NetNotes


lustration of why MIPs are dangerous. Te percent overlap based on a simple threshold followed by simple AND math is very easy to achieve in ImageJ and other soſtware. Including removing fea- tures such as stained nuclei from the analysis. Michael Cammer michael.cammer@med.nyu.edu


A MIP is a worst-case scenario for quantitation since it deletes


data, and it removes a dimension that may have useful informa- tion in it. As long as background subtraction and noise correction (that is, with deconvolution) are properly performed, an average or sum projection will at least do the job non-destructively. Hon- estly, if people are going to take a Z-stack and collapse it, then they might as well do the same thing in one shot on a wide-field with a lower NA objective if they can get enough lateral resolution. Add- ing a dimension can add power, but it adds complexity and makes segmentation more challenging. It’s not bad to start with a 2D plan and add the third dimension when the experiment needs it. On the other hand, you can’t just use the heuristic of asking whether 2D analysis returns a clear difference between control and experi- mental data! Tat skips the important step where experimental ar- tifacts created by the projection process are determined. Take, for example, a treatment that makes epithelial cells taller or denser, that is, by rounding up. Taller cells mean more information gets compressed or lost in a Z- projection. Taller cells compressed into a MIP will have a higher average Pearson’s or Mander’s score, even if the ‘real’ proximity of proteins was not changed at all. Doing a 2D analysis without considering artifacts can lead to embarrassing reversals that most investigators want to avoid. Before going to 3D, I’d also ask whether a system has enough Z resolution to make a useful distinction! If you picture the ‘fried egg’ profile of a cell on a flat coverslip, even a confocal with a 1.4 NA objective has a hard time telling the upper and lower cell membranes apart from the cy- toplasm in between them (aside from very close to the nucleus). Su- per resolution techniques (or TIRF) can do better, but it’s important to have a rationale for when to put in the extra time for 3D, or add even more extra time (and data volume!) and step up to super res. Timothy Feinstein tim.feinstein@gmail.com


For fun, I went ahead and made an ImageJ macro that can


demonstrate the difference between 2D and 3D colocalization, as well as the impact of object size on the degree of colocalization: https://www.youtube.com/watch?v=DjjL2EZxWCk. Te raw video can be found here: https://bit.ly/3uABSd7. Te macro used to create the movie can be found here: https://github. com/Llamero/Colocalization_demo_macro. In the macro, the size of the spheres and the degree of overlap can be adjusted to demonstrate different concepts and issues. Te macro requires the 3D library and TransformJ library to run. Ben Smith benjamin.smith@berkeley.edu


Tis is a fantastically simple and effective demonstration. I


appreciate the examples. Tese are good analogies to help con- vince the lab that using MIPs is completely inappropriate for their experiments and what they want to quantify (pre-synaptic and post-synaptic colocalization in 3D tissue). Not to mention the fact that they have the confocal pinhole open to 2X Airy and step size at 3x Nyquist. And yes, we are scheduling a lab-wide image analy- sis tutorial consultation soon. On a side note, a post-doc showed me the published reference where he obtained these specific im- aging parameters and colocalization quantitation workflow. So, I’m fighting against an already published paper. Kathryn Spencer kspencer@scripps.edu


62


Quantifying colocalization is a minefield. MIPs are clearly


nonsense and will change as the thickness of the Z-series increases. 1) Tere are too many colocalization coefficients and some are almost meaningless. In a recent article we made a case for aban- doning one group of coefficients. Te premise is that colocaliza- tion coefficients fall into two useful groups, those that measure co-occurrence, the degree to which molecules are found in the same place, and correlation, the intensity relationship when fluo- rophores found together. Both measures are informative. However, this scheme exposes a third group that combines the two types of measurements into an unintelligible mess that we propose dumping as the measurement can arise from widely differing distributions (https://doi.org/10.1002/cyto.a.24336). 2) A second serious prob- lem of noise arises with point scanning images as two images taken consecutively of the same fluorophore are not measured as being perfectly correlated. So, if two nominally identical images don’t cor- relate perfectly then correlations from two different fluorophores will not be measured correctly. We demonstrate a practicable so- lution in https://onlinelibrary. wiley .com/doi/abs/10.1111/j.1365- 2818.2008.01967.x. 3) Expressed proteins with a fluorophore can cause problems. Colocalization can be measured accurately, but the levels of expressed protein are usually superimposed on the en- dogenous protein and will distribute differently. Binding sites are saturable. Overall, there are serious problems to resolve including segmentation. Jeremy Adler jeremy.adler@igp.uu.se


I would support the fact that colocalization of MIPs is not


appropriate. I would add that before doing colocalization stud- ies, deconvolution should be applied to the volume of interest. Louis Villeneuve louis.villeneuve@icm-mhi.org


Everything shared by Jeremy provides a great overview of the


problem. Just to add, there are a bunch of resources on the Imaris website which try to make this topic as easy as possible to follow. Most recently, our American (East) support team member Mat- thew Gastinger did a great webinar on how to use Imaris to look at colocation and colocalization (both voxel and object based). You can find that here https://imaris.oxinst.com/learning/view/ article/various-ways-of-solving-the-colocalization-problem-in- image-analysis. It also sounds as if you need an example to get the message across to users that may need help understanding the fundamentals. If they are having a tough time understanding, I would use a picture taken from an Ames Room Model. While not particularly scientific in the biological sense, it is a fairly useful tool to explain how you can’t trust perspective in 2D. I previously used it in presentations when I was a post doc to explain why 3D image analysis was important. Nick Jones n.jones@bitplane.com


My go-to illustration of the unsuitability of MIPs, and the


effect of cell geometry and anisotropic voxels in colocalization analysis, is figure 2 in this article from 2008: https://pubmed.ncbi .nlm.nih.gov/18353895/. Chris Wood chris@ibt.unam.mx


As the discussion is spinning a bit off topic, let me contrib-


ute. Tere are far worse things than MIPs. I oſten encounter co- localization within the nucleus (transcription, translation, repair, etc.), see, for example, https://drive.google.com/file/d/1n91JKGr hVgN3YHW9luAbVUpAHINNRljg. While Pearson coefficients are one of my favorites (insensitive to gain, offset; no need for thresholding), the dark pixels do indeed count! What’s shown in the snapshot cited above is while the two labels (green and red) hardly colocalize in the nucleus (Pearson=0.19), when other


www.microscopy-today.com • 2022 July


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68  |  Page 69  |  Page 70  |  Page 71  |  Page 72