023-scripts. This is also downloaded into new Saturn Cloud resources started with the recipe linked to in the Saturn Cloud thread.\n\n2. For details about how the p-value is calculated in Limma, please consult the user guide https://bioconductor.org/packages/devel/bioc/vignettes/limma/inst/doc/usersguide.pdf\n\n3. For a guide to multiome data analysis, please consult [the Single-Cell Best Practices book](https://www.sc-best-practices.org/chromatin_accessibility/introduction.html) or the [Muon Tutorials](https://github.com/scverse/muon-tutorials/tree/master/single-cell-rna-atac/pbmc10k). For more details on the technology and biological context, consult [the 10x Genomics training module for Multiome](https://www.10xgenomics.com/videos/cloiq73tpm).\n\nIs there anything else I'm missing here? Please feel free to leave a note in the comments.", "2469600": "Thanks for the YouTube of the first info session. \n\nI think I understand that the goal of the competition is to predict how confident you can be that a compound has an effect on a gene. Your experience has indicated to you that the p-value is the most robust indicator that you have found that indicates that two distributions are different.", "2471831": "My understanding is similar,\nhowever I am still confused - since LIMMA is linear model its p-values basically the same as measuring log-fold change,\nthey cannot capture real differences between the distributions - as KS or chi2 or any other concordance criteria would in theory be able. \nSo despite , Daniel, comments here https://youtu.be/MaHQkT59oEs?si=591xRPvEGuRaeovQ&t=956\nare quite inspiring and clarifying,\nI am still confused whether the p-values from linear model like LIMMA are really able to catch difference in distributions, not just means.", "2475559": "Hi @danielburkhardt! When will the fixed RNA expression AnnData be uploaded to Kaggle?", "2476365": "To the best of my knowledge, in limma, the LFC estimates are only adjusted by shrinkage. The p-values are adjusted accounting for other covariates, like library or donor.", "2478796": "thank you for these invaluable resources. I am gaining more insights about the data" }, "source": "meta" }