![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12193662%2F95d6a050891ab0fb1f14cf7b8f189116%2FScreenshot%202023-12-14%20101132.png?generation=1702538048375329&alt=media\")
It seems that there are genes in the de_train with exactly the same value of expression change.
\nFor example, if we convert de_train to long format and calculate the number of rows for each unique combination of sm_name, cell_type, and value, we can see that e.g. for T regulatory cells, there are 433 genes with the same value.
\nTo illustrate I subsetted these 433 genes from the initial de_train, you can see that all have = 0.7218509894 in the first row, 1.2982483964 in the second row, and so on.
\n
What is the nature of these duplicates? An artifact of LIMMA processing?
", "messages": [ { "id": "2560982", "postDate": "12/14/2023 07:16:39", "content": "It seems that there are genes in the de_train with exactly the same value of expression change.
\nFor example, if we convert de_train to long format and calculate the number of rows for each unique combination of sm_name, cell_type, and value, we can see that e.g. for T regulatory cells, there are 433 genes with the same value.
\nTo illustrate I subsetted these 433 genes from the initial de_train, you can see that all have = 0.7218509894 in the first row, 1.2982483964 in the second row, and so on.
\n
What is the nature of these duplicates? An artifact of LIMMA processing?
", "rawMarkdown": "It seems that there are genes in the de_train with exactly the same value of expression change.\n\nFor example, if we convert de_train to long format and calculate the number of rows for each unique combination of sm_name, cell_type, and value, we can see that e.g. for T regulatory cells, there are 433 genes with the same value. \n\nTo illustrate I subsetted these 433 genes from the initial de_train, you can see that all have = 0.7218509894 in the first row, 1.2982483964 in the second row, and so on.\n\n\n\nWhat is the nature of these duplicates? An artifact of LIMMA processing?", "votes": null }, { "id": "2560987", "postDate": "12/14/2023 07:27:07", "content": "Great observation! The similarity after decimal point in that many genes is intriguing. It does appear to be an artifact from the LIMMA processing.
", "rawMarkdown": "Great observation! The similarity after decimal point in that many genes is intriguing. It does appear to be an artifact from the LIMMA processing.", "votes": null }, { "id": "2561442", "postDate": "12/14/2023 13:36:17", "content": "Hi @antoninadolgorukova If you look at the RNA counts, you'll see that these 433 genes have an RNA count of zero in T regulatory cells for all compounds. A gene which is never observed shouldn't have a differential expression value at all. This is an artifact of Limma indeed.
\nThe 269 genes which are never expressed in T cells CD8+ are discussed in that topic.
\nAnd it's kind of absurd that we try to predict differential expressions for these genes.
", "rawMarkdown": "Hi @antoninadolgorukova If you look at the RNA counts, you'll see that these 433 genes have an RNA count of zero in T regulatory cells for all compounds. A gene which is never observed shouldn't have a differential expression value at all. This is an artifact of Limma indeed.\n\nThe 269 genes which are never expressed in T cells CD8+ are discussed in [that topic](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661).\n\nAnd it's kind of absurd that we try to predict differential expressions for these genes.", "votes": null }, { "id": "2561452", "postDate": "12/14/2023 13:47:47", "content": "Omg, thank you for clarifying that. I did suspect they might be zero - expressed but I thought if so, they all would have same values as other duplicates, there are much more than I showed in the post… Agree that is crazy to predict limma artifacts…
", "rawMarkdown": "Omg, thank you for clarifying that. I did suspect they might be zero - expressed but I thought if so, they all would have same values as other duplicates, there are much more than I showed in the post... Agree that is crazy to predict limma artifacts...", "votes": null } ], "comments": [ { "id": 2560987, "author_name": "kishanvavdara", "author_url": "", "post_date": "12/14/2023 07:27:07", "content": "Great observation! The similarity after decimal point in that many genes is intriguing. It does appear to be an artifact from the LIMMA processing.
", "votes": null, "replies": [] }, { "id": 2561442, "author_name": "ambrosm", "author_url": "", "post_date": "12/14/2023 13:36:17", "content": "Hi @antoninadolgorukova If you look at the RNA counts, you'll see that these 433 genes have an RNA count of zero in T regulatory cells for all compounds. A gene which is never observed shouldn't have a differential expression value at all. This is an artifact of Limma indeed.
\nThe 269 genes which are never expressed in T cells CD8+ are discussed in that topic.
\nAnd it's kind of absurd that we try to predict differential expressions for these genes.
", "votes": null, "replies": [ { "id": 2561452, "author_name": "antoninadolgorukova", "author_url": "", "post_date": "12/14/2023 13:47:47", "content": "Omg, thank you for clarifying that. I did suspect they might be zero - expressed but I thought if so, they all would have same values as other duplicates, there are much more than I showed in the post… Agree that is crazy to predict limma artifacts…
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2560982": "It seems that there are genes in the de_train with exactly the same value of expression change.\n\nFor example, if we convert de_train to long format and calculate the number of rows for each unique combination of sm_name, cell_type, and value, we can see that e.g. for T regulatory cells, there are 433 genes with the same value. \n\nTo illustrate I subsetted these 433 genes from the initial de_train, you can see that all have = 0.7218509894 in the first row, 1.2982483964 in the second row, and so on.\n\n\n\nWhat is the nature of these duplicates? An artifact of LIMMA processing?", "2560987": "Great observation! The similarity after decimal point in that many genes is intriguing. It does appear to be an artifact from the LIMMA processing.", "2561442": "Hi @antoninadolgorukova If you look at the RNA counts, you'll see that these 433 genes have an RNA count of zero in T regulatory cells for all compounds. A gene which is never observed shouldn't have a differential expression value at all. This is an artifact of Limma indeed.\n\nThe 269 genes which are never expressed in T cells CD8+ are discussed in [that topic](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661).\n\nAnd it's kind of absurd that we try to predict differential expressions for these genes.", "2561452": "Omg, thank you for clarifying that. I did suspect they might be zero - expressed but I thought if so, they all would have same values as other duplicates, there are much more than I showed in the post... Agree that is crazy to predict limma artifacts..." }, "source": "meta" }