jatinarora-upmc
commented
3 years ago @shntnu thanks so much for putting this discussion here. also I promise you i will put all future discussion here instead of emails, so others can also benefit.
Closed shntnu closed 2 years ago
jatinarora-upmc
commented
3 years ago @shntnu thanks so much for putting this discussion here. also I promise you i will put all future discussion here instead of emails, so others can also benefit.
From: Arora, Jatin jarora1@bwh.harvard.edu
Hello Guys,
In isolate cells, rare protein-coding variation in PRLR (prolactin receptor) gene is associated with many traits. Some of those traits are. Cells_Intensity_MeanIntensityEdge_DNA, effect size = 1.46 Nuclei_Granularity_1_DNA, effect size = 1.04 Cytoplasm_Texture_InfoMeas1_DNA_20_02, effect size = -1.58 Cytoplasm_RadialDistribution_RadialCV_Mito_1of4, effect size = -1.83
However, I found it a bit difficult to make biological sense of PRLR’s associations. I was wondering if you already know this gene in any context of cellular morphology?
From: Anne Carpenter anne@broadinstitute.org
Can you remind me how you've defined isolated cells - do they have zero contact with another cell? If so, then features 1 & 3 are pretty odd, because for 1 the cell should not have any DNA staining at its outer edge (Beth can remind me whether it's the outer edge only, or the inner edge is also counted) and for 3 the cell should not have any DNA staining in its cytoplasm. Normally you'd only see those having any signal if the cells are crowded on top of each other, so here I would think they are unstable statistics (and in this worst case reporting on some artifact like specks of DNA stain where they aren't supposed to be). Feature 2 refers to how smooth the DNA staining is within the nucleus. Feature 4 is referring to how much mitochondria staining is in the ring just outside the nucleus. These all could have something to do with cells being severely rounded up. Have you looked at a few images? Can you send those along? (I've not looked up PRLR specifically, trying to stay a bit unbiased)
From: Beth Cimini bcimini@broadinstitute.org
It should just be the outer edge of the cell, so yes, agree that 1 and 3 are likely slightly suspect.--
From: Arora, Jatin jarora1@bwh.harvard.edu
Hi Anne, Beth, Thanks much for the info. I was wondering about PRLR gene but it’s really good that you pointed out the suspected associated traits. I focussed at one of those traits, Cells_Intensity_MeanIntensityEdge_DNA. I have put images of 3 cell lines carrying variant in PRLR gene (a slide attached), which look fine to me, but would like to know your thoughts. Also, other traits having correlation >0.9 are also related to the intensity of DNA in cell compartment (a screenshot in the slide).
From: Anne Carpenter anne@broadinstitute.org
Again, not sure how isolated cells were defined here, but I don’t see many in these images! Is the count of isolated cells low for these samples? I wonder if there is an image segmentation problem where the cytoplasm is dim and not recognized, such that the Cell outline is just at the same spot as the nucleus outline for these samples?--
From: Arora, Jatin jarora1@bwh.harvard.edu
Oh sorry, I forgot to answer this question: isolated cells are the ones without any neighbor.
From: Anne Carpenter anne@broadinstitute.org
Ok, I see zero such cells in those images so either the segmentation is bad or the results you are seeing are based on a tiny number of cells per sample and are thus unreliable I don’t know if we have an easy way to view the segmentation results but Cell counts should be easy to check.
From: Arora, Jatin jarora1@bwh.harvard.edu
Sure, indeed I should have noticed that I'm looking at isolate cells and such cells are not visible in the images before bugging you. Sorry about that.I agree that potential reason would be a few number of isolated cells in those cell lines. I will check the actual count. Thanks so much.
From: Anne Carpenter anne@broadinstitute.org
No problem, this is why we work in teams with different expertise! We are very familiar with how cells try to trick us
From: Arora, Jatin jarora1@bwh.harvard.edu
Thanks much. I looked at the number of isolate cells in cell lines having variant in PRLR gene and larger value of Cells_Intensity_MeanIntensityEdge_DNA. It seems those cell lines have low number of isolate cells, but not that low. (a slide attached here)
I was also wondering, as a cell should not have DNA staining at its outer edge, they why (at first place) there exist such traits like Cells_Intensity_MeanIntensityEdge_DNA and others in CellProfiler. If I am missing something, please excuse me for my naiveness.
From: Anne Carpenter anne@broadinstitute.org
In that case, my guess is that the segmentation didn’t work well for these samples: still might be an interesting phenotype because the stain used to find Cell edges may be much dimmer than usual for these samples causing poor segmentation. The measurement modules simply measure all the features for all the stains for all compartments (even if, as you say, some don’t make much sense!)
From: Anne Carpenter anne@broadinstitute.org
I took a look about the gene itself and there really are no reported interesting Cell phenotypes in GeneCard. The most specific thing I can find is proliferation, but Cell growth is itself a pretty vague phenotype so that’s not much to go on as far as stories go. I guess that’s good news because it might be a totally novel discovery but it does make it harder to know what’s going on! https://www.genecards.org/cgi-bin/carddisp.pl?gene=PRLR
From: Arora, Jatin jarora1@bwh.harvard.edu
Thanks so much. Yeah, i was also not sure about how i can make sense of PRLR’s associations, but I can dig more and we can discuss in up-coming v2f meeting. Best wishes.