![\"sirna](\"https://i.imgur.com/9n8MiYK.jpg\")
\nsirna 840:\n![\"sirna](\"https://i.imgur.com/Eo1qI3a.jpg\")
I have been trying to understand why my OOF predictions for experiment U2OS-02 are significantly worse than for U2OS-01 and U2OS-03. And it looks like U2OS-02 has just higher concentration of cells! It came to me as a surprise as I thought that the variation between the experiments was mainly in brightness.
\n\nLook at the two pictures below. Each column is experiments U2OS-01, U2OS-02 and U2OS-03 respectively, each row is site 1 and 2 respectively. First picture is for sirna 317, the second one is sirna 840. I selected two sirnas for demonstration, but I looked at more than a hundred of those, and almost invariably experiment U2OS-02 visually has more cells.
\n\nHonestly, I don't know what to do about it, but I think it can be interesting for others. Apparently, Recursion wants us to find a solution for this kind of variation.
\n\nsirna 317:\n\nsirna 840:\n
I have been trying to understand why my OOF predictions for experiment U2OS-02 are significantly worse than for U2OS-01 and U2OS-03. And it looks like U2OS-02 has just higher concentration of cells! It came to me as a surprise as I thought that the variation between the experiments was mainly in brightness.
\n\nLook at the two pictures below. Each column is experiments U2OS-01, U2OS-02 and U2OS-03 respectively, each row is site 1 and 2 respectively. First picture is for sirna 317, the second one is sirna 840. I selected two sirnas for demonstration, but I looked at more than a hundred of those, and almost invariably experiment U2OS-02 visually has more cells.
\n\nHonestly, I don't know what to do about it, but I think it can be interesting for others. Apparently, Recursion wants us to find a solution for this kind of variation.
\n\nsirna 317:\n\nsirna 840:\n
Does negative/positive controls for U2OS-02 have difference in number of cells ?
", "rawMarkdown": "Does negative/positive controls for U2OS-02 have difference in number of cells ?", "votes": null }, { "id": "622563", "postDate": "09/09/2019 20:24:32", "content": "Not very clear, but looks like yes, U2OS-02 has more cells for the negative control as well. Attaching the picture. U2OS-02 seems to be quite densely packed.
\n\n![\"sirna](\"https://i.imgur.com/IZ215sI.jpg\")
They probably gave more food to U2OS-02 ;)\nJoking apart, they do say that the concentration of reagents may vary, within other things, so the cells may grow slower/faster :/ \n> \"[..] Even when experiments are designed to control for technical variables such as temperature, humidity, and reagent concentration, batch effects unavoidably enter into the data, resulting in images that contain factors of variation due to either biologically relevant variables or irrelevant technical variables. [...]\" rxrx.ai
", "rawMarkdown": "They probably gave more food to U2OS-02 ;)\nJoking apart, they do say that the concentration of reagents may vary, within other things, so the cells may grow slower/faster :/ \n> \"[..] Even when experiments are designed to control for technical variables such as temperature, humidity, and reagent concentration, batch effects unavoidably enter into the data, resulting in images that contain factors of variation due to either biologically relevant variables or irrelevant technical variables. [...]\" [rxrx.ai](https://www.rxrx.ai/)", "votes": null }, { "id": "622565", "postDate": "09/09/2019 20:32:14", "content": "each of them require different growing soup =) Also if I remember from my biology class U2OS divide faster..
\n\nDivision time:
\n\nU2OS: 18-20 hours\nHepG2: 48 hours\nHuvec: 3-4 days\nRPE : 20-24 hours
\n\nBasically it means if you leave cells for 5 days U2OS and RPE will have more cells since they can divide faster =)
", "rawMarkdown": "each of them require different growing soup =) Also if I remember from my biology class U2OS divide faster.. \n\nDivision time: \n\nU2OS: 18-20 hours\nHepG2: 48 hours\nHuvec: 3-4 days\nRPE : 20-24 hours\n\nBasically it means if you leave cells for 5 days U2OS and RPE will have more cells since they can divide faster =)", "votes": null }, { "id": "622575", "postDate": "09/09/2019 20:43:55", "content": "Where did you get this data from? :) Don't remember this in my biology classes.
