Quantifying cellular microenvironments with spatial analysis

The rapidly evolving field of multiplexed imaging is revealing increasingly complex patterns of cellular positioning and cell-cell interactions with important roles in both cellular- and tissue-level physiology.

In this on-demand webinar, Dr. Michael Gerner, assistant professor at University of Washington, and Dr. Caleb Stolzfus, postdoctoral scholar at University of Washington, describe a spatial analysis toolbox capable of leveraging these information-rich datasets: the histo-cytometric multidimensional analysis pipeline (CytoMAP). CytoMAP incorporates multiple approaches for data clustering, positional correlation, dimensionality reduction, and 2D/3D region reconstruction to reveal features of cellular heterogeneity, quantify patterning of these cells across tissues, and identify large-scale principles of tissue organization.

In the first half of the webinar, Gerner and Stolzfus introduce the tools available in CytoMAP for high-dimensional image analysis and demonstrate the utility of CytoMAP in studying the microanatomy of phenotypically complex immune cell subsets in lymph nodes, organs with intricate cellular spatial patterning. In the second half, Gerner and Stolzfus present a live demonstration of the analysis workflow showing how CytoMAP can be used to understand tissue structure and reveal features of tissue organization in spatially resolved datasets.

Read on for highlights from the live Q&A session or register to watch the webinar at a time that suits you.

Q: During the raster-scan step in CytoMAP, how do you decide on the radius of the neighborhoods?

CS: This is a question we get quite often, and it depends on what you are looking at. If you are trying to define just big features of a lymph node or a tumor. On the slide, I wanted to differentiate between the T cell zone, which has all the red cells in the middle, and the B cell follicle on the outside. If I make my radius big, then I can get those features, but the border is blocking. If I make my radius too small, for example, 15 micrometers, then I start to pick up way too much detail, like these blood vessels and all these other sub-features.

Playing around with the size of that radius and trying out different types will help to answer the question you are looking for. Generally, if you look into the literature, you can see how chemokines diffuse and how cells communicate, somewhere between 30 and 100 micrometers is usually a pretty good distance for those neighborhoods.

Q: How do you decide on the number of clusters or regions?

CS: You can have as many unique neighborhoods as you like and that's as many different region types as you have in your tissue. You must think about what are you trying to answer biologically? Are you looking for all the diversity that is in your sample? Then, you are going to want to do a lot of different regions. If you are looking for just those very broad features, then you only want to do a couple of regions. I have some algorithms built into CytoMAP, which do statistics like what is the standard deviation within a group versus the standard deviation across the whole data set and try to minimize functions like that. Each algorithm performs different tasks to try to guess the number of regions. So, it really depends on the question you are asking.

Q: What is the limit of the number of antigens that can be analyzed by CytoMAP?

CS: The most I have loaded in was a sequencing data set, where I had 10,000 parameters that were antigens at that point, but the problem is all these interfaces are made for 100 or 200 antigens at most. When you start getting beyond that 100 limit, it gets tedious to interact with the data and turn those individual channels on and off. The real answer is it depends on how patient you are, but you can load in a lot of parameters for yourselves.

Q: Are there any limitations to the quantitative analyses, and can subcellular, submicron nanoscale structures be resolved if we can visualize with standard confocal?

CS: It has been something I have been wanting to try for a while, and I have not had the chance to try it yet. It should work if you segment intracellular components and CytoMAP does not know what you are loading into it. It just sees them as objects, therefore, you should be able to load those subcellular components into CytoMAP and do the same spatial statistics. I have not tried it, so I do not know what assumptions might run afoul on that subnanometer level, but it should work.

To find out more about how CytoMAP can accelerate your spatial analysis research, register for the on-demand webinar here>>