Spindle: How to handle cases with unusual DNA shortest axis?
Closed this issue · 9 comments
Depending on the experiment or biological question, users might not include sufficient slices to cover the entire z range of the chromatin plate.
https://www.dropbox.com/s/5a6jietubu8gkgr/Composite-01-NSP-01.tif.zip?dl=0
Original version, bright neighbour cells mess with initial DNA segmentation...
https://www.dropbox.com/s/r7990mbb9nz15b7/Composite_01-NSP-01_Clearoutside.tif.zip?dl=0
Same sample, but xy-cropped mitotic cell to exclude others.
The cell seems to have incomplete chromatin plate coverage, so the expected shortest DNA axis is not found correctly because the entire DNA object looks not as expected.
Would it make sense to pre-confine possible directions of shortest DNA axis to a "horizontal" range? Or should these cases be handled differently? Or maybe not at all?
Talking to the person providing us with the samples, it became clear that this is actually a physiological phenotype ("butterfly chromatin").
A proposal: the user should have the option whether to use shortest DNA axis as starting point (=default) or find the axis independent of the DNA signal via the tubulin channel?
The original idea was to avoid the assumption, that in the tubulin signal, the longest axis would always be the spindle axis (pole-to-pole axis), but maybe it's a fair alternative to explore if DNA shape is too weird?
The files you sent still are cropped:
I would say:
- If it really is a biological phenotype it is interesting to look into it!
- If the data is not fully acquired in 3D (like in above example) it is out of scope for our analysis.
Thus, to look further into it I would ask your collaborators to acquire the cell fully.
From what I've learnt, this particular sample is both: The chromatin plate indeed is in this unusual but phenotypical configuration, plus unfortunately not fully acquired in 3D - which might or might not have helped in this case.
I remember that we often had a similar problem in fixed cells, where the chromatin plate gets deformed quite a bit. On top of that, many users would be interested in measuring morphometry in Xenopus spindles, where the DNA can adopt weird shapes and the samples are often "squash-fixed" and thus lose three-dimensionality:
https://www.dropbox.com/s/hrpf60mst8q8x3s/xenopus_019.tif?dl=0
I think we could still get a meaningful spindle lateral and axial extent and so forth from these, if the right axis was found in the beginning, no?
So considering that it might be useful for the above cases and more, I think it's worth looking into it.
Off topic, but I'm wondering why initial DNA segmentation goes wrong in the "Composite-01-NSP-01" sample, because the signal seems quite nice and the mitotic DNA is spatially separated from the surrounding cells?
ok, I get your point.
However, currently I do not have the resources to develop a "new analysis".
I think I have to focus on getting stable (and published) what we currently have, before branching off to accommodate new imaging and sample preparation modalities.
@TobiasKletter
I think in such cases that are not really 3D maybe a 2D analysis could be an option? Somehow I feel doing 3D analysis on data that is not really fully 3D is a bit of a mess 😏
What do you think?
Yes this would be useful for the squashed Xenopus samples 👍
ok! are you motivated to translate my 3D Java code to a 2D python code? 🚀 😄
Motivated yes but this could take a couple more pandemic lockdowns :D
I am also motivated, but don't see either when I would find the time...(for doing it in Java)