", "rawMarkdown": "Where did you get this data from? :) Don't remember this in my biology classes.", "votes": null }, { "id": "622579", "postDate": "09/09/2019 20:49:59", "content": "U2OS https://bionumbers.hms.harvard.edu/bionumber.aspx?s=n&v=7&id=107993\nHepG2:http://www.hepg2.com/ (look for doubling time)\nHuvec: https://www.atcc.org/~/ps/CRL-1730.ashx (doubling_time)\nRPE: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2637796/
\n\nHere you go =)
", "rawMarkdown": "U2OS https://bionumbers.hms.harvard.edu/bionumber.aspx?s=n&v=7&id=107993\nHepG2:http://www.hepg2.com/ (look for doubling time)\nHuvec: https://www.atcc.org/~/ps/CRL-1730.ashx (doubling_time)\nRPE: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2637796/\n\nHere you go =)", "votes": null }, { "id": "622652", "postDate": "09/09/2019 23:47:27", "content": "This seems like typical variability to me, if anything it seems that the difference in sirna may have more to do with the density, given the control images, it is hard to say from visual inspection alone. But this is exactly the sort of thing recursion and others want to be able to control for.
", "rawMarkdown": "This seems like typical variability to me, if anything it seems that the difference in sirna may have more to do with the density, given the control images, it is hard to say from visual inspection alone. But this is exactly the sort of thing recursion and others want to be able to control for.", "votes": null }, { "id": "623068", "postDate": "09/10/2019 12:57:45", "content": "I'm surprised you find this in U2OS.. aren't the differences in HEPG2 much more extreme? To me the pictures of the two sites for the same sirna already look nothing alike. How come your model deals well with those but not U2OS?
", "rawMarkdown": "I'm surprised you find this in U2OS.. aren't the differences in HEPG2 much more extreme? To me the pictures of the two sites for the same sirna already look nothing alike. How come your model deals well with those but not U2OS?", "votes": null }, { "id": "623079", "postDate": "09/10/2019 13:12:08", "content": "There are only 3 train experiments for U2OS, it is easier to compare them. I haven't looked at other cell types from that point of view, don't know yet. I will pay attention to HEPG2 as you suggest.
", "rawMarkdown": "There are only 3 train experiments for U2OS, it is easier to compare them. I haven't looked at other cell types from that point of view, don't know yet. I will pay attention to HEPG2 as you suggest.", "votes": null } ], "comments": [ { "id": 622555, "author_name": "drhabib", "author_url": "", "post_date": "09/09/2019 20:10:07", "content": "Does negative/positive controls for U2OS-02 have difference in number of cells ?
", "votes": null, "replies": [ { "id": 622563, "author_name": "zaharch", "author_url": "", "post_date": "09/09/2019 20:24:32", "content": "Not very clear, but looks like yes, U2OS-02 has more cells for the negative control as well. Attaching the picture. U2OS-02 seems to be quite densely packed.
\n\n
This seems like typical variability to me, if anything it seems that the difference in sirna may have more to do with the density, given the control images, it is hard to say from visual inspection alone. But this is exactly the sort of thing recursion and others want to be able to control for.
", "votes": null, "replies": [] } ] }, { "id": 622564, "author_name": "hmendonca", "author_url": "", "post_date": "09/09/2019 20:28:53", "content": "They probably gave more food to U2OS-02 ;)\nJoking apart, they do say that the concentration of reagents may vary, within other things, so the cells may grow slower/faster :/ \n> \"[..] Even when experiments are designed to control for technical variables such as temperature, humidity, and reagent concentration, batch effects unavoidably enter into the data, resulting in images that contain factors of variation due to either biologically relevant variables or irrelevant technical variables. [...]\" rxrx.ai
", "votes": null, "replies": [ { "id": 622565, "author_name": "drhabib", "author_url": "", "post_date": "09/09/2019 20:32:14", "content": "each of them require different growing soup =) Also if I remember from my biology class U2OS divide faster..
\n\nDivision time:
\n\nU2OS: 18-20 hours\nHepG2: 48 hours\nHuvec: 3-4 days\nRPE : 20-24 hours
\n\nBasically it means if you leave cells for 5 days U2OS and RPE will have more cells since they can divide faster =)
", "votes": null, "replies": [] }, { "id": 622575, "author_name": "zaharch", "author_url": "", "post_date": "09/09/2019 20:43:55", "content": "Where did you get this data from? :) Don't remember this in my biology classes.
