A Cell Neighborhood can be defined as a region within a tissue with a set of characteristic cell type enrichments, which may have a unique role within a tissue microenvironment (example being tertiary lymphoid structures).
To define Cell Neighborhoods, Enable uses a k-nearest-neighbors approach, followed by k-means clustering. For each annotated cell in your entire dataset, the algorithm identifies the cell types of its k-nearest-neighbors in x/y coordinates. It’s often appropriate to start with 10 nearest neighbors, but this value can be adjusted. Each cell’s local neighborhood is defined by the cell types that are its neighbors. In order to identify which collections of cell types constitute a Cell Neighborhood in the tissue, we cluster the values of all the cells’ local neighborhoods using k-means clustering. For the first pass here, it’s often good to use 10 neighborhoods. After the first pass is done, try tuning these parameters until you get interpretable Cell Neighborhoods. NOTE: it is usually advisable to settle on your Cell Neighborhood definitions before performing statistical analyses. Doing otherwise could compromise the integrity of your analysis.
To see which cell types make up each Cell Neighborhood, and annotate the Cell Neighborhoods, open the Cell Neighborhood extension. This is located in the Analysis Toolkit.
You can generate new Cell Neighborhoods by clicking the New Run button. Here, you can configure the cell type classification as well as the number of neighbors to consider in the k-nearest-neighbors step and the number of neighborhoods to identify in the k-means clustering step. You can also assign a name and description to your new set of Cell Neighborhood annotations.
Selecting Regions
We also provide the ability to select specific regions for analysis in case you want to focus your analysis on a subset of your study.
In these circumstances you can similarly use the Regions dropdown to select each as an input of region(s) you wish to use for your analysis.
Additionally, same as the Visualizer’s Region selection dropdown, you can use the Regions Filter to limit the region options for your analysis based on metadata traits uploaded in the designer.
If your neighborhood run requires the results of a previous analysis, we will notify you if there are regions that are missing these previous results and what that means for your analysis.
Once you’re happy with the configuration, you can click Create to kick off the computational pipeline. This process typically takes a few hours, depending on the size of your data set.
Once your computation is completed, return to the Cell Neighborhood extension and select a run to view the full results.
The Cell Neighborhoods generated in the study are represented as a heatmap, where each row represents a Cell Neighborhood label, and the columns represent annotated cell types. The heatmap is colored by normalized frequency of these cell types in each Cell Neighborhood. Using this heatmap, you can rename the Cell Neighborhood labels as you see fit. For example, Neighborhood 8 has the highest proportion of epithelial cells, and might appropriately be labeled “Epithelium”.
Once the Cell Neighborhoods are labeled, click on Update Labels to save. Now, if you go back to the Explorer tab, the Cell Neighborhood analysis will show updated label names.
Return to Home Directory