", "votes": null, "replies": [] }, { "id": 622579, "author_name": "drhabib", "author_url": "", "post_date": "09/09/2019 20:49:59", "content": "U2OS https://bionumbers.hms.harvard.edu/bionumber.aspx?s=n&v=7&id=107993\nHepG2:http://www.hepg2.com/ (look for doubling time)\nHuvec: https://www.atcc.org/~/ps/CRL-1730.ashx (doubling_time)\nRPE: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2637796/
\n\nHere you go =)
", "votes": null, "replies": [] } ] }, { "id": 623068, "author_name": "context", "author_url": "", "post_date": "09/10/2019 12:57:45", "content": "I'm surprised you find this in U2OS.. aren't the differences in HEPG2 much more extreme? To me the pictures of the two sites for the same sirna already look nothing alike. How come your model deals well with those but not U2OS?
", "votes": null, "replies": [ { "id": 623079, "author_name": "zaharch", "author_url": "", "post_date": "09/10/2019 13:12:08", "content": "There are only 3 train experiments for U2OS, it is easier to compare them. I haven't looked at other cell types from that point of view, don't know yet. I will pay attention to HEPG2 as you suggest.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "622530": "I have been trying to understand why my OOF predictions for experiment U2OS-02 are significantly worse than for U2OS-01 and U2OS-03. And it looks like U2OS-02 has just higher concentration of cells! It came to me as a surprise as I thought that the variation between the experiments was mainly in brightness.\n\nLook at the two pictures below. Each column is experiments U2OS-01, U2OS-02 and U2OS-03 respectively, each row is site 1 and 2 respectively. First picture is for sirna 317, the second one is sirna 840. I selected two sirnas for demonstration, but I looked at more than a hundred of those, and almost invariably experiment U2OS-02 visually has more cells.\n\nHonestly, I don't know what to do about it, but I think it can be interesting for others. Apparently, Recursion wants us to find a solution for this kind of variation.\n\nsirna 317:\n\nsirna 840:\n", "622555": "Does negative/positive controls for U2OS-02 have difference in number of cells ?", "622563": "Not very clear, but looks like yes, U2OS-02 has more cells for the negative control as well. Attaching the picture. U2OS-02 seems to be quite densely packed.\n\n", "622564": "They probably gave more food to U2OS-02 ;)\nJoking apart, they do say that the concentration of reagents may vary, within other things, so the cells may grow slower/faster :/ \n> \"[..] Even when experiments are designed to control for technical variables such as temperature, humidity, and reagent concentration, batch effects unavoidably enter into the data, resulting in images that contain factors of variation due to either biologically relevant variables or irrelevant technical variables. [...]\" [rxrx.ai](https://www.rxrx.ai/)", "622565": "each of them require different growing soup =) Also if I remember from my biology class U2OS divide faster.. \n\nDivision time: \n\nU2OS: 18-20 hours\nHepG2: 48 hours\nHuvec: 3-4 days\nRPE : 20-24 hours\n\nBasically it means if you leave cells for 5 days U2OS and RPE will have more cells since they can divide faster =)", "622575": "Where did you get this data from? :) Don't remember this in my biology classes.", "622579": "U2OS https://bionumbers.hms.harvard.edu/bionumber.aspx?s=n&v=7&id=107993\nHepG2:http://www.hepg2.com/ (look for doubling time)\nHuvec: https://www.atcc.org/~/ps/CRL-1730.ashx (doubling_time)\nRPE: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2637796/\n\nHere you go =)", "622652": "This seems like typical variability to me, if anything it seems that the difference in sirna may have more to do with the density, given the control images, it is hard to say from visual inspection alone. But this is exactly the sort of thing recursion and others want to be able to control for.", "623068": "I'm surprised you find this in U2OS.. aren't the differences in HEPG2 much more extreme? To me the pictures of the two sites for the same sirna already look nothing alike. How come your model deals well with those but not U2OS?", "623079": "There are only 3 train experiments for U2OS, it is easier to compare them. I haven't looked at other cell types from that point of view, don't know yet. I will pay attention to HEPG2 as you suggest." }, "source": "meta" }